The atmosphere at FutureNet World 2026 last month in London was ambitious, but if you listened carefully, there was a clear underlying anxiety.
A thousand senior telecom executives in London, two days of sessions, and one theme dominating every conversation: AI-driven autonomous networks. But the most telling sign of the times wasn't the technology, it was the audience. They welcomed double the C-level executives compared to last year and saw a 25% surge in senior decision-makers.
When the boardroom shows up in force, you know the stakes have shifted. But if you listened past the keynotes and into the hallways, or the roundtables, a different, more uncomfortable question kept surfacing:
If everyone is betting on AI, why is no one confident it’ll work in production?
Netcracker even ran an entire panel on it: "Autonomous Networks: Separating the Myth from the Reality.". Not exactly a title that screams confidence in the current state of practice. Because here’s what FutureNet World 2026 actually exposed: the industry is building an incredibly sophisticated intelligence layer and has forgotten to build a reliable foundation underneath it.
The gap between insight and action
The industry has made genuine progress on AI’s ability to observe and recommend. AIOps platforms can now detect anomalies faster than any human team. The intelligence layer is maturing quickly.
What’s lagging is the execution layer, the part of the system that takes an AI recommendation and turns it into a verified, controlled, reversible action on a live network. This is the Autonomy Chasm. Most operators are stuck here and not because their AI isn't smart enough, but because their infrastructure isn't ready to act on AI decisions safely at scale.
But the execution layer isn't the only thing missing. In conversations with AI engineers on the exhibition floor, a more surprising problem kept coming up: the data itself. Raw data without service context is nearly useless to an AI — it has no way of knowing what a deviation means, or what action is appropriate. Reverse engineering that context after the fact adds enormous complexity.
The smarter approach is to treat AI as something you earn the right to expand into. Encode your best operational knowledge into your infrastructure first, so it handles immediate problems reliably and then decides, service by service, where AI is likely to find deeper patterns that justify the investment.
Prevent hallucinations with deterministic execution
When an Agentic AI builds an operational model, it relies on the data it receives/gathers and the APIs it is exposed to for its operational model. On a carrier-grade network, a hallucinated operation isn't a software bug to be patched in the next release. It's a live network action, potentially irreversible, potentially catastrophic.
The engineers who know your network best can't be in the loop for every AI-driven decision. At Inmanta, we encode their expertise directly into intent-based service models, the guardrails that ensure when an AI agent acts, it acts within boundaries that reflect real operational knowledge, not just what the API technically allows. The result is deterministic execution: precise, verifiable, and predictable by design.
Control is not the opposite of autonomy
There’s a tendency to frame control and autonomy as competing goals. In reality, true autonomy depends on increased control. By using self-healing mechanisms that detect and correct drift without human intervention, we move from "hoping" a script worked to "knowing" the intent is satisfied and maintained.
Where the industry goes next
FutureNet 2026 made one thing clear: the appetite for autonomous operations is no longer just a lab interest, it is a C-suite mandate. The question isn't whether AI belongs in network operations. It does. The question is whether your execution layer is robust enough to handle it.
The networks that get there first won't be the ones with the most advanced AI. They'll be the ones that built the right foundation beneath it.
Ready to cross the Autonomy Chasm?
Download our latest report, The Path to True Autonomy, to learn how to build a deterministic foundation for the AI-driven era.