By Roger Dewey

There is a lot of discussion right now about AI in telecom — AI for network optimization, AI for operations, AI for customer experience, AI for security. Every conference, article, and panel seems to include AI in some form.

But from my perspective, the industry may be focusing on the wrong part of the problem.

AI itself is not the hard part anymore.
We have more than enough data.
We have powerful models.
We have cloud infrastructure.
We have analytics platforms that can generate insights and recommendations.

The real challenge — the part that is much harder and talked about much less — is execution.

If an AI system determines that a device should switch networks, change its connectivity profile, enforce a new security policy, or alter its behavior based on context, how does that actually happen in the real world? How does that decision get executed across millions of devices, operating on different networks, in different countries, with different hardware and different software stacks?

Dashboards don’t execute.
Analytics platforms don’t execute.
Even orchestration platforms often don’t execute directly — they instruct something else to execute.

As networks move toward automation and autonomy, the industry will need to focus much more on the execution layer — the place where decisions turn into real, deterministic actions on real devices and real networks.

This becomes even more important as we move into a world of private 5G, edge computing, and AI-driven operations. In these environments, decisions often need to be made and enforced locally, securely, and in real time — not just analyzed in the cloud and acted on later.

In other words, AI can decide what should happen, but the network still needs a reliable and secure way to make sure it actually happens.

As an industry, we have spent years building the analytics and orchestration layers. The next phase of the industry will be about building and standardizing the execution layer that makes autonomous and AI-driven networks truly operational.

That’s the part of the conversation I believe we should be spending more time on.