Want to future-proof your telecom network? Let’s talk about the future of agentic AI in telecom

If you’ve worked in telecom lately, you know how fast things are moving. Artificial Intelligence has already begun to weave its way through network operations, IT, and customer experience but the next evolution isn’t just about AI models that predict and automate. It’s about agentic AI; intelligent, goal-driven agents that can reason, act, and collaborate autonomously across systems and domains.

Right now, most of the industry sits on the edge of a chasm; experimentation with individual agents is happening everywhere but scaling them across the enterprise, where every employee has a few agents working alongside them and fully autonomous agents operate inside networks, is another challenge entirely.

The future of telecom will depend on our ability to bridge that gap. Agentic AI isn’t about more automation; it’s about redefining how networks and organizations operate, how teams collaborate, and how intelligence flows between humans, systems, and machines.

Why is this all so important?

Because this is our chance to move from automation to autonomation, from one-off robotic tasks to self-governing, continuously improving processes that evolve with the network. This is the compound accelerator that will finally make a difference as it will refactor how telcos operate, opens the door to genuine growth, and delivers sustainable efficiency gains rather than incremental fixes.

For decades, telecom has chased automation, but it’s time to turn that effort into a living system that learns, adapts, and scales. By building autonomation into our foundations, we can achieve an industry-wide transformation that’s pragmatic, repeatable, and measurable and ensure it doesn’t turn into another half-baked, costly experiment.

So, what’s our big idea?

We realised this moment isn’t so different from when we first introduced other new technologies; for example, when we introduced video calling to every desk in the telco and then scaled it to homes worldwide. What made that revolution work wasn’t the gadget on the desk, it was the fact we had built the foundation underneath to support it, with its standards, interoperability, and the shared commitment to scale by design from the start.

Our big idea is to get the foundations right for agentic AI, and do it now, before the sprawl begins. To make sure that, as these agents multiply across customer operations, IT systems, and networks, they do so in a way that’s governed, interoperable, and scalable.

At TM Forum, we’ve been doing this for 30 years: building the frameworks that let the industry move forward together. Now, we’re doing it again, by making sure the introduction of AI is done safely, securely, and sensibly.

Introducing the industry’s AI-Native integration blueprint

Agentic AI is made up of three parts - interaction, reasoning, and execution - and each layer needs a strong foundation and that’s exactly what the 3 workstreams of the AI-native Integration Blueprint will focus on:

• Interactions: Agents don’t live in isolation. They talk to systems, humans, and each other across IT and network environments. That means interactions must be secure, observable, and standardized, ensuring agent-to-agent, agent-to-system, agent-to-human, and agent-to-device communication happens safely and consistently across zones.
  
  

This is what doing within our Agentic Security Interaction workstream which will be extending proven frameworks such as NIST and CSF 2.0 to the world of autonomous agents. We’re not reinventing security; we’re adapting what already works to the new realities of agentic connectivity

• Reasoning: Each agent needs to consume models, invoke reasoning services, and manage tokens that represent real cost and computational value.

Our Model Operations and Services (MODaS) workstream defines how models are stored, versioned, monitored, and consumed responsibly — so organizations can explain, control, and optimize every inference, from the network to the CFO’s dashboard.

• Data: Whether data is used by the model or the agent, it must have a clear lifecycle that captures its business value.

The Data Product Lifecycle Management workstream will define how data is curated, governed, and made discoverable across cloud and on-prem environments this allows for the true democratization of data and giving agents a common foundation to operate on.

So, What Can I Do?

Join one of TM Forum’s three collaboration projects and help shape the foundation of AI-native telecom:

1. Agentic Security Interaction: Define the standards for safe, observable, and trusted communication between agents, systems, and humans across telco zones.
2. Model Operations & Services (MODaS): Design how models are hosted, consumed, and governed, including token metering, lineage, and assurance.
3. Data Product Lifecycle Management (DPLM): Build the data backbone for agents and models to discover, share, and operate on information seamlessly.

By participating, you’ll help shape the standards that determine how agentic AI integrates into real-world telecom environments. You’ll ensure your organization isn’t left behind and that the future of AI in telecom is built with you, not just around you.

At the end of the day, TM Forum’s AI-Native Integration Blueprint isn’t just about technology. It’s about giving our industry the tools, trust, and confidence to embrace AI without fear, to build networks that are safer, smarter, and truly future-proof.