+ All Categories
Home > Documents > Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features...

Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features...

Date post: 21-Mar-2020
Category:
Upload: others
View: 1 times
Download: 0 times
Share this document with a friend
47
Trend Analysis Report AI AND ITS PIVOTAL ROLE IN TRANSFORMING OPERATIONS Author: Mark Newman, Chief Analyst Editor: Dawn Bushaus, Managing Editor December 2018 inform.tmforum.org Sponsored by:
Transcript
Page 1: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

Trend Analysis Report

AI AND ITS PIVOTAL ROLE IN TRANSFORMINGOPERATIONSAuthor: Mark Newman, Chief AnalystEditor: Dawn Bushaus, Managing EditorDecember 2018

inform.tmforum.org

Sponsored by:

Page 2: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 2

Section 1

e respondents speakWhere are CSPs on their journeys?

Contents

10Section 2: Therespondents speak –AIOps and opportunites

inform.tmforum.org

03The big picture

05Section 1: Analytics,algorithms and operations

25Section 5: Make ithappen – Strategies forimplementing AIOps

17Section 3: What are thechallenges toimplementing AIOps?

22Section 4: Developing andfinding AIOps expertiseand solutions

40TM Forum toolkit fordigital transformation

46Research & Media Team

42TM Forum Frameworx

27Additional features andresources

Page 3: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

In researching this report, wesurveyed 65 executives from 37different global, regional or nationalCSP operating companies in 25countries, and 48 suppliers from 33unique companies. We alsoconducted in-depth interviews withoperators and suppliers.

AI deployment getsunderwayNearly 70% of CSP respondents saidthey are either deploying AI in someparts of the business or testing it inproofs of concept and trials.Operators are deploying thetechnology on their own and in

collaboration with suppliers. Eitherway, the use cases typically focus onimproving how CSPs operate now.

However, as networks become morecomplex, the IT systems supportingthem also must evolve toaccommodate future demands.Networks and IT must be prepared toaccommodate billions of connecteddevices and waves of newtechnology, the most prominent being5G. CSPs recognize that AIOps will bethe only way to scale operations andmake them sufficiently fast andresponsive to profoundly changednetworks and businesses as digitaltransformation progresses.

As José Manuel de Arce, DeputyDirector OSS/BSS Infrastructure,WorkSpace and OSS Technology atTelefónica International WholesaleServices, puts it:

inform.tmforum.org 3

e big picture

A year ago, we published our first report on artificial intelligence, AI: The time is now. The titlereflected the urgency and excitement surrounding the potential of AI, and the report delved intothe most obvious application for it: improving customer experience. This year the focus andmood have shifted. Much of the enthusiasm has been replaced by caution and indecision ascommunications service providers (CSPs) grapple with the many difficulties of applying AI tooperations (AIOps). Still, they are beginning to make progress.

TM Forum, 2018

Who are the AIOps survey respondents?

Mobile operators23%Converged operators66%

1.5%

8%65

Data center/cloud operators

respondents from 37 unique companies 48 respondents from

33 unique companies

Fixed operators

CSPs Suppliers

16%BSS/OSS24%

20%Systems integrators

Consultants22%

AI is about working withdata and doing with itwhat humans would do,without the errors andfaster.

Page 4: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

e big picture

inform.tmforum.org 4

Caution and distrustMany CSPs seem uncertain aboutwhere and how to start with AIOps,and some are not convinced by what’son offer from suppliers. There isconsiderable tension and distrustbetween CSPs and establishedequipment providers in particular,fueled by operators’ great fear oflosing control of their networks anddata if they no longer understandwhat decisions AI is making (or why)in a supplier’s black box. Also, themore processes an AI solutiontouches, the more valuable andefficient it should become. By thesame token though, any faults causedby AI will spread rapidly, so isolatingthem quickly will be critical. How todo that is adding to CSPs’ longer-termconcerns.

Our research found that CSPs seehaving internal AI and analyticsexpertise as strategically important.Many are working to acquire anddevelop those skills and knowledge invarious ways, yet few appear to haveany kind of structured approacheither to implementing AI orcoordinating their initiatives to avoidduplicating effort and expense, andslowing progress.

Another big issue is that AI needsmassive amounts of clean data, butthe data CSPs have, which ismultiplying all the time, resides insilos and in formats that are notcompatible, consistent or easilyaccessible despite years of grapplingwith this issue. Lack of standardizeddata models and other universalapproaches are also blockingprogress, but CSPs have recognizedthis, and individual companies andindustry bodies are acting to addressit.

Go slow to go fastA lot is happening on many fronts,and while the progress of AIOpsseems piecemeal and slow, theapproach is arguably a perfectexample of Amara’s ‘Law’: Back in the1960s, Stanford University computerscientist Ray Amara said we tend tooverestimate the impact of a newtechnology in the short run butunderestimate it in the long run. Or toquote from a recent presentation byJerome Katz, Vice President ofService Provider, CustomerExperience at Cisco:

Read this report to understand:

n Which operational changes CSPsare undertaking that require AI

n The types of AI technologyoperators are using in operationsand for what purposes

n How reliant operationaltransformation is on AI

n The use cases driving AIOpsdeployment, including which ofthem operators are implementingtoday versus in the future

n Issues and gaps CSPs are finding intheir attempts to harness AI

Go slow to be able to gofast: You need to plan,and [it’s] critical to get[AI] aligned withbusiness processes,operational structure,organizational alignmentand skills.

Page 5: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 5

Section 1

Analytics, algorithmsand operations

Analytics and AIThe term ‘AI’ is often used to meanapplying analytics, but the use ofanalytics alone does not necessarilyinvolve AI, and indeed in many waysthis is at the heart of the issueregarding operations. If analyticsmonitor data looking for pre-definedpatterns and anomalies withoutapplying intelligence – that is, withoutevolving how or what it monitors, ordrawing more advanced conclusionsfrom what it has ‘learned’ – that is notAI. Analytics on their own cancontribute to better decision-making,but the decisions must be made andapplied by humans.

The infographic below defines manyof the terms that appear in thisreport, explaining how they’re relatedbut different.

What is AI – and what’s not? Analytics – monitoring datato look for patterns and

anomalies (without applyingintelligence) and applying thosepatterns towards effective decision-making

Artificial intelligence (AI) –the development of computer

systems capable of performing tasksthat normally require humanintelligence; this includes visual

perception (such as is required forself-driving cars), speech recognition(for example, Alexa), decision-making,and translation between languages(think, Google Translate)

Automation – withintelecoms this means

automation of processes that werepreviously carried out by people; AI isan enabling technology that may (ormay not) help with the process ofautomation

There is considerable confusion inside and outside telecommunications about what AI is, whywe need it, how to implement it and what impact it will have – never mind how to apply it toCSPs’ networks and operational and business support systems (OSS/BSS). We do not have fullanswers to many of these questions, but understanding what AI is (and is not) as well as why itis becoming essential to operations (AIOps) is a good start in the journey towards reaping thepromised benefits.

30%Reduc�on in mobile

infrastructure spending

$9 billionIn opera�ng profit from reduced frequency and

dura�on of network outages

$46 billionSavings in customer

acquisi�on costs and lost revenue through be er network performance

$14 billionGenerated from the sale of

self-organizing network solu�ons

30,000 tonnesIn CO2 emissions saved globally because of fewer field visits

$27 billionCumula�ve cost savings for the telecom industry over the coming decade

$12 billionIn consumer benefits from CSPs passing network cost savings on

2.5 billion hoursSaved and $3.8 billion in produc�vity gains through reduc�on in dropped calls

AI use cases for network and service management

TM Forum, 2018 (source for data: World Economic Forum)

CO₂

Page 6: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

Section 1

inform.tmforum.org 6

Machine learning – a type ofAI that gives machines the

ability to learn automatically andimprove from experience withoutbeing explicitly programmed

Deep learning – takesmachine learning further by

processing information in layers,where the result or output from onelayer becomes the input for the next

Cognitive computing – like AI,cognitive computing is based

on the ability of machines to sense,reason, act and adapt based onlearned experience, but whereas AIacts on its analysis to complete a task,cognitive computing provides theinformation to help a person decide

Robotic process automation– a software tool that allows

people to configure robots (computersoftware) to perform rules-basedtasks such as accessing programs andsystems, performing calculations,creating reports and checking files;within telecoms it can be particularlyuseful for processes that havepredictable and frequent interactionswith multiple applications

Decision managementengine – runs a process or

set of processes to improve andstreamline action items in businessprocesses and customer-facingapplications

Virtual agents – animatedcharacters (usually in an

appealing form) that act as onlineservice representatives and can have‘intelligent’ conversations with users,answering their questions

Self-organizing networks(SON) – a technology for

automating the planning,configuration, management,optimization and healing of mobileradio access networks; it wasdeveloped by 3GPP and is sometimesconflated with AI

Analytics for operationsWe asked CSP survey respondentshow they plan to use analytics inoperations over the next year, and alarge majority said they intend to usethem for customer care.

Customer experience is frequentlycalled out by operators as the biggestreason for digital transformation, andcustomer care is a key aspect. It isalso one of the most heavily criticizedelements of customers’ experiences indealing with operators, even though itremains primarily call center-basedand because of that is extremelyexpensive. Inexplicably, CSPs oftenview improving customer care asdelivering a better call centerexperience – when was the last timeyou called Amazon?

Read this report formore on improvingdigital customerexperience:

CSPs are under pressure to improvecustomer care while cutting costs, andusing analytics to do this is key.Customer experience is also behindthe second most likely application foranalytics in operations in the nextyear – predictive maintenance.

As Dave Salam, Director of Core andData Analytics, BT/EE, was quotedsaying in his keynote at a CambridgeWireless event in the UK inSeptember:

With predictive maintenance, the ideais to be able to ‘fix’ network elementsbefore they are broken by knowingwhen they are due for routinemaintenance and updating, ratherthan responding when customersstart to complain. Predictivemaintenance can also minimize truckrolls.

Verizon Communications has set upan internal team in its wirelinebusiness to deploy analytics, machinelearning and other types of AI in thefuture. Its first priority, though, is toaddress network problems and reducecustomers’ complaints.

Matt Tegerdine, Director of NetworkPerformance Analytics, Verizon,pointed out in a recent TM Forumreport, Building a data lake to drivedigital transformation, that the payoffsfor that project are multiple: Fasterresolution of problems leads to bettercustomer experience, and solving theproblem the first time around alsoreduces costs.

Where will CSPs use analy�cs?

TM Forum, 2018

Customer care Radio access network opera ons

Predic ve maintenance

Cybersecurity

81% 20% 61% 34%

When customers aretelling you there is aproblem, it’s alreadytoo late.

Page 7: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

Section 1

inform.tmforum.org 7

Securing the networkAbout a third of respondents plan touse analytics for cybersecurityoperations. This is an increasinglydifficult challenge to address becausethe ‘attack surface’ gets bigger as thenumber of network-connecteddevices and interdependenciesbetween them grow. For example, 5Gwill be a network of networks, andoperators are keen to pushinteroperability up the stack tosupport enterprise applications (formore about this see page 45).

Radio access network (RAN)operations is the next priority foranalytics. Again, 5G will make RANoperations more complex because itneeds forests of tiny antennas toexploit the very short wavelengths itwill run on. 5G coverage and capacitywill be provided through tightintegration with other kinds ofnetworks such as Wi-Fi and 4G, andincreasingly, RAN technology will bedeployed that can be configured andupdated remotely, with self-organizing and self-healingcapabilities.

Dr. Lester Thomas, Chief IT SystemsArchitect, Vodafone Group, notes thathis company is deploying self-organizing networks in ‘emerging’markets (the first one was Egypt)where electricity supplies can beunreliable, so sometimes cells gooffline then have to optimizethemselves. Using big data analytics,the surrounding sites reconfigurethemselves to accommodate thechanges. Thomas observes, “It’s notlike static radio planning,” adding thatVodafone built the autonomousapplication itself using open sourcetechnology.

Of course, data and analytics are twoof the basic ingredients that fuel AIwhich ‘learns’ from and builds on

processing massive amounts of data,in real time. Although operatorsgenerally have made little progress inharnessing ‘big data’ and analyticsdespite all the hype, they are nowlooking to AI to help them becomedata-driven organizations, which is amajor theme of this report.

