Why a Unified Namespace matters for operational AI
A Unified Namespace (UNS) gives AI a shared context for operational data. Connect that context to history and documentation, then decide where the analysis runs and what information may leave your company.
An AI assistant can explain refrigeration. That does not mean it knows whether someone left the cold-room door open.
To help investigate a warming room, it needs current measurements, equipment context and the right operating instructions. Those may already exist, scattered across systems that describe the same equipment differently.
A UNS gives those systems a shared way to describe the operation. That helps people and applications today, and gives AI something useful to work with.
A temperature reading is only the beginning
Consider a fictional cold-storage room where the temperature is rising. Is the door open, a defrost cycle running, or the sensor reporting an old value? The response also depends on what the room contains and how long it has been outside its expected temperature range.
Those facts must refer to the same room and period. Identities, units and relationships still have to be agreed and mapped; a UNS does not do that work for us.
AI does not require a UNS. Its advantage is reusing that agreed context across applications and assistants instead of rebuilding it for each integration. Archives and document libraries remain separate systems, connected by that context. This applies beyond manufacturing, too.
Connecting documents to the operation
RAG, or retrieval-augmented generation, brings document evidence into an answer. In UNS OpenHub, we link document sources to specific parts of the Unified Namespace and have tested combining that evidence with live data. This is private development, not yet a public Assistant release.
For our cold room, the assistant retrieves relevant passages from its procedure and queries operational tools for current values and history. A shared room identity ties both results together, helping the operator check whether the temperature rose when the door opened.
The answer must separate measured facts, document evidence and the assistant's interpretation. The operator should be able to check the readings, open the cited procedure and see where information is missing or stale.
From answering questions to investigating events
Once that evidence is available, why wait for someone to ask? A rule could detect that the room has stayed too warm. An agent could gather related readings and the procedure, then prepare an explanation for review. The rule detects the condition; the agent helps investigate.
Permission to investigate is not permission to act. Corrective actions require their own authorization, validation and safety limits.
Over time, the same history could support forecasting or anomaly detection. A separate predictive model might learn how temperature normally recovers after a door opening and flag an unusual pattern.
That could use statistical methods or a neural network, but an archive alone is not enough. It still takes reliable measurements, operating context and, where needed, labelled outcomes. Testing on later, unseen periods helps check whether the model works beyond the history it learned from.
Local AI: control should be verifiable
That shared context can reveal how a business works. Where should it be processed, and who controls access? Using AI should not require sending sensitive operational data to an external provider.
That is why we built our own assistant runtime in OpenHub to coordinate tool calls and access to operational context. Its configurable endpoint lets us choose a compatible local language model, or an external one where company policy permits. The application controls which tools and data are available.
For a fully local setup, the chosen model and the entire data path, including document processing, embeddings and logs, still need validation. A local endpoint alone is not proof that nothing leaves the company.
Models will change. The meaning of your data, and control over how it is used, should remain yours.