Your Plant's Institutional Knowledge, Finally Findable
Operating plants accumulate decades of critical asset information — buried in legacy PDFs, scanned drawings, outdated formats, and personal hard drives. Asset Intelligence ingests that chaos, extracts structured metadata, builds tag-level references, and makes every piece of engineering knowledge queryable in seconds.
“Show me the maintenance history and vendor manual for pump P-101 — urgent, 2am callout.”
Every answer cites the document it came from, so the engineer can open the source rather than trust a summary.
The failure modes
Four Ways Plant Asset Knowledge Gets Lost
Operating plants generate enormous volumes of critical asset information across decades of operation — but four recurring failure modes make that knowledge inaccessible when it's most needed.
Unstructured Digital Documents
Thousands of PDFs, scanned drawings, and legacy files with no consistent naming, no metadata, and no cross-references to the equipment they describe.
Outdated & Inaccessible Formats
TIFF scans, hand-annotated drawings, and obsolete file formats that modern systems cannot read — critical data locked away in formats nobody can query.
Critical Information Not Traceable
Maintenance histories, inspection certificates, and operating procedures exist somewhere — but nobody can find them when a decision needs to be made under pressure.
Not Referenced to Tag Numbers
Drawings, manuals, and specifications filed by project number or date — not by the equipment tag they describe — making equipment-level retrieval impossible.
Six core capabilities
What the Asset Intelligence Agent Delivers
Every capability addresses a specific failure mode in how operating plants currently manage their asset information — turning a liability into a live, queryable knowledge base referenced to every equipment tag.
01
Metadata Extraction from Document Content
AI parses every document — regardless of format, age, or naming convention — extracting project codes, document numbers, revision levels, discipline, system, and equipment references without manual entry.
02
OCR Title-Block Extraction from Drawings
AI identifies title-block locations on scanned drawings from images, extracts all fields using OCR, and populates the document register — turning illegible image files into fully indexed, searchable records.
03
Equipment Tag-Level Document Referencing
Every document is linked to the equipment tags it references. Access any tag and instantly retrieve all related drawings, datasheets, inspection records, maintenance histories, and vendor manuals.
04
Custom Attribute & Structured-Data Creation
User-defined attributes — manufacturer, model, capacity, material spec, operating parameters — identified and extracted from document content, creating a structured equipment database from unstructured sources.
05
AI Instruction-Manual Generation
GenAI synthesises content from vendor manuals, operating procedures, and maintenance records into a coherent, current instruction manual for any equipment item — preserving institutional knowledge that would otherwise leave with retiring staff.
06
Natural-Language Query Across the Knowledge Base
Operations teams query the entire asset knowledge base in plain English — "what are the last three inspection records for vessel V-305?" — receiving answers with direct links to source documents, not just search results.
Natural-language query interface
Ask Your Plant's Knowledge Base Anything
Operations team members query the full asset knowledge base in plain language — and receive answers with direct links to source documents. No training. No search syntax. Source references on every answer.
More example queries
How the agent works
From Legacy Document Heap to Queryable Asset Knowledge
A four-stage pipeline ingests whatever format your legacy documents exist in, extracts metadata using AI, builds equipment-tag relationships, and makes everything accessible through a natural-language query interface.
Multi-Format Document Ingestion
PDF, Word, Excel, TIFF, scanned drawings, and legacy image formats ingested via bulk upload — no pre-processing or manual conversion required.
OCR for Non-Searchable Content
Optical character recognition extracts text from title blocks, annotation fields, revision tables, and equipment schedules invisible to standard AI.
Filename Nomenclature Parsing
AI parses filenames against configurable naming standards — Project Code, Document Number, Revision, Discipline — without manual entry.
Content-Based Metadata Extraction
Where filenames are incomplete, AI reads document content to extract metadata — identifying document type, subject equipment, system, and project phase.
Drawing Title-Block Extraction
AI locates title blocks on engineering drawings, extracts all structured fields via OCR, and stores them as searchable metadata — eliminating manual register entry.
Equipment Tag Reference Mapping
AI identifies all equipment tag numbers across every document and builds a bidirectional reference: every tag knows its documents; every document knows its tags.
Cross-Document Relationship Graph
P&ID references, datasheet cross-references, and spare parts lists tied to equipment — access any document and see all related documents across the plant.