Automation and AISometimes the terms ‘automation’and ‘AI’ are used interchangeably,which is misleading. CSPs areautomating their networks because asthey become increasingly complex, itis beyond humans’ capabilities tomanage them in terms of scope, scaleand speed. EE/BT’s Salam explainedwhy EE started investing heavily inautomation several years ago:

EE’s reasons for transforming itsoperations through automationinclude:

n An exponential rise in the numberof devices connected to theirnetworks, driven by IoT

n Need for greater capacity onnetworks as the number ofsimultaneous connections in manylocations soars

n Requirement for increased speed asuse cases such as autonomousvehicles need a return path with alatency of 1 millisecond or less

n Continuing, staggering growth ofmobile and video traffic

n Changing traffic patterns, which aredemanding more flexible andefficient use of assets and arechanging how networks aredesigned and run (EE is moving to adistributed, cloud-basedarchitecture with more processinghappening the edge of networks toreduce latency and minimize thetraffic traversing the network)

n Need to accommodate 5G, a cloud-native technology that will provideseamless coverage through anecosystem of networks, such as Wi-Fi and 4G, and those of thirdparties, which creates a far largernumber of interdependencies

n More interdependencies make itharder to identify the source ofproblems and secure the network,the traffic it carries and devicesconnected to it

In short, the insatiable demand forgreater bandwidth capacity, coverageand speed, coupled with theproliferation of new services meansnetworks and the associated ITsystems are increasing in complexity.Over the next few years as CSPsmove networks and operations to thecloud, the age-old division betweennetwork and IT divisions within CSPorganizations will disappear assoftware-controlled, automatedoperations become the norm.

It became apparent thatyou can’t run a networkwhen…people reportingthings on Twitter is amuch quicker way to spotissues in the networkthan you can yourself.

Page 8: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

Section 1

inform.tmforum.org 8

What’s drivingautomation The graphic below shows where andwhy CSP respondents are looking toimplement automation over the nextyear.

With close to half of CSPs pickingnew services as a driver, it seemslikely that the imminent arrival of 5Gand the complexities it brings are onoperators’ minds. Even so, newservices like microservices andnetwork slicing are a considerableway off and unlikely to becomemainstream for up to five years.

In the near term, 5G will be aboutboosting fixed wireless connectivityand providing better mobilebroadband coverage, speed andcapacity.

It’s also intriguing that changingprocesses and settings is the second-place driver by such a margin –slightly more than 14% behind newservices. Arguably this should be aspressing as introducing new services,but operators are more used tocoping with incremental change.

Slow networktransformation The relatively low percentage of CSPspicking new technology likevirtualization as a driver forautomation reinforces the results ofour latest Digital TransformationTracker survey, DTT 3: Why is networktransformation so difficult?, whichshowed that almost half of CSPrespondents and two thirds ofsupplier respondents believe CSPshave not established a business casefor network virtualization.

Network functions virtualization(NFV) and software-definednetworking (SDN) have stalled andslipped down CSPs’ agendas,although operators know thateventually their infrastructure willbecome virtualized and cloud-based.Many blame the lack of progress onsuppliers that were slow to developproducts and solutions to implementNFV, as they saw little benefit indoing so according to their critics,preferring to leverage their existingportfolio. Operators are also awarethat unless their networks areautomated, the cost of implementingNFV and SDN will be prohibitivelyexpensive.

There is another big factor foroperators to take into account too:Automation as they have deployed itso far cannot meet the operationalchallenges that are hurtling towardsthem because it cannot scale. Todayautomation typically relies on static,network operations center-basedanalytics to manage repetitive tasks,which do the same thing endlesslyunless a human makes changes,having of course first worked outwhat changes are needed – no smallfeat in such a complex environment.AIOps are far more suited to goal orintent-based management, as weexplore in Section 3.

Machine learning isnecessary In order to scale their operations,CSPs must apply machine learning inconjunction with automation.Machine learning is a type of AI, andit is currently the most common typeof AI in use, although this is expectedto change in the next two years (seepage 14). Where machine learning isin use now and how it will besuperseded varies considerablyamong operations domains.

Machine learning works by ‘studying’the algorithms and mathematicalmodels that computer systems use tobuild and constantly refine how theyperform tasks without specific codingfor those changes. This and otherkinds of AI present an alternativeapproach to static, rules-basedautomation because they massivelyreduce the human effort needed togenerate the rules that enableautomation and are needed to devisebetter rules over time.

Drivers for introducing automa�on into opera�ons

20.5%

3%

28%

6.5%

42%

We are introducing new technology (e.g. virtualiza�on) that necessitates opera�ons at speeds not feasibile for manual processes

We need to change processes or se ngs (e.g. provisioning) more o!en and more quickly than manual processes allow

We have new process requirements that are more complicated and change more o!en (e.g. partner onboarding for IoT) than manual processes can cope with

We would like to introduce new services that require faster and more complicated responses from opera�ons (e.g. on-demand services) than manual processes can provide

Other

TM Forum, 2018

Page 9: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

Section 1

inform.tmforum.org 9

In particular, as networks and theecosystems they operate withinevolve, AIOps will have a key part toplay in helping us understand andmanage interdependencies. AsEE/BT’s Salam, puts it:

The path to making the fullest use ofAIOps is not clear. Perhaps thebiggest thing to grasp is that applyingAI to operations – with the ultimategoal of gaining autonomous control ofthe network – is not just a question

of applying technology. ImplementingAI demands a whole lot of otherchanges too, which on top ofeverything else CSPs must address aspart of their transformation efforts, isa lot to handle.

In the next section, we’ll look atwhere and how CSPs are deployingAI.

We need proactive fault management, anomaly detectionfor real-time network management. As we build in moreintelligence, we can optimize it, tune it in real time andultimately, for 5G and IoT, autonomous control will beessential to underpin, design and run operations…becausethe level of complexity you will get to will be notunderstandable by operations now – assistive intelligence isneeded

Page 10: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 10

Section 2

e respondents speakAIOps and opportunities

Who?For this report we surveyed 65 executives from 37 different global, regional or national CSP operating companies in 25countries, and 48 executives from 33 supplier companies. We also conducted in-depth interviews with operators andsuppliers.

5%North America

15%La n America/

Caribbean

43%Europe and/or Russia

8%Middle East and/or Africa

18%Asia-Pacific

11%Global

Loca�on of CSP

8%Fixed operator

1.5%Data center/cloud provider

66%Converged operator (some combina on of mobile, fixed voice and data, and TV)

23%Mobile operator

1.5%Other

Type of CSP

Page 11: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

Section 2

inform.tmforum.org 11

29%Fewer than

5 million

Number of subscribers

0%100 million to

150 million

12%25 million to

50 million

31%5 million to

25 million

11%50 million to

100 million

17%More than

150 million

Size of CSP

4%Hardware supplier

16%Network func on so!ware supplier

24%OSS/BSS supplier

20%Systems

integrator

22%Consultant

4%AI specialist

10%Other

Type of supplier

Page 12: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

Section 2

inform.tmforum.org 12

Gauging deployment ofAIOpsWe began our survey by askingcommunications service providers(CSPs) if they are deploying AI, and byasking suppliers for their view of theircustomers’ activity – or lack of it.There are some notable discrepanciesbetween their responses, especiallywhen it comes to how operators aredeploying AI, but it is at least

encouraging to see that a majority ofCSPs have embarked on some AIOpsactivities.

Vendors’ perception is that almost50% more CSPs are waiting to deployAI than the CSPs’ answers indicated,although admittedly the percentage inquestion is low. What’s moreinteresting is that vendors’ answerssuggest that about 30% more CSPsare working with them to build AI andproducts and services than the CSPs’

responses showed. CSPs’ repliesconcerning how many of them aredeveloping internal AI expertise ismore than double the numberprovided by the vendors.

We probed this further in interviewsand uncovered something of a chasmbetween vendors and their CSPscustomers regarding AI, which clearlyis having a profound effect on therate of deployment (see Section 4).

Are CSPs deploying AI?

CSPs Suppliers

TM Forum, 2018

11%

31%

20%

38%

We are taking a wait-and-see approach

We have started proofs of concept

We are working with suppliers who are building AI into some products and services

We have already built some internal AI exper se and we are incorpora ng it into our product and service roadmaps

15%

35%

19%

29%

2%

Customers are taking a wait-and-see approach

Customers have started proofs of concept

Customers are working with suppliers who are building AI into some products and services

Customers have already built some internal AI exper se and are incorpora ng it into their product and service roadmaps

Other

Page 13: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

Section 2

inform.tmforum.org 13

Where do CSPs needAIOps?It is perhaps not surprising thatimproving customer experience is thebiggest driver for AIOps (see graphicbelow). Interestingly, two thirds ofCSPs’ explanations in the ‘other’category were indirectly about bettercustomer service while the remainderconcerned cleaning databases. Buteven without adding those responsesinto the original totals, deliveringbetter customer experience is theclear winner by more than 15%.

Given the diverse range of drivers forAIOps mentioned by CSPrespondents, we delved deeper intothe numbers to see if their prioritiesare influenced by their size. Deliveringthe best customer experience is a toppriority among respondents from the

group of largest operators (whichhave more than 150 millionsubscribers), with more than 91% ofthem choosing it. This is 11% aheadof the next nearest groups which arefrom at the other end of the scaleregarding size – those with fewerthan 5 million customers (79%) andthose with 5 million to 25 million(80%).

The second most popular driver,reducing OpEx by using automationand closed loop systems, scored mosthighly among respondents from smallto mid-sized operators (those with 5-25 million and 25-50 millioncustomers). About 75% chose it asthe top driver. Interestingly, this wasthe least popular driver for largeoperators, with only 55% choosing itas a top driver.

Preventing failures and outages is apriority for operators with 25-50million subscribers, with half of themchoosing it. The lowest percentage(36%) was among the largestoperators, which perhaps indicatesthat they believe they have built moreresilience into their networks.

However, as we were completing thisreport, there were serious outages onthe O2 UK network and SoftBank’s inJapan due to software beingdecommissioned. Interestingly, onlyone operator chose the option ofidentifying services and timing fordecommissioning as a top driver forAIOps in our survey, and none of thevendor respondents cited it. Giventhe level of change that will be takingplace in networks over the nextcouple of years and beyond, theymight do well to revise their view.

Other

Iden�fying services and �ming for decommissioning

Suppor�ng decision making in opera�ons teams

Crea�ng new services or service flexibility (e.g. on-demand)

Improving revenue assurance

Iden�fying market trends and assis�ng in investment decisions

Crea�ng new products and services (e.g. automa�c voice transla�on as a service)

Upselling new products to exis�ng customers

Primary drivers for AI

77%

62%

45%

32%

28%

25%

18%

15%

15%

14%

5%

2%

Improving efficiency, agility and transparency

Predic�ng performance and preven�on of outages and failures

Reducing OpEx in the organiza�on by driving automa�on and closed loop systems

Delivering a be!er customer experience

TM Forum, 2018

Page 14: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

Section 2

inform.tmforum.org 14

Which AI use cases?We asked CSP respondents which AItechnologies they have alreadydeployed and in which operationaldomains, and which they intend todeploy in the next two years. Weranked the top four responses, whichaccount for the majority of activities.Use of machine learning in networkplanning and optimization is thebiggest use case today by far, but thatwill change over time.

As noted in Section 1, machinelearning is currently the mostcommonly used type of AI overall,with a strong presence across theboard. It is interesting to note though,that CSP respondents expect it largelyto be replaced by deep learning,which didn’t make it into the top fourAI technologies already deployed. Intwo years’ time CSPs expect roboticprocess automation to diminish,replaced in the top four by deeplearning.

Use of machine learning is alsoexpected to decline in revenuecreation and billing and assurance, aswell as in service creation andmanagement. The most dramaticchange predicted is the decline ofmachine learning for planning,optimization and management, from70% of AI tech deployed today to22% in two years’ time and the rise ofdeep learning. This is a reflection ofoperators needing more sophisticatedtools and insights as 5G and IoTdeployments proliferate.

The other notable predicted change isthe increasing use of decision

management engines. While itsprominence falls in customerexperience and remains stable innetwork planning, optimization andmanagement, it more than doubles inservice creation and management,and more than triples in revenuecreation, billing and assurance. This isbecause decision management

systems are designed to automateinteractions with customers,employees and suppliers. They areincreasingly popular in many sectorsto speed response times and makedecisions based on huge amounts ofinformation resulting from analysis ofhistorical behavioral data, previousdecisions and their outcomes.

Top uses of AI in opera�ons now and in 2 years

Top uses of AI in opera�ons now

Top uses of AI in opera�ons in 2 years

70% 6% 12% 6%

33% 10% 8% 23%

30% 9% 7% 23%

25% 33% 10% 4%

Network planning, op�miza�on and management

Service crea�on and management

Revenue crea�on, billing and assurance

Customer experience

22% 29% 5% 12%

13% 26% 13% 17%

21% 17% 11% 26%

24% 7% 20% 6%

Network planning, op�miza�on and management

Service crea�on and management

Revenue crea�on, billing and assurance

Customer experience

TM Forum, 2018

Machine learning pla orms Deep learning pla orms

Virtual agents Decision management engines

Page 15: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

Section 2

inform.tmforum.org 15

How are CSPsautomating AIOps?We also wanted to know how CSPsare using AI to automate operations.Almost half of respondents areautomating one process within asingle domain or business unit, afinding reinforced by our interviews.

The beauty of using AI to automate asingle process is that it is a goodlearning experience but with limitedramifications if it doesn’t go well. It’seasy to measure how successful it isand helps operators build confidencein the technology and approach, butmany small projects contribute tomainstream ops. Still, over a fifth ofrespondents told us they haveventured into automating multipleprocesses across multiple domains orbusiness units, and 20% have tackledmultiple processes within a singledomain or business unit.

