EnginAI empowers engineering teams with AI-driven insights across data, documents and designs — from concept to operations.
Live product Equipment tags located on the drawing, stored per project, feeding MTO and Revision Compare downstream.
Engineering organisations hold decades of value in P&IDs, specifications, revision histories and project correspondence — and almost none of it is searchable. Time that should go into engineering goes into finding things.
Cloud AI tools solve the search problem but fail the security review, because they take the decision about what leaves your network out of your hands. The tool that answers the question is the tool that cannot be approved.
An on-premise engineering intelligence platform. You set the boundary.
Your drawings, specifications and project knowledge live on infrastructure you own and control — no cloud tenancy, no shared platform. Suitable for air-gapped and security-restricted environments.
Two deployment modes, your choice per project
All inference on local models inside your network. No external API calls. No data egress.
Optionally route selected workloads to external models where you permit it, under your own policy.
Documents, drawings and project records become one queryable body of engineering knowledge — with the review gates and provenance an engineering organisation already runs on.
Ask one question, get one grounded answer with visible cited sources — drawn from the active project's ingested documents and its live platform data. The retrieval pipeline combines vector retrieval, structured-module retrieval over real project records, glossary expansion for engineering terminology, HAZOP handling, guardrails and output filtering. Voice dictation via built-in speech-to-text.
ANSWERS STRICTLY SCOPED TO THE ACTIVE PROJECT
The project repository nearly every other module reads from. Upload PDFs, specifications and audio; an automatic extract → chunk → embed pipeline builds the knowledge base with live per-file status, and audio is transcribed first. Documents are classified and assigned to a work package.
DRAFT → REVIEW → APPROVED → ISSUED
Turns extracted drawing tags into a material take-off: line items with category, description, tag, quantity, unit and a confidence score. Lines can be verified individually, filtered, and exported to CSV or Excel for the material plan.
Vision-model reading of engineering drawings. Tag extraction with bounding boxes locates equipment tags on the drawing and stores them per project, feeding MTO and Revision Compare downstream. Drawing-type classification sorts the sheet. Handwritten-markup transcription reads red-pen annotations off scanned drawings.
AI-assisted detection on dense P&IDs — engineer review required
Two revisions of the same drawing side by side, with additions, removals and modifications computed from the extracted tags rather than from pixels — plus an engineering-impact assessment of what the change actually means.
P&ID ↔ GA consistency runs a deterministic tag-set comparison between two drawings that should describe the same system, with an optional AI narrative over the top. Multi-discipline clash detection finds bounding-box overlaps between drawings from different disciplines.
An AI compliance sweep across the project's documents, producing findings against named engineering standards — each with a clause reference, a severity and a suggested remediation — tracked through to closure.
Safety-critical findings require an Approver — enforced at the API
AI Reports generates documents in five templated types, moved through draft → in review → locked and then exported; once locked, content is read-only and the report cannot be deleted. Timesheet Reports run a filtered query over logged time, exported to CSV or XLSX.
Discipline checklists generated for a specific document from a template library keyed on three independent dimensions — document type, discipline and project phase — then augmented with AI-suggested items. Every item is tickable, with progress tracked as a fraction.
A hierarchical blocks → work packages structure organising scope, documents and deliverables, with work-package status tracked separately from structure — Leads own structure, Approvers set status.
Live product Additions, removals and modifications computed from extracted tags — not from pixels.
A full conversational AI engine embedded directly in the platform — alongside, and distinct from, Knowledge Search.
Live product Every claim carries its source back to a project document.
Time, tickets, teams and clients in the same platform as the engineering intelligence — under the same identity, the same roles and the same server-side rules.
A live time clock, a month calendar carrying submission state, and the entry form with its table. The month lock is absolute: once submitted or approved, the period is immutable for everyone — including an organisation Administrator.
The monthly approval queue owned by the Workforce Manager, in four tabs: pending submissions, outstanding, approved history and rejected history. A rejection reopens the period for correction.