GenAI Instruction-Manual Synthesis
AI synthesises content from vendor manuals, operating procedures, and maintenance records into a coherent, up-to-date instruction manual for any equipment item.
Four-stage pipeline structure
- Stage 1 — Ingestion & OCR:any format, any age accepted
- Stage 2 — AI Metadata Extraction:content parsed, not assumed
- Stage 3 — Relationship Building:documents linked to tags
- Stage 4 — Query Interface:NL questions with source refs
Tech foundation
- OCR + Image AI for legacy formats
- NLP & NER for metadata extraction
- Knowledge Graph for tag references
- RAG + GenAI for queries & synthesis
Why switch?
Asset Intelligence vs Conventional Document Management
Every dimension of plant knowledge retrieval is transformed — from how engineers find information at 2am, to how MOC reviews assemble document registers, to how institutional knowledge survives retirements.
Document Retrieval
Critical at 2amEquipment-Tag Linkage
Topology GraphLegacy Format Access
Scans & TIFFsMetadata Population
GovernanceKnowledge Preservation
Institutional IPMOC Document Register
Audit SpeedOn-Call Information Access
Mobile ResponseReal-world scenarios
What Changes for Operations Teams
Three situations every plant operations team faces — and how the Asset Intelligence AI Agent changes the outcome. Each scenario draws on the same underlying capability: tag-level retrieval over a structured knowledge graph.
Incident escalationA Critical Pump Fails at 2am
emergency riskThe human bottleneck
The on-call engineer needs maintenance history and vendor manual immediately — and the call to bypass a safety interlock can't wait 90 minutes for someone to find the right PDF.
Query: "All maintenance records and vendor manual for P-101"
AI agent real-time intervention
Institutional IPA Senior Engineer Retires
years knowledge survivalThe brain drain hazard
Twenty-five years of institutional knowledge would walk out the door — annotated personal drawings, undocumented Excel files, hard-won operational context that prevented near-misses for a decade.
Every document the engineer worked with already ingested
Permanent topology indexing
Compliance assuranceA Management-of-Change Review
3 days → HoursThe archive bottleneck
An MOC review requires a complete document register for the affected system — every P&ID, datasheet, and inspection cert across 20 years. The document controller would otherwise spend three days searching five archive locations.
Queried by system or tag number
Deterministic compliance evidence
Why it matters
Four Outcomes for Operating Plants
The Asset Intelligence agent doesn't just speed up document search. It converts an unmanaged liability — decades of fragmented asset knowledge — into a structured, queryable, compliance-supporting resource that the entire operations team can use.
Knowledge Preservation
Institutional knowledge that currently resides in individual engineers' heads and personal stores is captured, structured, and made accessible — permanently. People retire; the knowledge doesn't.
Immediate Information Access
Any engineer — on-call at 2am, at a remote site, or in a management review — can retrieve any document or operating procedure for any equipment tag in seconds, from any device.
Operations & Safety Compliance
Complete, traceable document records for every equipment item — inspection certificates, test records, maintenance histories — means audits and permit-to-work are supported by evidence, not memory.
Full MOC Compliance
A queryable, tag-linked document base makes full Management-of-Change compliance achievable rather than aspirational — every affected document retrieved automatically, not manually assembled.
Who uses it
Built for Every Operations Stakeholder
On-Call Engineers
Retrieve maintenance histories, vendor manuals, and inspection records in seconds — from a mobile device, at any hour, referenced to the equipment tag.
Maintenance Teams
Tag-linked maintenance histories, vendor procedures, and AI-synthesised instruction manuals — no more searching shared drives for the right datasheet.
Operations Managers
A single queryable knowledge base across decades of plant history — supporting daily operations decisions with structured, traceable information.
Document Controllers
Automated metadata extraction, OCR title-block population, and bulk-upload numbering — eliminating manual register entry on every document.
MOC Reviewers
Generate complete document registers for any modification scope in minutes — cross-referenced, demonstrably complete, ready for review.
Plant Managers
Institutional knowledge captured, structured, and made accessible — protecting the plant against retirement-driven knowledge loss and audit exposure.
What's coming next
From Knowledge Retrieval to Predictive Operations
Today the agent delivers structured knowledge retrieval. The roadmap takes that foundation — a tag-referenced, structured operations data layer — into live operational intelligence and digital-twin capability.
