Facility Operations Workflow Guide

Use this guide to build an AI Agent workflow for facilities, campuses, factories, and utility areas. The workflow helps teams review asset status, understand alarms, prepare inspections, and draft work orders from the same digital twin and operating records used by DataMesh Inspector.

Prerequisites

Requirement	Details
Asset context	FactVerse Platform asset hierarchy, site, floor, area, equipment ID, owner team, and permission boundary.
Operating records	Inspector alarms, inspection notes, work orders, checklists, meter readings, and field feedback.
Data pipelines	DFS or customer connectors that provide signal timestamps, quality flags, and source-system references.
Spatial context	Designer scene or Inspector visualization that places the asset inside the facility and shows nearby systems.
Access control	A scoped API key with the minimum endpoint and scope set required for the workflow.

Endpoint and scopes

Start with /mcp/base/.

Scope	Use
base.read	Read asset context, documents, status, historical records, and related evidence.
base.compute.run	Run approved analysis such as energy calculation, rule evaluation, or operating summary generation.
base.action.write	Create or update an action draft after the operator approves the proposed record.

Keep the first version read-only. Add write scope only when the customer has defined the approval path and audit record.

Workflow steps

1. Set the operating boundary: identify tenant, site, area, asset group, time range, and operator role.
2. Resolve the asset: map the user's request to a Platform asset ID and verify the corresponding Inspector record.
3. Collect evidence: retrieve current status, recent alarms, inspections, work orders, documents, and source timestamps.
4. Use spatial context: locate the asset in the Designer or Inspector scene and check nearby equipment that may be affected.
5. Prepare the summary: separate confirmed facts, likely causes, missing evidence, and recommended next checks.
6. Draft the action: create an inspection or work-order draft only when the workflow has write scope and human approval.
7. Record the result: store approved actions, operator notes, and corrections so later runs can learn from completed work.

Recommended output structure

Section	Content
Operating summary	Site, asset, current status, severity, open alarms, and source freshness.
Evidence	Signals, inspection records, work orders, documents, and scene references used in the answer.
Missing data	Signals, documents, or records that were expected but not available.
Suggested checks	Field checks, safety checks, nearby equipment checks, or document reviews.
Draft action	Work-order or inspection draft with owner, location, affected asset, and reason.

Common failure modes

Symptom	Likely cause	Response
Asset cannot be resolved	Name mismatch across Platform, Inspector, or source systems	Ask for site, location, or equipment ID, then update the mapping table after confirmation.
Summary misses recent activity	Source pipeline is stale or timestamp is hidden	Show the latest available timestamp and route a data freshness check.
Alarm context is too broad	Query boundary includes too many assets or time ranges	Narrow by site, system, severity, and time window.
Work-order draft is blocked	Missing write scope or approval policy	Return the draft content for operator review without writing to the operating system.

Validation checklist

• Runtime tool discovery shows the expected /mcp/base/ tools.
• The response cites asset IDs, source timestamps, and work-record references.
• The workflow behaves clearly when evidence is incomplete.
• Draft actions are separated from approved actions.
• Approved work results are captured for future workflow improvement.