Among operators, both multipleprocess approaches are more popularthan applying AI to a single process inmultiple domains or business units,which suggests a certain level ofconfidence and experimentation.After all, the greatest benefits will be

reaped by applying AI in operationsfrom end to end, and indeed will benecessary to reach the end-goal ofhaving an autonomously controllednetwork. Nevertheless, fear of AIfaults spreading and resulting inundesired and unforeseenconsequences is reasonable.

Using AI for end-to-endmanagementWhen CSPs were asked to name theareas where they are tapping analytics,customer care headed the pack by abig margin. This is encouraging becausecustomer experience, of whichcustomer care is a key part, is thebiggest driver of digital transformation.But CSPs must also slash costs, whichnecessitates using AI to improvenetwork and service management.

We asked CSP respondents whichmanagement use cases are a toppriority for deploying AI, and perhapsnot surprisingly, using it to handle bigdata and detect patterns within it,ranked highest.

So far, CSPs have not been able tomake good use of their enormousquantities of data with analytics alone,and the volume of data will increase byan order of magnitude due to 5G and

IoT applications in the near and distantfuture. Combining analytics,automation and AI can transformoperations, giving operators increasinglevels of autonomous control, butdeploying AI technologies will bedifficult in static, rules-based systems— they are far better suited to workingwith systems that are organized aroundgoals or intent, which looks at thenetwork in terms of what users want toachieve through services (see Section4).

Operators need clean, valid data sets,standard data formats and a universaldata model to fuel AI, plus AI models forspecific use cases and affordable,pragmatic ways, to work alongsidelegacy systems (see Section 3 for moreon data-handling standards and models).

Tied for second place are: using AI topredict and prevent networks outages,and to proactively improve customerexperience. As networks become morecomplex, the fear of outages risesbecause it is harder to understand causeand effect in such environments –indeed, it is one of the reasons anumber of operators are making greatstrides in dealing with the tens ofthousands of alarms they deal with dailyin their network operations centers (seepage 16).

How are CSPs automa�ng AIOps?

TM Forum, 2018

44%

21%

15%

20%

Automa ng a single process within a single domain or business unit

Automa ng mul ple processes within a single domain or business unit

Automa ng a single process across mul ple domains or business unit

Automa ng mul ple processes across mul ple domains or business units

Other

AI use cases for network and service management

68%

43%

43%

33%

8%

Trigger automated resolu on to self heal

Trend analysis to predict and proac vely prevent outages

AI technologies used by the network itself to automa cally and

proac vely provide the best customer experience

Big data handling and pa!ern detec on to iden fy issues and

inform reac ve ac on more quickly

TM Forum, 2018

Page 16: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

Section 2

inform.tmforum.org 16

Outages can cause financial lossesand reputational damage, regardlessof who or what is responsible. Forexample, an O2 UK outage in earlyDecember affected 30 millionindividual customers, which is nearlyhalf of the entire population of theUK. While 25 million of the customersaffected were O2 customers, 5 millionwere customers of MVNOs TescoMobile, Sky Mobile, GiffGaff andLycamobile whose businesses havesuffered too. In addition, O2 UK hasmany large corporate customers suchas Transport for London, for which itprovides live location data for 8,500buses in the city.

Using AI to proactively maintaincustomer experience is also key. AsBT/EE’s Dave Salam so eloquentlynoted in his comment about findingout about network problems onTwitter (see page 7), CSPs’ currentalarms, metrics and key performanceindicators fail to provide an accuratepicture of how customers areexperiencing services at any giventime because they are, by design,network-centric not service-centric.

Automated resolution to self-healnetworks also will be essentialbecause with so many dispersedelements, there will be no other viableor affordable way of fixing failures inthe network. As noted, truck rollsmust be avoided because they areexpensive and slow.

Early adopters’ successSeveral CSPs have made substantialprogress in applying AI in networkoperations centers (NOCs), whichtypically receive up to 90,000 alarmsevery day. Although only perhaps 1%to 2% indicate significant incidents orfaults, it is difficult to pick them outfrom all that noise coming off thenetwork.

According to intelligent automationsupplier Cortex, it has enabled aEuropean CSP operating company toclose down three NOCs and reassign

the 250 skilled technicians whoworked in them to more valuabletasks.

Similarly, a large North American cableoperator has used Guavus Alarm IQanalytics to silence unimportant alarmnoise. It applies machine learning andAI to alarm streams to classify whichalarms will result in problems forcustomers, which are associated withopen tickets and which can bediscarded, with a claimed accuracylevel of 99.2%.

This leaves staff free to escalate alarmsthat indicate genuine troubles and gaingreater insight into how issues that

impact customers develop on thenetwork.

So-called silent failures that affectcustomers but don’t trigger networkalarms are notoriously difficult toidentify and address – NTT DoCoMoused to employ 8,000 to handle thetask. Now it uses AI to classify datatraffic within cells as ‘normal’ or‘deviant’, with the latter causing silentfailures. This has greatly cut costs andimproved workforce productivity whileimproving customer experience.

In the next section, we’ll look at thechallenges CSPs face in deployingAIOps.

In October Deutsche Telekomannounced a pilot scheme using AIto streamline fiber-optic rollout inBornheim, Germany. WalterGoldenits, Head of Technology atTelekom Deutschland, said in astatement, “The shortest route tothe customer is not always themost economical… The newsoftware-based technologyevaluates using digitally-collectedenvironmental data. Where wouldcobblestones have to be dug upand laid again? Where is there arisk of damaging tree roots?”

A measuring vehicle was sent outin Bornheim (near Bonn) thissummer, equipped with 360°cameras and laser scanners. Itcollects about 5GB of surface dataper kilometer. Says Prof. Dr.Alexander Reiterer, who heads theproject at the Fraunhofer Institutefor Physical MeasurementTechniques (IPM):

“Such huge amounts of data areboth a blessing and a curse. Weneed as many details as possible.At the same time, the wholeendeavor is only efficient if you canavoid laboriously combing throughthe data to find the information

you need. For the planning processto be efficient the evaluation ofthese enormous amounts of datamust be automated."

Fraunhofer IPM has developedsoftware that automaticallyrecognizes, localizes and classifiesrelevant objects in themeasurement data. The neuralnetwork used for this recognizes atotal of approximately 30 differentcategories through deep learningalgorithms. This includes trees,street lights, asphalt andcobblestones, right down to thesmallest detail: Do the pavementsfeature large pavement slabs orsmall cobblestones? Are the treesdeciduous or coniferous? The trees’root structure also has a decisiveimpact on civil engineeringdecisions.

Once the data has been collected,a specially-trained AI is used tomake all vehicles and individualsunidentifiable. The automatedpreparation phase then follows in anumber of stages. The existinginfrastructure is assessed todetermine the optimal route. ADeutsche Telekom planner thendouble-checks and approves it.

Deutsche Telekom uses AI for fiber deployment

Page 17: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 17

Section 3

What are the challenges toimplementing AIOps?

Inconsistent andfragmented data

It is no surprise to see inconsistentand fragmented data as the topchallenge. No readily accessible,usable data means no AIOps, and thathas significant repercussions: WithoutAIOps, CSPs cannot virtualize andcloudify their infrastructure, whichimpacts 5G because it is cloud-nativeand IoT because operations will notbe able to scale to handle millions ofdevices and the concomitant traffic.

The issue of data is so fundamentalthat it demands CSPs rethink howthey run their businesses andoperations. Dr. Lester Thomas, ChiefIT Systems Architect, VodafoneGroup, comments:

While CSPs know they need to embrace AIOps, implementing it can be difficult for manyreasons ranging from a lack of software and analytics expertise and cultural fear of automation,to a lack of mature network components, support systems and standards. We asked operatorsto rate the challenges they are facing. The graphic below shows how they rank based on thepercentage of CSPs who put the specified challenge in their top three concerns.

1 2 3 4 5 6 7 8 9

Ranking the challenges to deploying AIOps

Inconsistent and

fragmented data

Lack of mature network

components and support

systems

Lack of standards for end-to-end

management

Lack of data analy cs exper se

Lack of so!ware exper se

Overcoming fear that

automa on will limit control and result in

outages

Concerns about

security

Explainability (explaining the

decisions algorithms

make)

Concerns about

displacing staff

TM Forum, 2018

Telcos [unlike internetcompanies] haven’tcome from abackground of data forits own sake and seeingit as an asset in its ownright, but you need thatapproach to say, ‘Howcan you optimize andimprove what you do?’

“TM Forum, 2018

57% of CSP respondents said inconsistent and fragmented data is a

big concern for AIOps

Page 18: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

Section 3

inform.tmforum.org 18

Some operators are striving toemulate webscale providers’ data-driven modus operandi by adopting a‘future operations mode’, whichVodafone, among others, iscollaborating on within TM Forum.

“This is a cloud-like way of buildingyour organization, not just the IT, withsmall domains that offer what they doas a service to other parts of theorganization and to third parties,”Thomas explains. “Smaller domainsare very important, they cover theaspects in autonomous units not onehumungous one. This is amicroservices culture with the agilemethod of operation based on data,that you continually optimize as itmatures, rather than measure, toachieve digital transformation.”

In parallel, Vodafone and a number ofother operators are working towardsintent-based network management,which abstracts the complexity of thenetwork at a high level and then usesa customer’s intent along withanalytics and policy to manage it.Thomas explains:

The level of interest in andcommitment to this intent-basedapproach was underlined by anaward-winning proof-of-conceptCatalyst project in Kuala Lumpur in

November. The Mindreader projectdeveloped a prototype for howoperators can predict customers’intent to serve them better and faster,through improved personalization.CSPs have always captured intent andepisode data, but they haven’t beenable to track or utilize it efficientlybecause it tends to get lost ininteraction notes or call detail records.

The project was led by Telstra withparticipants CloudSense, Infosys andNokia Australia.

Watch the Catalyst team discuss theMindreader project:

Steven Guggenheimer, CorporateVice President — AI & ISVEngagement at Microsoft, urges, "Youhave to think about the data estatefirst: If your data estate is not in order,or you don’t have a data pipeline forthe future then those early bespokeimplementations are not going to besustainable for the long term.Honestly, we should talk about BI[business intelligence] before AI …. ifyou’re not using data to drive insights,it’s likely too early to be trying todrive intelligence (AI)…if you aredriving insight then we can talk aboutAI.

“Beyond the BI to AI transition, weshould talk about SaaS and where itfits in — spending energy building AIinto a line of business tools (e.g. like acustomer care system, or a salessystem) that are already in use mightnot be the best place to expendenergy. You have to assume that SaaSvendors over time will add AIcapabilities into their offerings thatyou can leverage, versus buildingeverything yourself. It’s best to

consider applying your scarce AIresources to areas that drive realdifferentiation….. [Consider] networkoperations management, or price andsales optimization — there are otherplaces where you do have uniquedata and assets and where you dowant to do things yourself, asopposed to assuming it will comealong as an SaaS.”

Lack of maturecomponents andstandards

intent-based management needstraining data sets, but it is difficult toextract usable data from differentlystructured data stored in diverse,siloed, incompatible sources. Hence,data availability combined with therapid pace of change in the networkare factors in the No. 2 and No. 3AIOps challenges: Lack of maturenetwork components and supportsystems and lack of standards.

These issues are closely linked to alack of standards for end-to-endmanagement and are drivers of theTM Forum Open API initiative,instigated early in 2016 by Vodafone,Orange and BT. The aim is “deliveringa practical approach to seamless end-to-end management of complexdigital services” by enabling operatorsto plug incompatible systemstogether to make them interoperablefor specific, defined tasks.

Instead of modeling thenetwork, we model thecustomer’scommunications needs.at way, you cancontinuously changeand improve theimplementation. is isthe process the bigcloud providers use.

“TM Forum, 2018

47% of CSP respondents believe that lack of mature network

components and support systems is a primary inhibitor

to AIOps, while

41% believe lack of

standards is

Page 19: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

Section 3

inform.tmforum.org 19

Lack of skills

Lack of in-house data analyticsexpertise and software experience arealso key challenges. Many CSPs seehaving software, analytics and AI skillsin-house as strategically critical if theywant to maintain control ofoperations, and therefore their keyasset – the network. We’ll look howoperators are addressing skills in thenext section.

Fear of losing control

The skills issue is linked to theconsiderable fear CSPs have aboutlosing control of the network asAIOps become increasingly prevalent,which CSP respondents chose as thesixth biggest challenge they face inimplementing AIOps. Again, seeSection 4 for a more detaileddiscussion about this issue.

Concerns about security

Worries about security are also linkedto the fears about losing control andvisibility, in parallel with the imminentarrival of 5G. The issue is not so muchabout the next generation of wirelesstechnology, rather it’s about thenetwork of networks designed tomassively increase capacity, coverageand speed over the next few years.

This greatly increases the number ofinterdependencies increasingpotential weaknesses for hackers andother criminals to exploit, as wellmaking it difficult to figure out theimplications of those dependencies inevery set of circumstances.

