ODA explained: Governance is key, especially for AI

At the end of March, TM Forum’s Research & Media team published its first Benchmark report about the Open Digital Architecture (ODA). As an introduction tothe report, this is the third article in a three-part series highlighting one of the sections: ODA explained. Here, we look at the important role governance plays, especially as communications service providers (CSPs) adopt AI. You can read the first article, which focuses on ODA tools for business and information systems, here, and the second article focusing on tools for implementation and deployment here.

Governance is essential to all aspects of ODA, from planning to deployment. It provides the necessary standards, policies and guidelines to reduce risk and ensure security and regulatory compliance. In addition to providing design and transformation guides, metrics, and maturity models, TM Forum provides tools specifically for governance of APIs, AI and data.

“TM Forum’s mission around AI and data focuses significantly on governance: How do you use AI? How do you make sure that rather than just bolting on loads of AI applications and generating more technical debt, you are managing it properly? How do you control the cost? How do you know where to make your investments?” says TM Forum’s Andy Tiller, EVP, Member Products & Services. “The ODA story is really about how all the tools, processes and frameworks are combined with best practice guidance and governance to enable transformation to a cloud-native, digital telco.”

The CSPs and hyperscalers collaborating in TM Forum’s Innovation Hub are looking at how to integrate AI capabilities into the ODA Canvas. “We want telcos to take that extra step forward to consider how to modernize their BSS/OSS with [a native-AI] approach,” says Priya Saxena, Strategic Cloud Engineer at Google. “By integrating AI into the ODA Canvas, it’s not just a plug-and-play model for your infrastructure. It’s also going to be a standardized model for your future-generation, AI-native OSS/BSS.”

Studying AI’s impact on ODA

Indeed, a huge focus for the ODA team going forward will be on understanding what AI-based automation and autonomous networks (AN) mean in the context of ODA.

“We’ve got a nice AN reference architecture which shows intent management functions for autonomous domains negotiating intent with each other, but it’s a reference architecture for design; it’s not yet an implementation architecture,” explains TM Forum’s Tiller. “So, how can I buy an intent management function? What component does it live in? Does it live in all the components? Is it something separate? Those are all questions we need to resolve.”

The team will also look at the impact of agentic AI on the ODA use case library, which provides reference examples showing implementation of various common OSS/BSS processes with ODA Components and Open APIs. Human involvement is currently crucial in designing and operating parts of this process, but agentic AI aims for autonomy. Indeed, in Level 4 autonomous networks, as defined by TM Forum’s six-level taxonomy, AI agents will act on behalf of users or systems.

To learn more about autonomous networks and CSPs’ progress up the levels, read our Benchmark report.

“The impact of agentic AI is first of all to replace the humans in, say, a product ordering process – but potentially to have the process being dynamically defined by machines as well,” Tiller explains. “This is a very different way of thinking about how we orchestrate complex processes, which is why it's taking time for our industry to get its collective heads around the potential impact of AI.”

One of the specific challenges for the teams developing ODA is figuring out how to enable the ODA data architecture to support AI. “We need a modern data architecture that can make data available in real time to AI applications, models and agents,” Tiller explains. “But we also need to think about how the components and Canvas and APIs are impacted.”

Tiller concludes: “This is the future ODA strategy, and members are working on it. There’s no clear path yet, no clear agreement on how it should all be implemented. But that’s our current, exciting work.”

Download the full ODA benchmark report to find out more.