AI in the RAN: From Policy Enforcement to Power

Why the future of telecom is no longer decided by optimisation — but by who defines value

“The future of 5G is no longer a technical discussion — it is a strategic one.”

For years, the evolution of mobile networks followed a familiar logic: faster radios, denser cells, smarter cores. Each generation promised better performance, greater efficiency and new services. Artificial intelligence, at first glance, appears to be the next step in that same lineage — another tool to optimise an already complex system.

But something fundamental has shifted. AI is no longer just improving how networks operate. It is beginning to define why they behave the way they do — and for whom.

Where user value really lives

In a recent discussion around AI-driven telecom architectures, a key technical point was raised with notable clarity. User value, strictly speaking, does not reside in the Radio Access Network (RAN) or even in the core. It lives in the Business Support Systems (BSS): in contracts, profiles, entitlements, charging models and behavioural history.

“The user lives in BSS. The network behaves according to policies derived from that value.”

From this perspective, functions such as PCRF (and later PCF), PCEF and OCS were never about intelligence in themselves. They were enforcement mechanisms — translating business decisions into network behaviour across the control and user planes, all the way to the antenna.

Crucially, none of this is static.

Users move. Their context changes. Their commercial importance fluctuates. Policy must therefore be dynamic, adaptive and end‑to‑end. This is where AI naturally enters the picture.

AI as orchestrator — and something more

Technically, the promise is compelling. AI can correlate real‑time network conditions with user context, predict congestion, adapt policy enforcement and optimise Quality of Experience based on business value.

In that sense, AI becomes the ultimate orchestrator: continuously aligning user value, network behaviour and service quality.

“AI can kick in to control everything and secure CSP assets by matching user value to business objectives.”

At this level, the argument is hard to dispute. AI does what static rules and preconfigured policies never could.

But this is also where the discussion subtly — and decisively — changes.

When optimisation becomes decision‑making

Once AI is no longer just executing predefined policy but dynamically interpreting user value, it stops being a control function and starts becoming a decision‑making layer.

At that point, several new questions emerge:

• Who defines how value is interpreted?
• Which assumptions are embedded in the models?
• How transparent are those decisions — to operators, regulators and users?
• And critically: who owns and governs the intelligence itself?

This is not an abstract concern. In 5G and beyond, AI is increasingly embedded directly into RAN architectures. Different vendors are offering fundamentally different approaches — not just in performance, but in how tightly intelligence is coupled to their ecosystems.

What looks like a technical choice today becomes a long‑term strategic commitment tomorrow.

A strategic fork in the network

This is where the debate moves beyond engineering excellence.

If AI defines how user value translates into network behaviour, then AI effectively defines:

• monetisation logic
• prioritisation and exclusion
• resilience and availability
• compliance and regulatory posture

In other words, AI becomes infrastructure.

And infrastructure, historically, is never neutral.

“The real question is no longer whether AI should control the network — but whose AI, governed by whose assumptions and locked into which ecosystem.”

Why this matters now

The divergence we are seeing between major RAN vendors is not cosmetic. It reflects different philosophies about control, openness and long‑term dependency. For Communication Service Providers, the implications stretch far beyond performance metrics:

• autonomy versus vendor lock‑in
• flexibility versus embedded logic
• optimisation versus governance

This is why AI in telecom can no longer be treated as a feature upgrade. It is a strategic choice that will shape networks — and power structures — for decades.

Conclusion

AI will undoubtedly make networks smarter. But intelligence without governance simply shifts control to whoever defines the models.

The future of 5G — and what comes after — will not be decided solely by faster radios or smarter algorithms. It will be decided by who gets to define value, embed assumptions and author the logic that runs the network.

And that makes AI in the RAN not just a technological evolution — but a question of power.