Making AI explainable

‘Explainability’ is relatively low on thelist of challenges for now, but it too isclosely linked to worries about losing

control and about AIOps leading tooutages. A CSP’s engineers need tounderstand and be able to explainhow AI systems – whether developedinternally or bought from a vendor –make their decisions so that they canbe checked to ensure they complywith the organization’s goals andpolicies. A ‘by-product’ of this is thatthe line between operations andengineering will blur.

Also, operators would do well to heedmathematician and author CathyO’Neil’s warning that algorithms are“opinions embedded in code”, or theycould unconsciously or otherwisereflect the opinions of the coder.While there are clearly deep ethicaland possibly legal and regulatoryramifications, it would not serveoperators well in their developmentof intent-based approaches whichinvolves a large degree of judgmentand interpretation.

Another recent TM Forum Catalystproject called Artificial intelligencemakes smart BPM smarter looked athow to incorporate AI-based decisionmodeling and explainable AI (XAI) intotelecom business processes such asprovisioning, fault management,assurance and customermanagement. The team used standarddecision modeling and notation(DMN) to create a layer of XAI, orwhat amounts to an AI supportsystem.

Watch the Catalyst team discuss theproject: TM Forum, 2018

28% of CSP respondents fear losing control through automa on

TM Forum, 2018

22% of CSP respondents are worried about security threats

with AIOps

TM Forum, 2018

19% of CSP

respondents are concerned about how to make AI

explainable

TM Forum, 2018

39% of CSP respondents believe lack of analy cs skills is a major issue for AIOps, and

38% believe lack of so!ware

experience is

Page 20: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

Section 3

inform.tmforum.org 20

Worries about displacingstaff

‘Anxiety about staff being displacedby AIOps is inevitable, but as we’vementioned briefly above and explorefurther in Section 4, many operatorsare busy retraining as well asrecruiting staff and the ‘right’graduates to gain expertise inanalytics, automation and AI. Many ofthe migrations underway – forexample, the move to distributedmodels like edge computing pluscloud-based architectures andvirtualization in parallel with adoptingAIOps – are likely to require more, notfewer, staff during the transition.

Greatest data barrierTo implement AIOps, CSPs mustexpose the operational data neededfor network automation andefficiency, but this too comes withchallenges. Microsoft’s Guggenheimernotes, “Most telcos have so muchdata, the question becomes ‘Of whatyou have, what do you want to use?’Then, ‘Of what you have, what do youwant to try and make usable?’, and ‘Ofwhat you have what do you not wantto try and make usable, so let’s startover and try to collect it in a betterway?’ This is the dialog, and we areseeing [telcos] trying to implementthat, but right now they are busybecause there is so much they canuse and they are still doing a bunch ofthat work”.

The graphic below shows whichbarriers CSPs feel are most difficult toovercome. By a considerable margin

(almost 16%) they indicated thatbeing unable to access data spreadacross operational silos is the mostsignificant barrier to the effectiveexposure of operational data.

This issue has dogged CSPs for years,and they are taking variousapproaches to addressing it. Korea’sSK Telecom, for example, collapsed150 OSS platforms into one to help itunify data formats.

Two years ago, Telefónica created itsdata and AI unit LUCA by pulling itsexpertise into one unit. It is nowintegrating the Stratio Data Centricplatform: The combination will deliverwhat the company calls a Big DataPlatform as a Service (PaaS) solution.It is designed to speed up corporatecustomers’ digital transformation oftheir operations. Telefónica said it willstart deploying the new solution insome Latin America markets “soon”.

The operator claims that its newplatform differs from others on themarket because it can work withmultiple data sources, at large scale. Itsays the multitenant platform“revolutionizes the application-centricmodel through a unique interface thatguarantees data consistency and

provides a unified vision andadvanced intelligence for myriad usecases. Stratio Data Centric bringstogether the analytical andoperational capabilities of theenterprise, enabling data intelligencefor real-time operations.”

LUCA has relationships with morethan 150 corporate customers in over20 countries and is working on 200projects. Advertising and media,financial services, retail, tourism andtransport sectors are its key markets.The unit was positioned as a marketleader in a report published byForrester Research in October.

Diving into data lakesAn increasingly popular approach togaining access to usable data is toextract it in its native form and storeit in a data lake, often usingarchitectures like Hadoop because ofits massive storage capacity,extensibility and scalability. Datalakes, which allow for data to be leftin its original format making themmore flexible and less expensive thanother options such as datawarehousing, should be a key part ofany AIOps strategy.

Having extracted data from theirmultifarious systems, some large CSPshave taken it upon themselves toaddress the thorny matter of makingtheir data fit for purpose. Verizon’sMatt Tegerdine pointed out that oncehis team started investigating dataproduced by Verizon’s networksystems, they realized:

Chief barriers to exposing opera onal data

TM Forum, 2018

44.5%

1.5%

21%

5%

29%

Inability to access the data needed across opera onal silos

The lack of clear standards for exposing data streams

The lack of relevant APIs for exposing data streams (e.g. open event streaming integra on APIs)

Lack of provisions for/concerns about security

Other

e data was reallyprimarily being tappedby vendor systems, soessentially we’re payingvendors to go out andmine our resources thensell them back to us.

TM Forum, 2018

16% of CSP

respondents have concerns about

AIOps displacing staff

Page 21: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

Section 3

inform.tmforum.org 21

In addition, Verizon cannot find asuitable AI solution on the market touse the data and help the operatorgain greater autonomous control ofthe network. Consequently, alongwith others, such as internationaloperator Telia Carrier, it is looking todevelop its own solution. We lookfurther into the vexed issue ofoperators’ somewhat fraughtrelationships with equipmentproviders in Section 4.

Is collaboration theanswer? Lack of standards again rears its headin CSPs’ answers to the questionabout exposing operational data, withalmost a third identifying the lack ofclear standards as the biggest barrier,followed closely by lack of the rightAPIs. Given these concerns, we askedCSPs and suppliers how useful ornecessary they feel collaboration onstandards and APIs is. We asked themto rate the need for collaboration infour areas as either essential, veryuseful, moderately useful, slightlyuseful or not useful at all. The graphicbelow compares the percentage ofCSPs and vendors who saidcollaboration is essential.

Perhaps not surprisingly, CSPs aremore enthusiastic than suppliersabout collaboration, which reflectsthe strategic importance operatorsare attaching to developing expertiseand skills in-house and the fear ofbeing overly reliant on suppliers(again, we’ll cover this relationship inmore detail in the next section.)

The growing industry consensus isthat operators need common datamodels and a phased standardizationof how data is classified, collected,distributed, managed and used todrive decisions. TM Forum’s

Information Framework data model(previously known as the SID – seepage 43) is used by most CSPsaround the world, and indeed hasbeen developed and refined by theoperators themselves and theirsupplier partners over many years.The Forum is now working on afunctional AI data model, starting bypackaging all the relevant entities,then identifying the gaps and buildingextensions to address them.

Percentage of respondents who believe collabora�ng on AI is essen�al 27%

32%

18%26%

16%24%

16%22%

The development of an AI data model for sharing data streams

Suppliers

The crea on of an AI maturity methodology and model that enable CSPs to assess how

far along the journey they are in implemen ng AI in their company, networks

and opera ons

The development of AI management standards that enable AI to be structured,

monitored and managed effec vely

The crea on of a repository of anonymized data sets for training and valida on purposes

TM Forum, 2018

CSPs

Page 22: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 22

Section 4

Developing and finding AIOpsexpertise and solutions

We asked CSPs about how they aredeveloping solutions for AIOps. It isstriking that more than half of them,54%, said they are primarily relying ontheir own internal resources todevelop their AI solutions (see graphicbelow). This begs big questions. First,is building AIOps solutions what CSPswant to be pouring their resources andtime into, as opposed to running theiroperations and businesses with 5G, IoTand so much more that is imminent?As Utpal Mangla, Vice President &Partner, Global Leader – Watson AI,IoT & Blockchain at IBM, puts it: “Theircore competency is telecoms, notbeing an AI product company.”

Where to find the skills?A second question is: Where will theinternal expertise come from as theseskills are in short supply the worldover? Huawei has recognized thisopportunity and is offering to trainoperators’ staff, and this trend couldspread to other suppliers and systemsintegrators. Some CSPs are addressingskills shortages in creative ways, fromretraining/reskilling staff (such as thosedisplaced when network operationscenters are disbanded) to graduaterecruitment policies.

Our impression is that smallercompanies generally are more decisive

and active on the AIOps front,including building up internalexpertise. Christian Yde, Lean Agile ITleader at Danish operator TDC Group,says that top management at hiscompany has “bought into AI” andcreated a specialist division with ateam of 70 — a mixture of datascientists, mathematicians,statisticians, developers, graduates andconsultants.

The line of command here isinteresting too: The unit is run by theVP of AI and Robotics, who reports tothe Chief Digital Officer, who in turn isanswerable to the CEO of the OpCo.Note that TDC Group is a joint venturebetween the OpCo and the Netco,

which owns the infrastructure. Thetwo became separate legal entities inJune. BT, Telefónica and TIM have alleither separated their ops andnetwork, or are exploring the option.

This approach is gaining traction withEuropean telecom regulators inparticular as, despite years of whatmany view as heavy-handed regionalregulation, they have failed to curtail themarket dominance of the formerincumbents and/or oversee the creationof competing national infrastructures.Instead, they have ended up with manycompeting networks in denselypopulated areas and little alternativeinfrastructure elsewhere.

Survey responses and interviews about finding AIOps expertise, and the relationships betweenCSPs and suppliers, turned out to be some of the most intriguing information we collected. Atbest the relationship between CSPs and vendors can be characterized as mismatchedexpectations, but at worst it could signal deep-rooted distrust on the part of CSPs, which iscreating an impasse. One thing seems certain: Although operators know AIOps will be critical,they are determined to move at their own pace and ‘not break things’, and this means notgetting tied to vendors’ roadmaps.

With an exis�ng technology partner(s)

How are CSPs developing AI solu�ons?

54%

40%

38%

37%

32%

With newly recruited internal resources

With a new technology partner(s) that specializes in AI

With new technology partner(s) that also has/have a tradi�onal

telecoms-orientated product por olio

With exis�ng internal resources

TM Forum, 2018

Page 23: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

Section 4

inform.tmforum.org 23

The benefits of creating separate legalentities between ops/business and thenetwork is that building and runninginfrastructure has much longerinvestment and return cycles, withinitial high costs, in sharp contrast tothe much shorter cycles of retailbusinesses.

Yde explains the new division’s solepurpose is to build and spread AIacross the organization, although sofar, the only thing that has gone intoproduction is next-best-thing offers tosupport call center staff.

Gorazd Hribar Rajterič, Telecoms OSSExpert, Telekom Slovenije, observesthat larger operator groups “havegreater resources to invest in AI andgrow competition in AI and employspecialists, but smaller ops need tospend money carefully, yet don’t wantto be at the mercy of vendors, so wehave to find some middle ground.” Hiscompany is working with universitiesand TM Forum to get to know thetechnology. Telekom Slovenije alsohosts students who bring freshknowledge, defines projects to bestudied at university and has joined anumber of European Union AI projects.

Who’s more experienced?A third issue is experience with AI. AsSholom Weglein, Product Managerwith Amdocs, points out: “CSPshaven’t got the experience we haveas the result of working with many ofthe world’s largest service providers.

You need the expertise to build datamodels, and to know which tools arebest suited to each task and whichareas to focus on.”

Internal expertise aside, on thesubject of working with newsuppliers, a senior executive from onesuch vendor comments:

Rajterič doesn’t see it quite like that.“Operators are ready to let go of thecomfort blanket of the big vendors,but while in our core networkoperations we have been workingtowards open source, our network isEricsson-based and there is nosudden escape from that,” he says.

Working with vendorsTelefónica International WholesaleServices (TIWS) is a good example ofa telco working on AIOps with anestablished equipment provider. Thecompany is just starting to replace theentire national infrastructure of Braziland will run an AIOps testbed in

parallel, looking at the performance ofthe network between cities, accordingto Jose Manuel de Arce, DeputyDirector OSS & BSS Infrastructure,WorkSpace, OSS Technology. TIWSwill be deploying Juniper Networks’NorthStar Controller in the core tohelp it automate operations, which isdesigned to balance traffic andservice delivery priorities intelligently.

Or as de Arce puts it: “We are movingfrom SNMP and REST to telemetry;for the first time we’ll haveinformation in real time, as thingshappen, not when we retrieve them— you could say we are moving topush instead of pull.” The companyalso will be deploying Hadoop forabig data approach in 2019, butstressed that his company wouldmaintain control of the data at alltimes.

Vodafone is working in open sourcegroups, but in an effort tocomplement what its vendors aredoing, not replace them. Says Dr.Lester Thomas, Chief IT SystemsArchitect at Vodafone Group: “Wehave never attempted to buildsolutions ourselves, we rely onvendors, but it’s constant negotiationand evolution – for example, we havejoined ONAP but not to build an opensource alternative, but to buy fromdifferent commercial suppliers andorchestrate their solutions, askingthem to follow ONAP standards andmechanisms. We have the vision, butit takes as long as it takes.”

