What it does

What it does with the systems you own.

It makes them one sourced record, then answers, forecasts, and recommends on it, and never touches control.

A long row of medium-voltage switchgear panels in a substation building.

Capabilities

Six things it does, in detail.

Each reads from your systems through their own interfaces, resolves identity and time, and stops at the Execute boundary.

It makes one record out of many.

Across SCADA, the historian, GIS, and the asset register, the same physical asset carries a different tag in each. Identity resolution reconciles them, even where no common key exists; bi-temporal validity separates valid time from transaction time.

Identity resolution across systems, even with no shared key
Bi-temporal record: valid time separated from transaction time
Superseded values retained and queryable, not overwritten

One asset ID, one timeline, reconciled once instead of by every analysis.

It shows where every answer came from.

Every value in the record carries its source system and the timestamp it was true (provenance and lineage), so any answer resolves to the exact fact and moment behind it.

Provenance and lineage on every value (source system and timestamp)
A queryable audit trail tied to what was true at the time
Investigation evidence assembled by query, not manual export

Traceability is a property of the record, not a report assembled after the fact.

It reconstructs an event in minutes.

A sequence-of-events reconstruction across SCADA, historian, GIS, and asset records, correlated on the reconciled identity and timeline, with alarm clustering to cut the flood.

Sequence of events correlated across every source
Alarm flood clustered and deduplicated into ranked events
Root-cause chain, each step traceable to a source

The manual stitching an engineer does across systems, run on the reconciled record instead.

It answers in plain language.

A grounded copilot answers natural-language questions from the record and returns the sourced facts behind each answer. The same grounded answers reach third-party copilots through an MCP grounding gateway.

Natural-language queries answered from the reconciled record
Every answer returns its source facts and timestamps
MCP grounding gateway: the same answers under third-party copilots

One grounding layer serves a human surface and a machine surface, read-only.

It forecasts and optimises, and only recommends.

Forecasting and optimisation compute on the reconciled record, bounded by simulation and engineering constraints. Each output carries its evidence and a confidence score. UtiliSys™ never actuates.

Outputs bounded by simulation and engineering constraints
Every recommendation carries its evidence and a confidence score
The Execute step is a handoff to the system that owns the action

Read-only intelligence: it computes and recommends; the owning system executes and owns rollback.

It runs agents that stop at handoff.

Agents run with a declared scope, tool permissions, and approval gates. Each release is scored against replayed historical events before promotion. No agent writes to field gear.

Declared scope and tool permissions, enforced at runtime
Approval gates from notify-only to supervised
Scored against replayed events; regressions block release

Every agent's Execute step is a handoff. It never actuates.

Why it holds together

Ground the record first. Everything above it inherits the trust.

A copilot, a forecast, or an agent is only as trustworthy as the record beneath it. UtiliSys™ reconciles and sources that record first, then runs read-only intelligence on top.

Talk to us

See what it does on your systems.

Bring an incident that took too long to reconstruct. We will run the sequence of events on your own systems, sourced end to end. One working session, no commitment.