From transformer evidence to human-reviewed APM decisions.

GridAPM is a local-first, offline-capable Windows workbench concept for controlled transformer APM pilots where AI drafts remain reviewable and engineers stay in control.

Five workspaces a pilot team can evaluate.

The tour follows the review path utilities actually need: approved evidence, quality gates, AI assistance, engineering approval, and audit-ready reporting.

1

Evidence intake

Load approved DGA, oil, PRPD, SFRA, thermal, inspection, maintenance, work-order, and criticality evidence into a structured pilot record.

2

Quality review

Check source provenance, timestamps, units, asset identity, missing context, and reviewer ownership before AI drafts are trusted.

3

Agentic AI draft

Use bounded AI to summarize evidence, highlight contradictions, draft questions, and prepare rationale for engineering review.

4

Engineer approval

Approve, edit, reject, or escalate AI output before recommendations become reportable decisions.

5

Evidence pack

Export a pilot-ready package with source links, uncertainty notes, reviewer comments, decision history, and next-step actions.

Built around transformer evidence streams.

GridAPM should be evaluated around the evidence types, review states, and export packages that matter to transformer teams.

DGA and oil quality workspace
PRPD and partial-discharge evidence workspace
SFRA and electrical-test evidence workspace
Thermal/loading and aging context
Health/risk and lifecycle context
Maintenance planning and work-package review
Standards-aware evidence references
Human-reviewed agentic AI workflow

What to inspect in a first pilot.

A strong pilot avoids vague AI claims and makes the workbench visible: source evidence, assumptions, output states, edits, approvals, and gaps.

Product tour boundaries.

Is the product tour describing a production control system?

No. GridAPM is positioned publicly as a local-first transformer APM workbench and controlled pilot scaffold. It does not claim autonomous control or final diagnostic authority.

Can a pilot start offline?

Yes. The public positioning supports local-first and offline-capable evaluation using approved historical evidence before live integrations are considered.

What is the main output of a pilot?

A practical pilot should produce evidence packs, reviewer notes, missing-evidence lists, AI draft rationale, and human-approved work-package context.