Over the past year, I’ve had more conversations about AI than I can count—with operators, cloud vendors, platform builders, and startups alike. And while the excitement is real, so is the confusion. Everyone wants to “do something with AI,” but few telcos know where to start—or more importantly, how to make it real inside their networks, business units, or customer offerings.

What’s become clear to me is this: the biggest AI opportunity in telecom is not ChatGPT clones or flashy dashboards. It’s enterprise-grade, edge-deployed, sovereign AI that actually runs where the action is—on devices, in networks, and inside the telco data center.

The gap between what telcos want and what’s on offer

Most telcos I speak to are hungry for AI—but cautious. They’ve been burned by pilot fatigue, worried about data sovereignty, and unsure who to trust when every vendor suddenly claims to be “AI-first.” But here’s what they really want:

Practical automation: real-time fraud detection, call summarization, customer experience orchestration, predictive ops. Data control: the ability to train, test, and run models on their own infra—not send data to some mystery region. Integration-ready tools: AI that plugs into what already works—IMS, Open Gateway APIs, RAN nodes—not weeks of glue code.

In short, they want agentic AI that works like infrastructure, not like magic. That means inference under 10ms, compliance built-in, and modular designs that work across different networks, vendors, and clouds.

A new wave is forming

This is where I believe the next big wave is forming. Not in some futuristic AGI lab, but in modular, containerized AI that telcos can deploy at the edge, on sovereign cloud, or even directly in the IMS path. The tech stack to do this is just coming together—low-latency silicon from Nvidia, Intel and AMD, platforms like Azure Operator Nexus, telco-native SDKs from players like Radisys, and frameworks like the one at TGDF that help compress, containerize, and audit AI for real-world deployment.

We’re already seeing this come to life:

  • A voice AI agent that plugs into a live call via IMS core.
  • A planning agent that predicts and prevents RAN alarms in under 5ms.
  • A privacy-compliant customer assistant running in a telco’s national cloud.

These aren’t proof-of-concepts anymore—they’re the beginning of a real shift.

Why this matters

AI won’t transform telco unless it’s built for telco. That means understanding not just the tech, but the business, compliance, and cultural layers too. And that’s why I’m excited. Because we finally have the pieces to do this right—and the urgency to act.

We don’t need one giant player to own the whole stack. We need the right coalition of partners, each bringing strengths:

Intel and Microsoft enabling sovereign edge compute. Radisys unlocking access to the voice and messaging core. TGDF translating enterprise AI into telco-ready, agentic systems. And others.

Put together, this is where telco AI gets serious—and starts making a difference. For operations, for customer experience, and for new revenue at the edge.

Let’s build it.