The project help desk. Every issue is routed to a receiving team and escalated up a configured team chain until completed, with every transition written to a per-ticket escalation history that stays visible for the life of the record.
PENDING → ESCALATED → COMPLETED
Client records with an activate / deactivate lifecycle; team structures with membership and the escalation tiers that underpin ticket routing.
A cross-project personal column — my tickets, my hours, my projects, independent of the active project — beside a project-scoped column of live project metrics.
Organisation-wide parameters owned by the Workforce Manager — maximum hours per day, submission deadline day and default break minutes — enforced server-side.
Live product What is mine, and what is the project's — on one screen.
One table component, sixteen data views. Every list in the platform renders through the same shared component — consistent sorting, filtering, pagination, empty states and accessibility everywhere. The interface behaves the same way in every module, so training transfers and mistakes do not.
Authority in EnginAI is deliberately split, and every rule is enforced where it cannot be bypassed — at the API, not in the interface.
Chosen per project. Air-gapped: all inference on local models inside your network — no external API calls, no data egress. Hybrid: selected workloads may optionally be routed to external models where you permit it, under your own policy.
Every artifact — document, drawing, tag, knowledge chunk, chat, audit entry — carries a project ID. All queries bind a server-validated project. Context cannot cross projects.
Permissions are enforced at the API, not merely hidden in the interface. Unauthorised writes are refused even when called directly.
Three organisation roles — Administrator, Workforce Manager, Staff — plus four project roles: Lead, Approver, Contributor, Viewer. Organisation authority and project authority are deliberately separate.
Document transitions, safety-critical findings, report locking and timesheet approval each require a specific role — and each is enforced on the server.
TOTP-based 2FA, with both enrolment and challenge built in.
A hash-chained audit log, with a Recent Activity view for administrators.
A built-in OIDC identity provider; the conversational engine is a relying party.
Live health of every platform service in a single administrator view.
| Capability | Admin | Workforce Mgr | Lead | Approver | Contributor | Viewer |
|---|---|---|---|---|---|---|
| Global settings, integrations, audit log | ✓ | · | · | · | · | · |
| Create / edit users, assign org roles | ✓ | · | · | · | · | · |
| Create / edit projects, clients | ✓ | · | · | · | · | · |
| Approve monthly timesheets | ✓ | ✓ | · | · | · | · |
| Workforce parameters | ✓ | ✓ | · | · | · | · |
| Manage blocks, work packages | ✓ | · | ✓ | · | · | · |
| Manage teams and project membership | ✓ | · | ✓ | · | · | · |
| Approve documents / resolve findings | ✓ | · | ✓ | ✓ | · | · |
| Set work-package status | ✓ | · | ✓ | ✓ | · | · |
| Upload documents, run extraction | ✓ | · | ✓ | ✓ | ✓ | · |
| Log time, submit own timesheet | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Raise tickets, generate checklists / reports | ✓ | · | ✓ | ✓ | ✓ | · |
| Read project data | ✓ | · | ✓ | ✓ | ✓ | ✓ |
Separation is the point: a Workforce Manager has no engineering authority, and a project Lead has no approval authority over timesheets. The month lock follows the same principle — once a timesheet month is submitted or approved it is immutable for everyone, including an Administrator. The only way back is a Workforce Manager rejection, which reopens the period for correction.
Live product Stage transitions move one step at a time, and every step beyond submitting a draft requires an Approver.
Deployed as containers on your own hardware. Offline fallback models included.
Automate time-consuming tasks to deliver projects faster.
Reduce errors with AI-powered validation and verification.
Enable engineers to focus on innovation, not information.
Optimise performance, reduce costs and maximise asset reliability.
Deployed on your infrastructure. Evaluated on your data.
Screenshots are of the live product. AI-assisted outputs — drawing tag detection, compliance findings, generated reports and checklist suggestions — require engineer review.