Operators talk and talkabout bringing in newvendors and using opensource, but when it comesto the writing the check,they lose their [courage].

Page 24: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

Section 4

inform.tmforum.org 24

Improving processesArnold Buddenberg, EnterpriseBusiness & IT Digital TransformationArchitect, Orange Group, notes thatit’s important for CSPs to improveprocesses as part of their marchtoward AIOps. He says:

Amdocs’ Weglein advocates buildingopen solutions so operators canintegrate their own preferences with nopressure to change the way theyoperate at the beginning. Instead theidea is to help identify issues and dealwith them much faster as the first babystep. Then introduce AI to help with thesolution while the operator is still incontrol, only moving to full automationonce the operator is comfortable andconfident.

“AIOps is so broad and affects so muchthat it can help improve operations nowand then play a role in IoT businessmodels [or anything else],” he adds.

Differing organizationalapproachesWe also asked CSPs about theorganizational approaches they’reusing to develop AIOps solutions.With hindsight, it seems it may havebeen too soon to ask this questionbecause most operators are notsufficiently advanced in deployingAIOps to answer it. As Orange’sBuddenberg comments:

Although 42% of CSP respondentssaid they are developing solutions tomeet specific needs in a particularbusiness unit, it’s a concern that only13% have a structured, centralizedmechanism for sharing experienceand knowledge. This was also borneout in our interviews.

Encouragingly, a quarter of operatorsare developing solutions centrally topush out to other business andoperational units, although the riskhere is that one AI size does not fit allpurposes and situations.Nevertheless, this is a good way toexperiment and build expertise.

Vodafone Group’s Thomas, explainsthat although his company’s strategyis centralized around the future modeof operations (see page 18 and thereis guidance on all design, “The generalculture is to leave them to it…usingthe central blueprint of what we thinkAIOps should be. We use opensource and recommend certainvendors for certain things.” The teamsalso have access to centralizedresources such as Hadoop and sometooling.

Vodafone is encouraging its operatingcompanies to set up communities topull experience and knowledge fromdeployments, but Thomasacknowledges that each company hasits own priorities based on itsparticular B2B2X model. He alsopoints out that different teams needdifferent AI expertise. Vodafone hasset up a central team of about 10 AIexperts to look after the technologicalaspects of AI, while another team ofthe same size focuses on thecommercial elements.

Risks of informal sharingThe 19% of respondents who saidtheir companies are relying oninformal means of sharing what theylearn, risk failing to pool expertise andexperience, and building on it. Thiscould result in CSPs duplicating effortand spending more time and moneythan they need to, as well as runningthe risk of ending up withincompatible systems and approaches– the opposite of end-to-end control.

This situation could prove expensive,in every sense, in the longer term. AsIBM’s Mangla puts it: “Fine on dayone, not so much on day three.”

For now there is a lot ofthinking, notdeployment – althoughthe feeling is [AIOps]will become critical.

Which organiza�onal approaches are CSPs using to AIOps?

TM Forum, 2018

42%

26%13%

19%

AI solu ons and technologies are being developed centrally and pushed out to business and opera onal units

AI solu ons are solely being developed to meet specific requirements within a business unit/opera onal unit

AI solu ons are being developed to meet specific requirements within a business unit/opera onal unit with lessons and solu ons shared with others informally

AI solu ons are being developed to meet specific requirements within a business unit/opera onal unit with lessons and solu ons shared with others formally through structured processes

All the vendors talk aboutAI as much as 5G, so it’shard to know the rightstarting point… eindustry is not used tothinking in process terms,and you need loops andprocesses for AI – and youhave to really understandand be able to describe thatprocess, particularly in thenetwork. If it’s in a blackbox, you don’t understandit. We need to improveprocesses… Do we eitherwait for vendors to comeand tell us how to do it witha black box or I understandmy processes and have awhite box that I own?

Page 25: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 25

Section 5

Make it happen Strategies for implementing AIOps

Go slow to go faster laterAlthough it feels as if the level ofenthusiasm for AI has dippedcompared with our research last year,implementing AI in operations is amajor undertaking by any measure.Hence progress might appear slow,but CSPs are determined to becautious and not risk the integrity ofthe network, because getting AIOpsright is critical to the future of theiroperations and businesses.

Automate, automate,automateNetwork functions virtualization(NFV) and software-definednetworking (SDN) have slipped downoperators’ agendas for the time being,although they are inevitable. The bigtakeaway from their slow progress isthat unless the network is automated,NFV and SDN will likely increase, notdiminish, operational costs.

Sharing is essentialIn our research for this report, wefound that operators typically havelittle formal structure concerning thesharing of AIOps expertise andexperience internally, which risksmuch reinventing of wheels,unnecessary expenditure and slowerprogress. Operators recognize that AIexpertise and skills are an importantstrategic asset if they are to controlthe network by autonomous meansrather than have control of theirnetworks wrested from them. In theshort term at least, AIOps are likely torequire more rather than fewerpeople at least through the transitionphase, and developing talentinternally will be necessary.

AI is integral to datastrategyAll kinds of AI are fueled by data, andlots of it, so AI need to befoundational to every organization’sdata strategy. To provide the requiredfuel, CSPs needs to natively exposedata, remove it from silos then createdata lakes and make the data storedthere accessible to the entirecompany via open APIs. Simple to say,less easy to achieve.

Lead from the topWhile experimentation and runningsmall specific AIOps projects are agreat start, CSPs need leaders with avision about how to reach the end-game of achieving autonomouscontrol of the network. While this willinform use cases and progress, andensure they are aligned withcorporate strategies, it does notpreclude experimentation in pocketsor making decisions now about theultimate centralization of AI acrossthe organization in these early days.But we do urge CSPs to centrallycoordinate the areas that are key to AIsuch as data management andgovernance, warehousing and datalakes, and commonality of AI andanalytical capabilities.

The most likely candidates for leadingthe corporate AI charge are the CIOor Chief Data Officer, or the CTIO, asnetwork and IT merge. It matters lesswhich position they hold than beinggiven the authority to carry thestrategy through – read more aboutthis in the recently published TMForum report, Redefining the CTIO:Essentials for the digital age.

Page 26: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

Section 5

inform.tmforum.org 26

Collaborate andstandardizeAI is so integral to every CSPs’success, it is a prime candidate forcross-industry collaboration to createstandards, models and best practicesby sharing knowledge, skills andexperience. As we have learned fromour survey results, operators tend tobe keener on this than their suppliers,but it is up to operators to encouragetheir participation. TM Forum, ETSI,and the Telecom InfrastructureProject (TIP) initiated by Facebook areall working on AI initiatives.

Join a TM ForumCatalyst projectAnother option for collaboration isparticipation in a ground-breakingproof-of-concept Catalyst project. Arecent award-winning project calledArtificial Intelligence for ITOperations, which was demonstratedat Digital Transformation Asia, scoreda number of firsts. This includes ChinaMobile, China Telecom and ChinaUnicom collaborating to see how theycould improve users’ mobile internetexperience by optimizing networksand IT operations. They used AI andbig data analytics to examine faultsand route around them, predictincidents and look at usage profiles togain continuous insights across IToperations management. They weresupported by participants BoCo Inter-Telecom, Huawei and SI-TECH.

Watch the team discuss the project:

Page 27: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

Additional features and resources28 | Driving an enhanced customer experience with artificial intelligence

32 | The reality of Artificial Intelligence and Intelligent Machines

34 | Succeeding with big data, AI and analytics – Real-world examples

37 | The promise of AI is holistic transformation

Page 28: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 28

Driving an enhancedcustomer experience withArtificial Intelligence

As we can see from the TM Forum’sstudy, service providers who havestarted implementing AI-drivenoperations have seen an increase incustomer satisfaction, a vital tool inthe battle against customer churn.Now that communications serviceproviders are transforming into digitalservice providers, they are obligatedto take a fresh look at their operationsand focus on how the can leverage AIin their operations to advance theirdigital business.

And they do: over 75% of operatorscited better customer experience asthe driver for intelligent automation,while another 45% (multiple choiceswere enabled) cited the ability topredict and prevent outages, whichhas a direct impact on customerexperience, as the main rationale anddriver for implementing AI.

At the same time, AI is being used todrive new revenues and services.New digital services require fasterprocessing, and with a dramaticincrease in the number oftransactions, automation is a must.When asked what is drivingautomation in operations, customerscited new services (42%) andfrequent changes to processes andsettings (28%), highlighting the needfor automation in the faster-paceddigital world. This is the demand

today, even before theimplementation of 5G, which willsignificantly increase demands onoperating systems.

Another aspect mentioned is theability of automation and AI to reducecosts. Over 60% of respondentsmentioned OPEX reduction as one ofthe three main drivers forimplementing AI and automation.

e impact of operationson customer experienceCustomer experience is far more thanthe services offered or the waycustomers communicate with theservice provider. Even if a new appuses the latest design, and the serviceprovided is in great demand,customer experience can still be poorif operations do not maintain servicequality. This is what encouragescustomers to implement AI-drivensolutions that impact customerexperience. The top two responses,when asked where they are investing,relate to experience, with 80% citingcustomer care, and another 60%focused on predictive maintenance(multiple choice question).

The importance of a good customerexperience is even more pronouncedin today’s environment, given thegrowth of media services and partner-

based solutions. Customers expectseamless, error-free service,regardless of the type of service orwho is providing the solution.Operations need to be fast, accurateand must cover the completeecosystem. Operations can impactthe customer experience through:

Service quality – in many aspectsservice quality is more important thanthe application it is supporting. If youlaunch a cool new self-service app,but the billing system is down formaintenance, the customerexperience will suffer. Similarly, ifthere are any errors in the systemsimpacting the successful resolution ofprocesses like payment, customerswill become annoyed, and startlooking elsewhere. Such issues can beprevented through good operations.

Speed of operations – the amount oftime it takes for a service provider toresolve an issue, or simply how long ittakes to complete routinetransactions, such as new serviceactivation or payment, can often beanother source of customerfrustration.

Bottom line – the faster and moreaccurately operations take place, themore satisfied the end user, and themore value operations can drive.

Artificial intelligence (AI) has a key role to play in the smooth operations of any business today.And by enhancing the customer experience through improved service quality and speed ofproblem resolution, well-run operations bring business value and company growth.

Page 29: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 29

Improving speed andquality through AITo enhance customer experience,operations can leverage various AIcapabilities, in several different usecases, improving both the quality andthe speed of operations.

Preemptive issue identification andresolution – advanced monitoring andanalytics can help discover issuesbefore they impact customers. Trendsare analyzed, comparing actual datato previous “normal” behavior, lookingfor anomalies that indicate a potentialissue, and taking measures to preventthe “glitch” before the customerexperience is impacted. The systembasically heals itself beforemalfunction.

Automated operations – the use ofrobotic process automation across alloperations, whether planned, such asbill processing, or unplanned activities(for example ticket resolution), helpsdrive faster and more accurateoperations.

Auto-ticket handling – with naturallanguage processing capabilities, youroperations control system canautomatically read and classify ticketscreated by users across theorganization. This means tickets nolonger need to wait hours, or evendays, just to be assigned to the right

expert or team. Auto-ticket handlingmeans that as soon as a ticket iscreated, it can automatically beprocessed and sent for handling,thereby shortening time to aminimum for issue resolution.

Combine all three capabilities (seefigure 1) and you get predictive andpre-emptive, zero-touch, self-healingoperations. The system will detectissues before they impact customers,automatically classify the type ofissue, and select the right action totake to avoid the issue -automatically.

Another use of AI is in auto-rootcause detection, using deep learninganalytics to group common issuesand identify the root cause for amalfunction or a breakage in theservice or process, whether it stems

from a defect in the code or fromproblematic infrastructure issues. Itcan then suggest a fix to the stemproblem, thereby solving that specificcause for problem, helping to reducethe overall number of issues thatoccur, and improve the overall systemstability.

One further AI-technology, machinelearning, can be leveraged forcontinuous improvement. Withmachine learning, each interaction is alearning opportunity for the system,either to confirm its analysis, or torefine the models for more accurateanalysis the next time. This allows forautomation of the automation, withthe AI system able to create its ownautomation routines, based onobservation of the human behavior ineach iteration.

Page 30: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 30

Getting started withautomationAs we have seen, AI-drivenautomation can add tremendousvalue to your operations. But thevalue can only be achieved ifperformed correctly, using the rightapproach to implementing AI andautomation. As we see from theTMForum study, 45% of serviceproviders are currently focused onlyon implementing automation in asingle process in one domain.

For the service provider who wants,or more accurately needs, toaccelerate their automation, there area few key points to consider to ensurea successful project.

Don’t do it alone!When implementing AI-drivenoperations, one of the biggestobstacles service providers face is alack of expertise, whether in dataanalytics or software. Serviceproviders have an abundance of dataavailable to them but without theright expertise, it is difficult to knowhow to build the right data models torun the operations. And with a wealthof technology to choose from, serviceproviders find it hard to decide whichsolutions are best for whichapplications.

