Data Analytics & AI

Do you know how you’re going to use AI?

What are the best use cases for artificial intelligence (AI) and machine learning in communications service providers’ businesses? What’s the best architectural approach to implementing AI? Do we need an industry-agreed architecture? If so, what should it look like?

These are questions TM Forum is setting out to answer. At Action Week in September, Forum members took the initial steps toward forming a new collaboration working group on AI and machine learning, which is a specific type of AI.  Potential focus areas for the new team are:

  • defining AI use cases for customer experience, service optimization, network optimization and security, and reflecting the customer journey and business process enhancements needed to implement them;
  • developing a reference architecture for AI and determining how to reflect AI capability in Frameworx;
  • considering how AI impacts digital trust;
  • identifying the cultural changes necessary to embrace AI;
  • articulating how AI can be used in telecom networks to vendors who may be new to supplying the telecom market;
  • explaining how the Forum’s work on AI fits with that of other standards-development organizations (SDOs) and open source communities; and
  • developing an AI maturity model to measure progress.

What are potential AI use cases?

At the Action Week workshop, attendees discussed many potential use cases for AI and the possibility of developing a ‘playbook’ of use cases similar to the Forum’s Big Data Analytics Use Cases guide book.

Potential use cases cited include:

  • chatbots
  • contact center optimization and compliance
  • sentiment analysis
  • voice sentiment analysis
  • knowledge portals for agents
  • fraud detection and prevention
  • self-healing for software-defined networks
  • anomaly detection for OAM&P (operations, administration, maintenance and provisioning)
  • automated resolution of trouble tickets
  • prediction of network faults
  • performance monitoring and optimization
  • service assurance, where AI suggests thresholds and rules
  • churn prediction and prevention
  • 5G network orchestration

Preliminary research shows interest

The TM Forum Research & Media team has been conducting research into AI and machine learning for a Trend Analysis report to be published next month. We asked more than 175 individual respondents from telcos about potential use cases. About a third said their companies are already using or starting to deploy AI in automated customer-facing systems to deliver first-line technical support. However, many others said their companies have no plans to deploy the use cases we asked about.

Separately, we asked about using machine learning and analytics for network management. While some respondents indicated their companies are using AI to predict problems, for self-healing and even to automate policy changes, most are not yet using it for network management.

Developing a blueprint

It’s clearly early days for AI. While operators are beginning to experiment, most have not decided on a specific architecture for using the technology. They haven’t determined whether to deploy a centralized AI platform. Some operators describe this as a single “brain” that controls all customer experiences and handles operations and networking. Others are thinking about deploying lots of specialized solutions and figure out how to connect them when required.

“Is there a reference architecture for AI? What should be the architectural approach? Do we go for point solutions – many vendors have been developing AI versions of their products – or do we want to look at a more open approach where we bring in an AI platform and develop applications on top of the platform?” Vincent Seet, Head of Enterprise Architecture, Globe Telecom, asked during a presentation at Action Week. These are the kinds of issues a working group can help Globe and other CSPs address, he added.

The Forum’s Open Digital Architecture (ODA), formerly known as ODES, is tackling many of these architectural issues. ODA reimagines operational and business support systems (OSS/BSS) for future networks that are automated and rely on AI and machine learning. It’s a data-centric architecture (meaning it uses a consistent and holistic data architecture across all the layers of the architecture) and it implements automated, intent-based management from end to end.

The 80-20 rule

Globe has been experimenting with AI in customer-facing scenarios as part of the TM Forum Catalyst Program and is one of the most vocal proponents of the new TM Forum working group. During his presentation, Vincent explained that at Globe AI use cases typically can be categorized using an “80-20 rule”, which means that 80 percent of the time they can be developed on top of and be supported by a centralized platform. The other  percent of the time, they need to be tightly coupled point solutions.

He then presented the reference architecture below as an example of something he’d like the working group to address.

“We feel the AI platform should be divided into two parts, one is the platform services where you really see the manifestation of the AI…and then the underlying layer that’s providing the foundation of services – this is machine learning, whatever you have learned and put into knowledge management, and of course AI integration with the other systems in the enterprise.”

In the future, Vincent noted, nearly every system in a CSP’s environment will use AI, so the question becomes: What is the best way to manage them together holistically?

How does AI impact trust?

CSPs also must consider the effects of AI on digital trust. In July, TM Forum published a new technical report called Digital trust challenges and opportunities, which includes seven trust challenges CSPs and their partners face. One of them is: How can businesses establish safeguards and provide transparency to digital trust decisions based on AI and machine learning?

Autonomous digital trust decisions made without human involvement need safeguards to protect against incorrect or questionable decisions. Using AI and machine learning to make these decisions increases the difficulty of ensuring proper decisions, because of the complexity, dynamic nature and potential opacity of these decision-making technologies.

We need better digital decision-monitoring capabilities and decision-tree traceability. It is especially important that customers can ask for the rationale behind decisios to gain confidence in the technologies.

Culture shift

The working group also must address the huge cultural change that necessarily comes with implementing AI technology. The volume and velocity of changes in software-defined networks rely on automation and use of AI, which will have a major impact on telcos’ culture .

The will have to abandon tried-and-true manual processes and lose many staff: Some CSPs have said publicly that they expect automation to eliminate up to a third to half of jobs.

“We have to change people’s mindsets,” Vincent said. “When people hear about AI, they have a fear of their jobs being taken away. Are we really talking about replacing humans immediately? Or is it about complementing the human in their role and making it more simplified for them? What will be the new competencies required for us to actively employ AI in the organization?”

AI ambassadors

Finally, the Forum could be an AI ambassador of sorts, explaing telcos’ needs to suppliers that may not be familiar with telecoms business processes, and to other SDOs and open source communities that may be working on other parts of the future telco architecture.

Many suppliers of AI technology are startups that have no idea how their solutions could be used in telecom businesses. It’s important to show them why telcos are potential buyers. Some workshop attendees suggested holding a hackathon for AI developers or “brain challenges” for companies to participate in. The idea would be to provide an AI simulation environment where suppliers could create and test solutions, possibly in Catalyst projects.

Similarly, Forum members want to connect with and educate other SDOs and open source groups about our activities to avoid duplicated effort and take a collaborative approach to solving telcos’ AI challenges.

What do you think about AI? Join the TM Forum Community to discuss your ideas – it’s open to members and non-members, alike. Members who’d like more information about participating in the new AI working group should contact Andy Tiller.


    About The Author

    Managing Editor

    Dawn Bushaus began her career in technology journalism in 1989 at Telephony magazine, which means she’s been writing about networking for a quarter century. (She wishes she didn’t have to admit that because it probably gives you a good idea of how old she really is.) In 1996, Dawn joined a team of journalists to start a McGraw-Hill publication called, and in 2000, she helped a team at Ziff-Davis launch The Net Economy, where she held senior writing and editing positions. Prior to joining TM Forum, she worked as a freelance analyst for Heavy Reading.

    Leave A Reply

    Back to top