Email-driven, AI-routed engineering document control

How manual filing, spreadsheet registers and hand-built transmittals were replaced with an AI-routed system built entirely on the existing Microsoft 365 environment.

Document control portfolio overview: 519 documents across 22 projects, overdue ageing, resubmission turnaround and live review-outcome charts

Document register showing the current revision of every document, with routing source, reviewer, OK-to-send state and the full filter set

Submission queue batching documents per project on a one hour window, with auto-send times and no manual transmittal building

Pivot view grouping the register by project, customer, discipline or reviewer, with submitted, awaiting and overdue counts per group

The challenge

Engineers filed documents by hand, needing to know exactly where each one belonged.
The register lived in spreadsheets, no concurrency, no audit trail, breaking down at scale.
Customer-returned markups were read, compared and actioned by hand.
Transmittals and submission packages were assembled manually for every issue.
Commercial deviations buried in technical comments were easy to miss until they became disputes.

What we built

Built inside the existing tenancy, SharePoint for files, Dataverse for data, Power Automate as the engine, an LLM for the intelligence.
A returns mailbox: the workflow reads each markup (even handwritten), pulls the issued revision and the PO, and diffs them.
Every comment is checked against the purchase order; anything out-of-scope is flagged immediately.
Confidence scoring routes each comment, high-confidence auto-filed, mid to a reviewer, low held for a human.
A single drop-zone: engineers drop documents in one place and the workflow files them to the right project.
An append-only event log, every submission, return and revision is a new row, never edited.

The results

Zero manual filing, engineers interact with one drop-zone.
A single, real-time register across all projects.
Commercial deviations caught the moment they appear in a technical comment.
A complete audit trail by user, time and change, suitable for regulated and contractual work.

The economics

Runs inside the Microsoft 365 the business already pays for, no new platform. Files and data stay in the tenancy; only the pages being read go to the AI model.
The intelligence layer is a low-cost, pay-per-use AI API.
Specified and iterated by a domain expert, not a multi-month systems-integration project.
Reached a working state at a fraction of the cost of a conventional document-management rollout.

What transfers to your business. Any project-based business that issues and receives controlled documents can separate the files from the data, make the register an event log, and let AI read, compare and route, on tools most organisations already own.

Other case studies

Product configuratorA 3D switchboard configurator in a single browser file Estimating & quotationAutomating the estimating-to-quotation workflow Sales intelligenceSales intelligence without a BI project Enquiry handling & RFQ triageAutomating enquiry triage and RFQ logging

Could this work on your process?

For one tool, a free 90-minute process walk and a fixed price within 48 hours. For bigger work, a two to three week assessment, and you keep the roadmap either way.

Book a Call See Starter Builds