To overcome this and accelerate yourjourney, you need to partner with anAI expert who can help you with:

1. Building the right data models –leveraging the data you have iscritical to success. Knowing whichdata is important and which youcan ignore, can be the differencebetween success and failure

2. Choosing the right technology –not all tools are created equal.Some may be better for someapplications, while others may bebetter in other areas. Being awareof which technology to use canimpact the success of your project

3. Knowing what to automate –

domain experts have theexperience and knowledge to helpdrive a value-driven plan. You wantto ensure that your automationdrives the optimal value, and thatyou are not simply automating forautomation’s sake

4. Learning from others – in somecases, you may not haveexperience in a certain domain (e.g.IoT). Having a partner who workswith other service providersinvolved in those domains can helpfast track your deployment throughthe use of data from other sourcesto help build your initial datamodels.

Building confidence with a step-by-step approachAnother key challenge is the fear of“explainability” -- the inability tounderstand why the system made thedecisions it did, which makes it harderfor humans to hand over control toautomation. This fear isunderstandable but can be overcomeby a step-by-step approach designedto help the organization learn to trustthe system. (see figure 2)

1. Use AI/analytics to help identifyproblems faster – the first step is toharness the analytics capability tofind issues faster. This can bethrough constant monitoring andanalytics of real-time data andspotting the anomalies whichindicate issues. This can be takenone step further with predictive

analytics that look for trendsindicating a future issue. Furtheranalytics can be applied to filter outnoise and focus on the alerts thathave real business value. Eitherway, the issue is identified quicklyand can be resolved faster.Additionally, the system learnsfrom each incident to see how theoperations team reacted, andrefines its models through machinelearning, leading to more accurateissue identification in the future.

2. Use AI to identify problems andrecommend solutions – once aproblem has been identified, thesystem can identify the rightresolution path. The system canguide a human on the correct stepsto take. Through machine learning,the system can learn which of itsrecommendations areimplemented, and which are not.This helps the system becomebetter at selecting the correctaction.

3. Use AI-driven automation toimplement the resolution – thefinal piece to the puzzle:automating the resolution of theissue. This creates a complete zero-touch environment. Operationscan be fully automated, driven byintelligence.

This whole process can be repeatedfor each operations domain, givingyou complete control of what, andwhen, to automate your operations.

Page 31: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 31

Taking it one more step –predicting the unknownDeep learning is an area where wesee limited investment [only 10-15%of service providers cited deeplearning investments in a couple ofuse cases] but its potential istremendous. With deep learning, AIsystems can train themselves to spottrends, correlations and so on fromamounts of data so vast that humansare unable to make sense of it. Thiscan be a critical tool for serviceproviders as they launch newservices, such as IoT-based solutions,for which they have no historical data

to build data models. Deep learning-based systems would help identifypotential issues and alert theoperations team, along with providinganalysis on which to base theconclusions. Humans would then beable to make educated decisionsabout how to proceed, and help thesystem become more accurate in itsassessment.

Amdocs Global SmartOps –driving business value throughIntelligent automationAmdocs Global SmartOps harnessesoperations to enable digital

transformation within a hybridecosystem, driving business growthand enabling service providers todeliver a superior, seamless customerexperience. This is achieved throughour unmatched domain expertise andan end-to-end operations approach,including AI-driven automation, self-healing and pre-emptive issueresolution. Amdocs Global SmartOpsis powered by Amdocs’ uniqueatomIQ platform, a standardizedplatform of tools, global best practicesand capabilities, which infusesartificial intelligence, analytics andautomation into all aspects ofoperations.

About AmdocsAmdocs is a leading software and services provider to communications and media companies of all sizes, accelerating theindustry’s dynamic and continuous digital transformation. With a rich set of innovative solutions, long-term businessrelationships with 350 communications and media providers, and technology and distribution ties to 600 content creators,Amdocs delivers business improvements to drive growth.

Amdocs and its 25,000 employees serve customers in over 85 countries. Listed on the NASDAQ Global Select Market,Amdocs had revenue of $4.0 billion in fiscal 2018.

For more information, visit Amdocs at www.amdocs.com

Page 32: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 32

e reality of ArtificialIntelligence and IntelligentMachines

Uncertainty in the markets alwayscreates opportunity, but marketuncertainty for CommunicationService Providers over the last 10years seems to have constrainedinnovation. Considering theheightened challenges faced fromhigh value content providers; over thetop services; and diversity in markets,now more than ever innovation andimproved customer service must beof primary concern.

There will always be challenges in amarket that is so capital intensive.Whether coming from vendors whoincite CSP’s to invest in the latest 5Gtechnology, IoT offering, SDN/NFVsolution, or the next “best” thing. Thisaccentuates the temptation toconfuse big capital projects, withinnovation or market shifts. Supplierinnovation rarely in itself createssufficient value in new products, newmarkets, or new opportunities, etc.Some will window dressremanufactured services such as“quad play” as ground breaking, somewill laud incremental, faster, better,cheaper, developments such as 5G asrevolutions, whilst many acquiesce tothe demand of investors withstrategically significant acquisitionsand divestments, none of these arecompetitive innovation. Trueinnovation in complex and maturemarkets such as CSP’s takes the fullfocus of clever and experiencedpeople in those markets. Like mostinnovation most attempts will fail, but

without testing sufficientopportunities, success will neveremerge. This determination comesfrom those with deep marketexperience and service knowledgethat are buried deep inside thecurrent operations.

So how are CSP’s supposed to seekinnovation and competitive edge in aworld where the best placedmembers of their organisations areburied in network design, operations,and customer services? Consigned todaily firefighting just to maintain thelegacy leviathan capitalisedinfrastructures to support ever morestressed services to increasinglydemanding clients… more… faster…cheaper.

AI FictionFrom Artificial Intelligence guruscomes a new generation of promisewhich has been generated by vastarrays of cheap storage; theabundance of low quality, fragmented,datasets combined with theexponentially increasing computepower. The accuracy, reliability,durability, resilience of the results isusually high enough to be interestingbut too low to be useful. Examplesare abound of chatbots that are easilydefeated, analytics that producepretty patterns but misleading results,neural networks that can reliablydemonstrate human weakness, andbias.

The problem with “shock and awe” isthat it rarely delivers the claims andusually creates a baptism of fire forthe uninitiated. The benefit of “shockand awe” is that it can often providesome political “air cover” diverting theprying eyes of judgement; giving timeto the less fantastical and morepractical revolution happening on theground.

AI FactWhilst the AI press makes all thenoise and many outlandish claims, thequiet revolution that is reallytransforming operations in some ofthe less conventional corners of CSP’shas accelerated less noticed… Whilstthe boardroom is having its headturned by main news streams carryingthe “shock and awe” stories about the“Artificial Intelligence revolution”, or“Robots taking over the workplace”,some real operations have beenmaking transformational strides usingthe less sensational, and morerealistic, machine intelligencecombined with dynamic processorchestration known as “IntelligentAutomation”. Not bearing thesequined jazzy coats of the ArtificialIntelligence movement, IntelligentAutomation deftly implements reliablemachine intelligence capabilities thatare robust, accurate and scalable.These release skilled operationsteams to do significantly higher valuework; most importantly tackling theinnovation challenge that is presentedto all CSP’s.

“Wasn’t it your mother that said you should learn to walk before you run?”

Page 33: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 33

Man MachineCollaborationAs we are a long way from AI doingeverything, if ever, the promise of AIwill only be delivered in acollaborative environment by buildinga trust relationship between man andmachine. In every case, this trust mustbe first built between man andmachine, then extended by theparticipants in forming the confidenceto allow the machines to operateunaided in the environment. Thecurrent rolling media squall ofadvanced and, the currently mythical,general AI promising to transform theworld is failing to build the trustrequired to formulate a road toautonomous operations. At the otherend of the scale, the basic actionbased “Robotic Process Automation”which seeks to integrate intelligentaction to establish the trust to actautonomously, is so fragile that itundermines the very trust required tooperate autonomously. The road totrust, let alone the road to confidence,does not support unpredictable orgeneralist techniques.

Based on well proven and reliablemethods of operational analytics anddeterministic decision models,Intelligent Automation implementstechniques that provide morepredictable outcomes rather than the

probabilistic approximations thatmore advanced AI methods propone.This produces a rapid route from the“Human in the Loop” type operationto the Supervisory mode where realtransformation releases significantresources to innovation and customerservice.

Sense, Analyse, Decide,ActIntelligent Automation is “theapplication of analytical and decisionsystems to take actions to achievepredictable outcomes.”. This is oftenconfused with AI which the ForbesInsights survey defines as "the scienceof training systems to emulate humantasks through learning andautomation.". We would argue thisdefinition is riddled with issues: Notleast because, AI is typically numericalanalysis applied in discretemathematics, it is more anengineering discipline than a science.Also, the word “emulation” and“learning” do not sit comfortably inthe same sentence in the same senseas “learning” by rote is moreemulating than learning.

However, even with this very loosedefinition, Forbes draw a similarconclusion for AI as Cortex make forIntelligent Automation that “Trust willplay perhaps a larger role in the

evolution of AI than it has for anytechnology in recent memory."

Taking the Next StepsCortex Intelligent Automation is thefirst platform specifically built to solvethe challenges that preventorganisations accelerate down theroad to an autonomous future. Cortextechnologies build the trust andconfidence that create the ultimatecollaboration between human andmachine. Advanced AI will eventuallyprovide predictable outcomes thatcan transform operations to releasemore resources to innovation andcustomer services, however, to takeadvantage of this organisations muststart to build trust and confidence intoday’s predictable technologies.There is no shortcut to advanced AI,and leading innovation.

The road to autonomy is littered withpainful lessons that has been well-trodden by the aircraft and roboticsmanufacturers alike who understandthe necessity to build trust throughdeterministic human in the loopactivities, and confidence throughpredictable automation with a humansupervisor, ultimately gaining highlevels of machine autonomy in a man-machine collaboration built on trustand confidence.

About CortexCortex Intelligent Automation is the first unified platform specifically built to solve the challenges preventing organisations’acceleration to an autonomous future. Cortex rapidly creates value, using multi-purpose intelligent automation software totransform telecommunications operations.

A unified, no-code, automation and orchestration platform, Cortex delivers Workflow, Orchestration, Automation,Reasoning, Integration and Event processing. Unique, decision-driven, closed-loop, and self-adjusting automationtechnology seamlessly integrates into existing and legacy technologies, automating processes to increase accuracy, speed,agility, and to deliver tangible ROI.

With strategic partners including Capgemini, TCS, and Tech Mahindra, Cortex applies proven strategies and methodologiesfor Intelligent Automation deployment, together ensuring that the most successful outcomes and ongoing autonomousoperations are achieved.

For more information, visit Cortex at www.cortex-ia.com

Page 34: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 34

Succeeding with big data, AI andanalytics – Real-world examples

Acceleratedtroubleshooting, millionssavedA top North American cable operatorusing customer care analytics hasreduced costs by more than $102million while improving customersatisfaction levels.

Here’s how:

Slashed problem resolution times.The MSO was receiving large volumesof customer service calls, and theoperations team was struggling toidentify the root of the problem.Technicians were dispatched tocustomer locations, only to frequentlyfind that the glitch wasn’t in thecustomer’s set-top box, but in aheadend device instead. The situationfrustrated both the customers andcable operator, wasting time andmoney.

The MSO used the Guavus Reflex®solution with Guavus Live Ops to

adaptively correlate data frommultiple sources – technical supportcalls, subscriber trouble tickets, andtruck rolls – allowing their Care Opsteam to quickly discover issuescommon across micro-populations ofsubscribers. Using machineintelligence, Guavus helped the teamimmediately triage and pinpoint theproblem and resolve it.

Results: Not only were customershappier but by slashing truck rolls andcustomer service calls, the operatorsaved approximately $70 million inthe first year alone.

• Proactive maintenance – GuavusLive Ops has now been integratedinto the same operator’s step-by-step methods of procedure (MOP)that its technicians follow whenimplementing a network change orupgrade. Guavus Live Ops allowsthe MSO to proactively monitor anypotential negative impacts of anupgrade across the operator’sinfrastructure and if any are found,

to take measures to correct thembefore they become a widespreadproblem.

• Prioritization of capital expenses –When the cable operator rolled outa new video service, analyticsenabled it to prioritize which piecesof equipment to upgrade first. TheGuavus solution correlatedsubscribers to each type ofequipment, examined failure ratesover time, and identified the impactof these failures on customers.Using with this information, theteam was able to prioritizeequipment changes, minimizecapital expenses, and roll outenhanced video-on-demandservices faster.

Results: The operator realized acapital expense savings of $32million while maintaining highcustomer satisfaction levelsthroughout the deployment of thenew services.

Communications service providers (CSPs) that have succeeded at monetizing data haveimproved their operations and customer care in ways that have substantially benefited theirbottom lines. Let’s look at a few examples of CSPs that have used AI-driven analytics to realizesizeable cost savings, measurably improve customer satisfaction and Net Promoter Scores,accelerate new service rollouts, and drive digital transformation in their organizations.

