All industry field guides

Applied research 02 · Engineering & field services

Know what the field needs before the field waits.

EPC, maintenance, fabrication, shutdown, and long-term service work succeeds through thousands of connected commitments. AI should protect that chain—from scope and material to crew, permit, quality record, and handover.

Decision domainBid, plan, mobilise, execute, inspect, and hand over
Evidence surfaceContracts, drawings, BOQ, schedules, work packs, QA/QC, field reports
Control boundaryPrepare and coordinate; accountable engineering and site roles authorize

Our point of view

The hidden product is certainty.

A contractor does not sell labour hours or documents. It sells confidence that the right work will be completed safely, to specification, inside a constrained window. Yet the evidence for that promise is distributed across estimating, engineering, procurement, planning, stores, permits, field supervision, quality, and commercial control.

We model the project as a live graph of commitments: scope item to drawing, material, work pack, competency, access, inspection point, progress evidence, and contract obligation. Agents can then prepare decisions across functions without losing the source, revision, or authority behind them.

Design principleA field-services agent should never optimise one crew’s task by creating an unsafe interface, a quality escape, or an unrecorded commercial obligation elsewhere.

This applied research field guide was informed by the public service model of multidisciplinary Gulf contractors, including STS Group. It does not describe work performed for STS or any other customer.

Applied research outputs

Three handoffs where margin and trust disappear.

These research patterns show how scope, readiness, and proof could become computable while leaving technical and contractual authority with the named role.

01

Bid-to-plan continuity

Carry tender assumptions into execution before they become surprises.

This research pattern structures requirements, quantities, exclusions, interfaces, deliverables, and risk allowances into a traceable scope graph that could survive the handoff from estimating to project controls.

The decision

Is the opportunity executable at the proposed price and schedule, and which assumptions must become controls after award?

The evidence

ITT, contract conditions, specifications, drawings, BOQ, vendor offers, clarifications, method statements, productivity norms, and lessons learned.

System behavior

Builds a requirements matrix, compares revisions, detects scope gaps, maps obligations to WBS and owners, and generates cited review questions—not invented estimates.

Human boundary

Estimating, engineering, commercial, and management approve price, contingency, qualification, and contractual position.

02

Turnaround & shutdown control

Re-plan the constraint before it becomes lost shutdown time.

The reference design connects scope freeze, isolation, access, materials, specialist tools, crew competency, permits, predecessors, and emergent work in one readiness and recovery model.

The decision

Which work fronts are truly ready, and what is the safest feasible recovery when inspection changes scope?

The evidence

Work list, critical path, work packs, blind/isolation list, scaffold and crane plan, materials, manpower, permits, NDT findings, and shift reports.

System behavior

Continuously checks readiness, predicts constraint collisions, classifies emergent work, simulates recovery options, and drafts the coordination pack with impact trace.

Human boundary

Shutdown leadership approves scope, sequence, resource movement, schedule baseline, and every change affecting plant or personnel safety.

03

Quality-to-handover evidence

Build the turnover dossier while the work is happening.

This research pattern links installation evidence to tag, drawing revision, ITP hold point, weld/NDT record, test pack, punch item, redline, and client acceptance so completion could be proven rather than reconstructed.

The decision

Is this system mechanically complete, what evidence is missing, and which defect prevents release?

The evidence

ITPs, RFIs, inspection reports, weld maps, NDT, material certificates, calibration, test packs, punch lists, redlines, and vendor documentation.

System behavior

Extracts and reconciles identifiers, validates revision and completeness rules, routes exceptions, and assembles a source-linked dossier by system/subsystem.

Human boundary

QA/QC, engineering, commissioning, and client representatives sign acceptance. The system cannot close a hold point or waive a nonconformance.

Reference architecture

One project memory from contract to close-out.

The architecture follows each commitment through systems and revisions while respecting commercial, technical, and HSE authority.

01 · Capture

Field evidence fabric

Documents, ERP, schedule, mobile forms, photos, scans, and telemetry arrive with time, location, author, and revision.

02 · Connect

Project knowledge graph

Scope, WBS, tag, drawing, material, crew, permit, inspection, progress, obligation, and change are linked.

03 · Decide

Constraint & rule engine

Readiness logic, specifications, ITP gates, competency, sequence, and commercial rules test proposed actions.

04 · Coordinate

Role-bound agents

Estimating, planning, site, quality, and handover agents prepare work in their own permission and approval lanes.

The line we do not cross.

  • No false progressProgress requires the agreed physical evidence and acceptance state—not a generated narrative or unverified percentage.
  • No approval fictionAgents cannot sign permits, ITP hold points, variations, timesheets, completion, or client acceptance.
  • Revision firstSuperseded drawings and specifications are quarantined; conflicts stop the workflow and route to document control.
  • Commercial traceEvery scope interpretation and change preserves notice dates, source clauses, correspondence, owner, and approval state.

Validation before scale

Test it against a finished project first.

Historical projects contain the late materials, revised drawings, emergent work, rejected inspections, and handover gaps a clean demo avoids.

  1. 01

    Back-test

    Reconstruct a completed bid, shutdown, or system turnover and compare the inferred constraints with the actual record.

  2. 02

    Break it

    Inject duplicate tags, superseded drawings, unreadable certificates, multilingual field notes, missing approvals, and changed scope.

  3. 03

    Shadow

    Run on one live workstream with no approval authority; compare readiness calls and evidence packs with the project team.

  4. 04

    Release by role

    Enable bounded drafting and routing only after accuracy, latency, and exception behavior meet the agreed gate.

What earns the right to scale.

Requirement recall
Critical obligations, interfaces, deliverables, and exclusions are captured and tied to their exact source.
Readiness precision
A “ready” call is supported by current material, access, competency, permit, predecessor, and quality state.
Evidence completeness
Dossier elements resolve to the correct tag, revision, inspection point, and acceptance status.
Coordination value
Teams measure reduced search, fewer avoidable handoff loops, earlier constraint discovery, and faster exception closure.

Public sector signals

The operating model studied here.

  1. STS Group overview

    Public sector reference spanning multidisciplinary construction, manufacturing, maintenance, brownfield projects, and energy/industrial services.

  2. STS service contracts

    Shows the full chain from engineering and procurement through construction, commissioning, preventive/corrective maintenance, integrity, and shutdown management.

  3. STS plant maintenance, turnarounds, and shutdowns

    Details mechanical, electrical, instrumentation, static, and rotating-equipment work that drives the page’s field evidence model.

  4. STS HSE approach

    Reinforces why safety authority and verifiable controls must remain explicit in construction, fabrication, and maintenance workflows.

Continue exploring

Related industry research.

Bring us the handoff where the field loses time, evidence, or margin.