Capability

Applied AI & Automation

Document and filing workflows, agents, retrieval, and human-in-the-loop review.

Discipline: 02
Typical duration: 5 to 10 weeks, after the evaluation
Model: Fixed scope
Ownership: Yours

What it changes

In a professional firm the expensive hours go into forms: the return, the claim, the report, the same fields copied out of records you already hold. We build the system that assembles the document from those records and puts it in front of the person who has to sign it. Capacity goes up without headcount going up, and nobody signs something they did not read.

In practice

The system drafts. The specialist signs.

We build AI into workflows that already exist, grounded in the organisation's own records so the output can be checked. The pattern that works is not autonomy, it is division of labour: the system does the drafting, the reading and the sorting, and a qualified person owns the decision and signs it. It is the same engineering whether the document is a clinical report, a statutory return or a funding claim. What changes is the records it reads and whose name goes on the bottom.

A drafted return, waiting on the accountant's signature

What that includes

Everything in the box, itemised.

Interface design, deployment and handover are in every engagement, so they are never quoted as extras.

Document and report generation pipelines
Statutory returns, claims and forms, assembled for review
Retrieval-grounded assistants and agents
Human-in-the-loop review and approval workflows
Classification and computer vision, including on-device

The division of labour

The system drafts. The specialist signs.

Automation is worth having where it removes the blank page, the re-keying and the search. It is not worth having where it takes the decision, because the decision is the part somebody has to answer for. So we build the two apart: the machine does the labour, a qualified person owns the judgement.

It is the same system whether the document is a clinical report, a statutory return or a funding claim. What changes is the records it reads and the professional whose name goes on the bottom.

What it drafts, and who signs it

Signed by

An EEG report, structured the way the department already writes them
The neurologist reviews, corrects and signs
A statutory return or filing, assembled from the client's own ledger and supporting documents
The accountant checks the figures and files it
A funding or grant claim, filled from the records the institution already holds
The certifying officer verifies and submits
A support answer, drawn from resolved tickets and citing the source it used
The support lead can audit every answer sent

Two of these are running today and link to the story behind them. The rest are the same engineering: read the records the organisation already holds, assemble the document, and put it in front of the person whose signature it needs.

A named human owner

The person who reviews and signs is recorded, by name, against the output.

Grounded in your records

Answers come from your own documents and ledgers, so every one is checkable.

It shows its working

The citation, the query, the draft before the edit. All of it kept.

How it runs

A typical engagement.

A two-week evaluation on your own documents, tickets or records before anything is committed, so the accuracy question gets answered with your data instead of a demo. If it clears the bar, a fixed-scope build follows with the review and sign-off step designed first, not bolted on.

The operating rhythm behind it