Page 35: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 35

Detecting irregularbehaviors, improvedQoE, stronger securityA large North American mobilenetwork operator is using AI andanalytics to detect traffic-patternirregularities to improve overall userquality of experience (QoE), shortenthe time needed to troubleshootproblems, and reduce securitybreaches.

Using the Guavus Reflex solution, thisMNO is able to detect unauthorizedtethering use, a situation in which auser with an unlimited data planallows multiple users to piggyback onthat connection with high-usageactivities such as movie downloads.Such use can result in radio-accessnetwork (RAN) clogs, which impactsthe experiences of other subscribersserved by that user’s cell site.Knowing which subscribers are usingunauthorized tethering, the operatorcan suggest a different data plan thatbetter meets that customer’s needs orthrottle usage from that user, ifnecessary, to improve performanceand overall QoE for all subscribers inthat area.

The operator’s security team alsouses Guavus analytics to trackmalware, through the detection oferratic traffic patterns and behaviorsindicative of malware or accessbreach. The security team is able toview a comprehensive dashboardallowing them to query, analyze andgenerate reports on subscriberactivity and then take the appropriateaction. The operator may elect to shutdown a certain Internet port or send amessage to the subscribers’ phonerecommending that he update hisphone with the latest securitysoftware version.

Results: The MNO has reduced thetime needed to discover malwareinfection from days to hours.Additionally, common problemsacross mobile devices are rapidlyidentified by correlating device anddropped call statistics. Feedback isprovided to the appropriate team ordevice manufacturer to correct theproblems before they becomewidespread.

Pinpointing root causesfast Like the North American MSOdiscussed, a European cable operatorhas turned to Guavus AI and analyticsto identify and resolve networkproblems faster, particularly thosethat could impact customer servicelevels. With the cost of a truck roll inthe MSO’s country running about€60 to €70 and the handling ofincoming customer service callsrunning about €5 to €10 each, theprovider implemented Guavus LiveOps as a way to ID root issues quickly– and, at the same time, to reducecustomer service costs and improvecustomer satisfaction.

As the barriers between data siloswere removed, the European MSOdiscovered hidden insights and rootissues. By correlating massiveamounts of disparate data andrunning advanced analytics on it inreal-time, the MSO’s Net Ops andCare Ops teams learned that theycould identify and troubleshoot issuesfaster – sometimes even before theyoccurred.

Results: By reducing customer calls,trouble tickets, and truck rolls, thecompany says it is saving “sevendigits” annually and has improved itsNet Promoter Score (NPS) by a fullpoint.

Which network alarmsindicate real issues? Most CSP network operations centers(NOCs) receive a minimum of onealarm per second (or about 86,400each day). However, only 2% to 3% ofthe alarms actually lead to trueincidents or problems; the rest aresimply noise that can and should beignored. The problem, of course, isdistinguishing which is which – or, ifthat’s not possible, how to arbitrarilydecide which alarms to ignore.

A large North American cableoperator has used the Guavus AlarmIQ analytics to eliminate thisconundrum. Using machine learningand AI, Guavus Alarm IQ takes inalarm streams and classifies whichalarms will lead to customer problems,which ones are already associatedwith open tickets and which ones aremerely noise – with a 99.2% accuracylevel! Now the cable operator canfocus on resolving a much smallersubset of alarms and confidentlyignore the rest.

Results: The operator has reducedalarm volume by more than 90%,while providing increased visibilityinto customer-impacting issuesdeveloping in their network.

Turning big data intogold Hear how Unitymedia is“Driving Value from Big Data inPractice” in this videotapedpresentation by Philipp Gröne,Senior Delivery Expert atUnitymedia.

And read more about howComcast as well as other CSPsare achieving strong resultsusing Guavus’ big data analyticssolutions.

Page 36: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 36

Breaking down big datasilos, succeeding with AIand analytics across yourbusiness Most telecom analytics and AI usecases to date have been focused onstreamlining operations and improvingcustomer care. Early adopters havelearned that getting analytics to workmeaningfully across the variousformats in their vast data lakes wasmore difficult than they anticipated,particularly when using off-the-shelfsoftware. It’s essential to haveanalytics products that crossorganizational boundaries to linkoperations with marketing insightsand customer QoE levels for a holistic– and often real-time – approach toautomation and decision-making.

One of the only vendors to have AI-based solutions that span customercare, marketing, networking, andsecurity operations, Guavus is asupplier to six of the seven toptelecommunications providers andthree of the four top MSOs globally.The company has a long track recordof working with service providers toapply AI and advanced analytics tokey areas across their businesses.

They know that just having big datadoesn’t take you where you need to

go. You need an analytics and AIplatform that brings the importantdata to life – and filters out the datathat’s insignificant. Their team has thelong-standing CSP and real-timeanalytics domain expertise thataddresses CSP challenges at the scale,speeds, and resiliency levels thatcarriers require.

These are the reasons that the CSPs

discussed here have turned toGuavus. As a result, they areimproving customer experienceswhile dramatically reducing costs andattaining the scale and security theyrequire to support the blossominginternet of things (IoT).

To learn more about how Guavus canhelp you get results like these, go towww.guavus.com

About GuavusGuavus is at the forefront of AI-based big data analytics and machine learning innovation, driving digital transformation at6 of the 7 world's largest telecommunications providers. Using the Guavus Reflex® solution, customers are able to analyzebig data in real-time and take decisive actions to lower costs, increase efficiencies and dramatically improve the end-to-endcustomer experience – all with the scale and security required by next-gen 5G and IoT networks.

Guavus enables service providers to leverage both customizable ‘self-service analytics’ and out-of-the-box analyticsapplications for advanced systems planning and operations, mobile traffic analytics, marketing, customer care, security andIoT. Discover more at www.guavus.com

Live Ops: adaptive analytics thatcorrelates separate data sources toidentify and understand customer-impacting events in real time andthen recommends steps to repair.

Proactive Ops: proactivelyanticipates events that may causenetwork problems, identifies whichones will have the biggest customerimpact, and takes automatedactions as needed.

Security Intelligence: automaticallydetects anomalous behavior toshow security analysts wherethreats may be imminent withoutoverwhelming them with falsepositives.

Smart Care: integrates with existingcustomer care systems and

recommends resolutions throughadvanced predictive algorithms andAI.

Marketing Insight: creates customersegments for custom campaigns inreal time to increase acceptancerates.

Smart Industry & IoT: out-of-the-box solution that automaticallypinpoints customer behaviors andpreferences, classifies asset usage,identifies performance issues androot causes, and takes closed- loopactions.

Alarm IQ: harnesses the power ofAI to eliminate alarm “noise” withoutchanging NOC operator workflows.

Guavus AI/ML-powered Analytics Solutions Lineup

Page 37: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 37

e promise of AI is holistictransformation

Throughout Microsoft’s 43-yearhistory, we have focused ondemocratizing the benefits ofcomputing to overcome barriers tohuman progress, enabling everyperson and organization on the planetto achieve more.

Microsoft AI is not a product thatcomes in a box. It is our vision toempower every developer,organization and individual toinnovate and transform the worldwith AI.

Empowering developersto innovateMicrosoft has been creating thebuilding blocks for AI for 25 years andour relentless focus on innovation hasculminated in an explosion ofbreakthroughs. More than everbefore, we are investing heavily inputting this innovation into the handsof developers around the world. Sincethe announcement of Project Oxfordmore than three years ago, which wasthe beginning of what ultimatelybecame Azure Cognitive Services, wepioneered the creation of AI as aservice.

With Azure Cognitive Services, weoffer pre-built AI Services for anydeveloper to use, surfacing innovationand breakthroughs in speech, text,

vision and translation – and theseinnovations are now included in theAzure AI Platform, which bundles thelargest set of pre-built andcustomizable AI services.

It has been thrilling to see the growthin the ecosystem of Microsoft AIdevelopers — more than 1 milliondevelopers are now using AzureCognitive Services. The usage of AIservices and the pace of innovation israpidly increasing. We are committedto helping every developer becomean AI developer with our investmentsin AI School and AI Lab, which aretargeted at making it easy fordevelopers to experiment, learn andcreate.

Empoweringorganizations totransformGartner reports that while 85 percentof enterprises will be using AI by2020, only 25 percent have started orare planning to start their AI initiativesin the near term. The foundation forAI is data. The volume of data isdoubling every year, so it’s no surprisethat the task of turning chaotic dataestates into knowledge is daunting.We consistently hear from customersthat their data is siloed, and much ofit is unstructured, making it hard to

move and even harder to get insightsfrom. In a world where time to marketis key, productivity is king andorganizations are struggling to growand nurture the development skillsneeded in their organization toembrace AI. There are also concernsregarding just how ready AI is forprime-time in an organization wherereliability, security, trust and scale aretable stakes.

Every company is unique and there isno single path to driving AItransformation. Our approach withMicrosoft AI is to meet you whereveryou are on your path to transforming.A major differentiator in our approachis that Microsoft’s data platformenables you to reason over data nomatter where it resides. Whether datais on premise, in the cloud or on theIoT Edge, you don’t need to moveyour data to reason over it. Weempower developers to use the toolsand frameworks they know and lovewith Azure AI, meaning developerscan use skills they already have andcan deploy AI models anywhere, inthe cloud or their own datacenter.The Azure platform is engineered forthe future with the enterprise-gradeservice-level agreements, scale, built-in privacy, compliance and securitythat your organization needs to makeAI real for every application, everybusiness process and every employee.

The era of artificial intelligence is upon us and has the potential to transform our lives,industries and society in ways that may be difficult to imagine today. AI offers us new ways toboost employee productivity and creativity, increase business agility, improve customerengagement and jumpstart new product innovation – and that’s just the beginning of what ispossible.

Page 38: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 38

Redefining the intelligentsoware-defined telco through AIAt Microsoft, we believe that theenterprise can achieve far more witha comprehensive AI strategy ratherthan incremental changes throughisolated use cases. Our vision for theenterprise is to enable every companyto transform by bringing AI to everyapplication, every business process,and every employee—and as a result,achieves more than they ever thoughtpossible.

Every application Quickly and easily develop intelligentapplications to create engaging userexperiences and surface unprecedentedinsights. Just like the internet explosioncatalyzed organizations to createcorporate websites, companies willstart to infuse AI into existingapplications and create completelynew applications with cognitiveservices. Through this infusion, weforesee companies revolutionizinghow they engage externally withcustomers and internally withemployees. For example,breakthrough innovations areenabling organizations to push theboundaries of what’s possible withconversational AI. Customers such asVodafone, Telefónica and TIM areusing Microsoft AI to customize andpersonalize their customer andemployee engagements, thustransforming the telecommunicationsindustry.

Every businessprocess

Enhance every business process withintelligence to expand customerengagement, optimize operations, andimprove products and services. At the other end of the spectrum,organizations are looking to AI to helpre-invent and re-define complexbusiness scenarios. Microsoft’sapproach is to empower complexbusiness process transformation with

enterprise-ready, out-of-the-box, AI-infused solutions, like the newDynamics 365 AI for CustomerService Insights. With a robustecosystem of 300,000 partnersaround the world, Microsoft haspartnered with organizations across awide range of industries to meetvertical-specific needs. Leadingtelecom suppliers like Amdocs,Ericsson, and Blue Prism areembracing Microsoft AI and Cloudservices to improve cuistomerexperiences and optimize operationsacross the datacenter, network,contact center, and more.

Telecom customers are transformingtheir business processes with AI:

Next Best Action (NBA) – AI-powered NBA solutions usesophisticated rules, analytics, andalgorithms to better predict customerneeds and in turn offer more relevantactions and promotions, leading toimproved wallet share and loyalty.

Revenue assurance and fraudprevention – Real-time datastreaming and machine learning makemore accurate fraud, and customermodels possible, enabling fasterdiscovery of revenue leaks andpotential fraud, recovering millions inlost revenue.

Efficient network capacity planning –Anticipate capacity needs andmaximize CAPEX spend across thenetwork, creating just-in-timecapacity that results in cost savingsand improved customer experiences.

Predictive maintenance – Estimatethe remaining useful life for customerpremises equipment, machines,switches, radio-equipment and theircomponents, enabling maintenancetechnicians to be proactive aboutrepairs and reduce costly downtime.

Robotic Process Automation –Streamline and improve manual taskslike order entry with intelligent,configurable automated workflowsthat increase productivity, enableemployees to focus on moreimportant tasks, and reduce OPEX.

Every employee Foster innovation and collaborationacross the enterprise by placing AI in thehands of every employee. Imagine the possibilities available toyour business when the hugeamounts of siloed data across yourorganization become connected andenriched with world data, becomingthe center of gravity for knowledge.Now expand upon this and imaginethe possibilities if you were able toexpose that knowledge to everyemployee in your organization with aconversational AI interface that makesit intuitive and simple to access andquery. It’s about empoweringemployees to participate with a self-service AI approach, helping them tobecome citizen data scientists andenabling them to get informationfaster and in ways that previouslywould have been highly difficult, if notimpossible. We believe the fullpotential of AI is realized when it’struly democratized and for us thatmeans putting it into the hands ofevery employee. This is the nextfrontier of fully democratizing AI.