Predictive Maintenance from Historical Data
Maintenance histories, inspection records, and failure-event data — now structured and tagged — feed predictive models that identify which equipment is approaching failure before it happens. Reactive maintenance becomes scheduled before it becomes emergency.
Real-Time Operational Decision Support
The structured asset knowledge base provides live context for operational decisions — when a process parameter deviates, the AI surfaces the relevant procedure, the last inspection result, and the applicable safety instruction simultaneously.
IoT Sensor Integration & Live Digital Twin
Asset records enriched with real-time sensor data — vibration, temperature, flow, pressure — linked directly to equipment tags and document history. The knowledge base becomes a living digital twin of the operating plant.
The business case
Why Plant Knowledge Management Needs AI
Operating plants accumulate critical knowledge over decades — but the way that knowledge is stored makes it functionally inaccessible at the moments it matters most. Manual document management cannot solve this; structural change is required.
2am Maintenance History Access
2am call-outs decided without the full maintenance history — because nobody can find it in time
Tag-level retrieval in seconds, from any device, at any hour — full document set on demand
Senior Engineer Retirement IP
Senior engineers retire taking 25 years of plant knowledge with them — handover captures 20%
Institutional knowledge captured, structured, and synthesised into AI instruction manuals
MOC Register Generation
MOC reviews start with document registers everyone knows are incomplete
Complete MOC document registers generated in minutes, cross-referenced, demonstrably complete
TIFFs, Scans & Paper Records
TIFFs, scans, and paper records effectively unreadable by any modern system
OCR + AI extraction makes every legacy format queryable — TIFFs, scans, paper records included
Operating Procedures Under Pressure
Inspection certificates and operating procedures exist somewhere — but not findable under pressure
Natural-language queries answered with direct links to source documents
Permits & Compliance Verification
Compliance audits supported by memory and partial registers, not evidence
Audits and permit-to-work backed by complete, traceable evidence — not memory
Frequently asked questions
Everything You Need to Know
Ingestion, human supervision, tag linking, manual generation, and where this agent sits in the wider Wrench AI suite.
What document formats can the agent ingest?
Any format — PDF, Word, Excel, TIFF, scanned drawings, hand-annotated images, and legacy image formats. OCR is applied automatically to non-searchable content like title blocks, annotation fields, and equipment schedules. Paper documents are also supported via the physical-document digitisation pipeline. No pre-processing or manual conversion required.
Does the agent make operational decisions on its own?
No. The Asset Intelligence agent retrieves, synthesises, and presents — engineers verify and act. Human supervision is built into the architecture: engineers validate AI-extracted metadata, approve AI-generated instruction manuals before they become controlled documents, and retain authority over all operational and safety decisions. The AI provides the information; engineers make the call.
How are equipment tags identified and linked across documents?
The AI identifies equipment tag numbers wherever they appear across every document — drawings, maintenance records, inspection reports, vendor manuals — and builds a bidirectional reference in a knowledge graph. Every tag knows all its associated documents; every document knows all the tags it references. Cross-document relationships are then built on top: P&ID references, datasheet links, and spare-parts ties to equipment.
How does AI instruction-manual generation work?
Generative AI synthesises content from multiple source documents — vendor manuals, operating procedures, inspection records, and maintenance histories — into a coherent, current instruction manual for any equipment item. Source references are preserved. Engineers review and approve before any manual is issued as a controlled document. This is how institutional knowledge that would otherwise leave with retiring staff gets preserved.
How does it fit with the rest of the Wrench AI suite?
Asset Intelligence handles operating-plant knowledge retrieval. Its siblings in the WAA suite cover capital-project domains: the AI Analyzer (conversational queries on PMIS data), the AI Control Tower (predictive project governance), and the AI Reviewer agents (Technical Reviewer for engineering documents, Bid Analyzer for tenders, Contract Review for commercial governance).
Is the agent useful before our document library is fully cleaned up?
Yes — that is exactly its purpose. The agent is designed to ingest your library as it currently exists: inconsistent filenames, missing metadata, TIFFs alongside modern PDFs, paper records mixed with digital. Its first job is to extract structure from that chaos. There is no clean-up prerequisite; the agent does the clean-up as it ingests.
Stop searching for documents. Start querying your plant.
Twenty years of drawings, certificates and vendor manuals, answerable in one question — with every answer citing the document it came from.