The Data Contract: What Agentic AI Actually Needs
In the first article of this series, we explored why simply adding AI to existing OSS environments has failed to deliver autonomous networks. The issue isn't the intelligence of today's AI models - it's the fragmented operational data they're forced to work with.
But identifying the problem is only the beginning. What does Agentic AI actually require before it can operate a telecommunications network autonomously?
The answer isn't another Large Language Model, a more sophisticated reasoning engine, or more dashboards. Autonomous AI needs a trusted operational foundation built upon a common understanding of reality - a data contract.
Artificial intelligence is only as reliable as the data it can trust
Most AI readiness discussions focus on data quality—whether information is accurate, complete, and clean. Those characteristics matter, but they aren’t enough.
Autonomous AI continuously observes changing conditions, reasons about cause and effect, predicts outcomes, takes action, and validates results. To do this safely, every operational system must describe the same network reality. It needs a data contract.
A data contract isn’t another database, integration layer, or API specification. It’s a shared operational framework that defines how every system describes the network, synchronizes information, and relates operational events over time. Rather than maintaining multiple conflicting versions of reality, every system—and every AI agent—works from a consistent operational truth.
The four pillars of trusted operational data
As we discussed in the first installment of this blog series, operators have invested heavily in OSS integration, but connectivity alone does not create operational consistency. Hundreds of interfaces can still produce conflicting operational views.
The missing element is shared operational context, which is exactly what the data contract provides. Trusted operational data relies on four pillars:
- Spatial and temporal alignment – Information from different systems must be aligned in time to prevent incorrect conclusions caused by stale or asynchronous data.
- A reconciled Digital Twin – AI requires an always-current understanding of devices, services, customers, and topology.
- A causal knowledge graph – AI must understand which events triggered others to identify true root causes instead of reacting to isolated alarms.
- Action-outcome observability – Every automated action must be verified so AI can learn whether the intended outcome was achieved.
Before an AI agent can recommend or execute an action, it must understand the current network state, what changed, why it changed, which customers are affected, what dependencies exist, and whether previous actions achieved the desired outcome.
The data contract provides this trusted operational context, allowing AI to move from probabilistic recommendations to confident autonomous decisions.
Looking Ahead
A trusted data contract explains what AI needs. The next article in our series explores how operators build that foundation through the combination of a real-time Digital Twin, AXON Cortex, and AXON Neura, transforming trusted operational data into intelligent closed-loop operations.