Transform yourorganization bybringing AI to…

Every application Quickly and easily develop

intelligent applications to createengaging user experiences andsurface unprecedented insights.

Every business process Enhance every business

process with intelligence toexpand customer engagement,optimize operations, andimprove offerings.

Every employee Foster innovation and

collaboration across theenterprise by placing AI in thehands of every employee.

Page 39: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 39

Empowering people totransform societyMicrosoft’s vision of empowermentgoes far beyond business. In fact, theimpact of AI will be the greatest whenused to empower all of us topositively transform society. Societalchange requires a deep commitmentto ensuring the benefits of AI arespread throughout society. While AIprovides us with massiveopportunities, it also creates newsocietal challenges and concerns thatwe must address head on. AtMicrosoft, we’ve identified sixprinciples – fairness, reliability andsafety, privacy and security,inclusivity, transparency, andaccountability – to guide thedevelopment and use of AI in waysthat would increase itstrustworthiness. We established aninternal committee, AETHER, toproactively establish internal policiesand ensure our AI platform andexperience is grounded in these sixprinciples. However, the task ofbuilding responsible and trusted AI isnot the responsibility of any onecompany. We’re actively engaging andparticipating with governments,institutions and associations includingthe Partnership on AI, theOrganisation for Economic Co-operation and Development, and theInternational Standards Organization

to help shape policies and standardsto guide trusted and ethical AIdevelopment.

Through our AI for Good initiative,Microsoft has committed over $115million to AI for Earth, AI forAccessibility and, most recently, AI forHumanitarian Action. Our goal with allof the AI for Good programs is toempower and accelerate the impactthat people around the world canhave in solving some of society’sbiggest challenges.

Get started todayWe recognize, however, that everyenterprise is unique and you will haveyour own path to transforming yourorganization. To help you take the firststep towards your own AItransformation, we have created theAI Ready assessment tool. This toolevaluates your own organizationalreadiness for adopting AI-basedsystems and provides customizedrecommendations around appropriateAI implementations for your business.

The future we create is a choice wemake. Learn more about howMicrosoft is empowering developers,organizations and society to harnessand reap the benefits of AI at ourrecently published Microsoft AI WhitePaper.

Top considerations toensure an AI-readyculture: Adopt a data-driven culture Ensure your AI solutions arefounded on high-quality data.

Share knowledge Commit to breaking down datasiloes across the enterprise andmaking data accessible to all.

Choose the right AI solutionAlign the AI solution to yourunique enterprise and idealbusiness outcomes.

Adapt AI to your enterpriseTest AI with minimally viableproducts and improveapplications and processes thatalready exist—while alwaysputting the customer experiencefirst.

Plan aheadCommunicate the AI strategythroughout your business, beproactive about AI training, andbe cognizant of ethical concerns.

About MicrosoMicrosoft enables digital transformation for the era of the intelligent cloud and intelligent edge. The telecommunicationsindustry is striving to innovate with new services and improve their customer experience, business insights, andoperational efficiency. Microsoft empowers the industry to achieve more with enterprise-class platforms and solutions thattogether with a rich partner ecosystem, help support the mission-critical operations of todayís communications serviceproviders (CSPs). Please visit www.microsoft.com/telco for more information and customer stories.

Company website:https://www.microsoft.com/telco

Privacy policy:https://privacy.microsoft.com/en-us/privacystatement

Page 40: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 40

TM Forum toolkit fordigital transformation

Agile & Virtualized

TM Forum Digital Maturity Model

The DMM is a ‘living’ maturity modeland set of metrics to help companiesmeasure their true digital maturity.

Members can access a guidebook aswell as an Excel spreadsheet

containing the actual model. It is alsoavailable as iOS app.

Agile OSS/BSS ToolkitThis toolkit includes a complete

blueprint for a platform for managinga multi-vendor hybrid/NFV

infrastructure, which includes openAPIs, information models, best

practices and deployment guides.

Open Digital ArchitectureDeveloped collaboratively by the

world’s largest telecom operators andtheir partners, the ODA provides a

common operations and ITmanagement ‘blueprint’. It combines

proven cloud-computing bestpractices with TM Forum’s work onzero-touch orchestration operationsand management; digital ecosystemmanagement; data analytics; AI and

Open APIs.

Open & Partner Effectively

Open APIsTM Forum offers more than 50 APIsto manage services end to end andthroughout their lifecycle in a multi-

partner environment.

Digital Trust Challenges andOpportunities Standard

This technical report outlines the keyconcepts of digital trust and identifiesthe top seven digital trust challenges.

Monetizing the Internet ofEverything Guide

This information guide describes astandardized approach and a

monetization template for new,innovative services.

Customer Centricity

Customer ExperienceImplementation Suite

This set of tools consists of aguidebook, hundreds of metrics, a

maturity model, lifecycle model, ROImodel and more than 54

implementation use cases.

Big Data Analytics SolutionSuite

This set of tools includes a big datareference model, a guidebook

containing more than 65 use casesand 1700+ pre-defined metrics.

360 Degree View of aCustomer

This guidebook offers a 360-degreeview of a customer and explains how

to put customers at the center ofconsiderations and actions.

Page 41: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 41

Toolkit

Research & Media

Digital TransformationTracker 3Why is networktransformation sodifficult?

Digital TransformationTracker 1e race is on

Trend Analysis ReportAI: e time isnow

Trend Analysis ReportVision 2020:Future CSPBusiness models

Trend Analysis Report5G monetization:Operationalimperatives

White PaperAI & customerexperience:Emerging bestpractices

Quick InsightsMicroservices:Piecing together astrategy

Quick InsightsData analytics &AI: Key to end-to-end management

Digital TransformationTracker 2How to fix thecultural divide

Quick InsightsWant to drivebusiness benefits?Improve customerexperience

Quick InsightsSmart analytics paydividends acrossthe customerlifecycle

Quick InsightsBuilding a datalake to drive digitaltransformation

ebookPlatforms: How tojoin the revolution

ebookTM Forum OpenAPIs: Enabling azero-integrationAPI economy

Page 42: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 42

TM Forum Frameworx

TM Forum Frameworx is a suite of best practices andstandards that, when adopted, enable a service-oriented,highly automated and efficient approach to businessoperations. Frameworx provides hundreds of standardizedBusiness Metrics that allow for benchmarking, as well as asuite of interfaces and APIs that enable integration acrosssystems and platforms. Frameworx also includes adoptionof best practices to help companies implement and use thestandards and management best practices to ensureongoing conformance.

Frameworx has been widely adopted and proven tosignificantly improve agility in IT and operations, resultingin increased margins, lower costs and optimal customerexperience. Frameworx is created and evolved by TMForum members who participate in the Forum'sCollaboration Community.

8 things Frameworx can do for you:

1. Reduce transformation risk by delivering a provenblueprint for agile, efficient business operations

2. Innovate and reduce time-to-market with streamlinedend-to-end service management

3. Create, deliver and manage enterprise-grade servicesacross a multi-partner ecosystem

4. Improve customer experience and retention usingproven processes, metrics and maturity models

5. Optimize business processes to deliver highly efficient,automated operations Download latest files Get training

6. Reduce integration costs and risk through standardizedinterfaces and a common information model

7. Gain independence and confidence in your procurementchoices through conformance certification andprocurement guides

8. Gain clarity by providing a common, industry-standardlanguage

Page 43: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 43

Frameworx

Business Process Framework (eTOM)The Business Process Framework (eTOM) is acomprehensive, industry-agreed, multi-layered view of thekey business processes required to run an efficient,effective and agile digital enterprise.

6 things you can do with the Business ProcessFramework:

1. Create a common language for use across departments,systems, external partners and suppliers, reducing costand risk of system implementation, integration andprocurement

2. Adopt a standard structure, terminology andclassification scheme for business processes to simplifyinternal operations and maximize opportunities topartner within and across industries

3. Apply disciplined and consistent business processdevelopment enterprise-wide, allowing for cross-organizational re-use

4. Understand, design, develop and manage IT applicationsin terms of business process requirements soapplications will better meet business needs

5. Create consistent and high-quality end-to-end processflows, eliminating gaps and duplications

6. Identify opportunities for cost and performanceimprovement through re-use of existing processes andsystems

Download latest files Get training

Information Framework (SID)The Information Framework (SID) provides standarddefinitions for all the information that flows through theenterprise and between service providers and theirbusiness partners.

5 things you can do with Information Framework:

1. Reduce integration costs by adopting standards-basedinformation models and using them in applications andinterfaces

2. Save hundreds of design hours by starting with a matureframework and 1500 entities developed and vetted bysubject matter experts

3. Speed time to market by using well-understoodintegration interfaces based on the InformationFramework, eliminating the need for data translationbetween systems

4. Avoid wasting precious development time on debateswith your team, partners, or vendors by adopting awidely proven, industry accepted, rich and extensibleinformation model

5. Mandate conformance to the Information Frameworkand save time and money during vendor evaluation andprocurement

Download latest files Get training

Page 44: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 44

Frameworx

Application Framework (TAM)The Application Framework (TAM) provides a commonlanguage and means of identification for buyers andsuppliers across all software application areas.

5 things you can do with the Application Framework:

1. Streamline procurement by using common definitionsand language to specify and evaluate solutions

2. Document and then rationalize your applicationinventory during transformation projects or mergers andacquisitions

3. Integrate faster and with lower costs by defining andclearly communicating the functions provided withineach application

4. Reduce custom development costs with modular,standard application requirements

5. Increase automation and efficiency with standard,deployable components

Download latest files Get training

Page 45: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 45

Frameworx

Open APIsTM Forum's 50+ REST-based Open APIs have been developed collaboratively by communications service providers (CSPs),government organizations and their partners. When used internally, the Open APIs help companies transform their IT,increase operational agility and improve customer centricity. Externally they enable end-to-end seamless connectivity,interoperability and portability across complex digital ecosystems.

To date, 42 of the world’s leading CSPs and technology suppliers have signed the Open API Manifesto publiclydemonstrating their endorsement of TM Forum’s Open APIs. CSPs that adopt the Open APIs can position them as apreferred requirement in their IT requests for proposal, and technology partners can commit to using the Open APIs inrelevant product applications. Together they can unlock many growth opportunities, including dramatically improvingbusiness and IT agility, reducing the cost and complexity of operations, and reducing integration cost, risk and time for theentire supply chain.

The Open APIs are often tested, improved and extended through TM Forum’s Catalyst Program. Catalysts are proof-of-concept projects that bring together companies large and small to create innovate solutions to common challenges,demonstrating how solutions can be achieved by leveraging key TM Forum best practices and standards. Catalyst teamswork on the projects for four to six months before demonstrating them at TM Forum’s flagship events.

Access the Open APIs Learn more

Best PracticesTM Forum members have collaborated to produce an extensive library of standards, best practices, guidebooks, technicalreports and much more covering the most important topics for companies operating in the digital economy.

We have arranged these resources into toolkits by topic. Click on the link below to access the full toolkits and download*all the available resources.

*Downloads are available to employees of TM Forum member companies only. Interested in joining as a member? Clickhere.

Access the Toolkits

Page 46: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

inform.tmforum.org 46

Research & Media Team

Advisor:Aaron Boasman-Patel, VicePresident, AI & CustomerExperience, [email protected]

Report Design:Intuitive Design UK [email protected]

Published by:TM Forum4 Century Drive, Parsippany, NJ 07054USAwww.tmforum.orgPhone: +1 973-944-5100Fax: +1 973-944-5110ISBN: 978-1-945220-41-8

Report Editor:Dawn Bushaus,Managing [email protected]

Report Author:Mark [email protected]

Head of Research & MediaProduct Delivery:Paul [email protected]

Editor, DigitalContent:Arti [email protected]

Global Account Director:Carine [email protected]

Head of Research &Media Sales:Denise [email protected]

Director, SolutionsMarketing:Charlotte [email protected]

Publications MarketingManager:Jan [email protected]

© 2018. The entire contents of this publication are protected by copyright. All rights reserved. The Forum would like to thank the sponsors and advertisers who have enabledthe publication of this fully independently researched report. The views and opinions expressed by individual authors and contributors in this publication are provided in thewriters’ personal capacities and are their sole responsibility. Their publication does not imply that they represent the views or opinions of TM Forum and must neither beregarded as constituting advice on any matter whatsoever, nor be interpreted as such. The reproduction of advertisements and sponsored features in this publication does notin any way imply endorsement by TM Forum of products or services referred to therein.

Page 47: Trend Analysis Report AI AND ITS PIVOTAL ROLE IN ... · TM Forum Frameworx 27 Additional features and resources. In researching this report, we surveyed 65 executives from 37 different

For more about TM Forum’s work on AI please contact AaronBoasman-Patel, Vice President, AI & Customer Experience, via

[email protected]


Recommended