Someone spends half their week reading documents and retyping what is in them.

AI is very good at a narrow set of jobs: reading documents, writing a first draft, sorting things into piles. It is unreliable at anything that has to be exactly right with nobody checking. We build the first kind, we say no to the second, and we tell you which one you are asking for.

Tell us what's broken

IN PRACTICE

What this looks like in a real business.

A wholesaler buried in supplier invoices

Hundreds of PDFs arrive each month, every supplier laying theirs out differently, and someone keys them in one at a time. The work here reads each invoice, pulls out the supplier, dates, line items and totals, matches them against the purchase order, and puts anything it is unsure about in a queue for a person. The job changes from typing everything to checking the exceptions.

A not-for-profit writing the same letter forty times

Acquittal reports and member letters say much the same thing with different names, dates and numbers in them. The work here drafts each one from records the organisation already keeps, then hands it to a staff member to read, edit and send. Nothing leaves the building unread by a human.

A clinic with an inbox nobody owns

Referrals, results, cancellations and sales pitches all land in the same address, and the urgent ones get found late. The work here sorts arrivals into piles, flags what needs attention today, and shows why it made each call so staff can correct it when it gets one wrong.

WHAT YOU GET

Every project includes this.

  • A blunt assessment first: which parts of the job AI will handle well, which parts it will not, and whether the maths works out in your favour.
  • The feature built into software your team already opens, not another tab to remember.
  • A person in the loop wherever the answer matters, with anything uncertain queued for review rather than guessed at.
  • The running costs explained before you commit, including what happens to them if your volume doubles.
  • A written note of where your data goes, which provider handles it, and how long they keep it.
  • The code and the accounts, in your name.

HOW WE WORK

No surprises, at any point.

  1. 1

    Conversation

    Free, and there is no pitch in it. You tell us what is painful. We tell you if we are the wrong people for it.

  2. 2

    Written scope

    Fixed price, fixed deliverable, in writing before anyone commits to anything.

  3. 3

    Build in stages

    You see working software every couple of weeks, not at the end. You can stop after any stage and keep what is built.

  4. 4

    Handover

    Code, accounts and documentation are yours. Nothing is held hostage.

STRAIGHT ANSWERS

The questions everyone asks.

Where does AI not pay off?

Anywhere the answer has to be exactly right and nobody is going to check it. Working out a payroll figure, deciding anything clinical or legal, or replacing a person whose real job is judgement. It also does not pay off on a task you only do twice a month — the effort of building it never comes back. We would rather tell you that in the first conversation than take the work.

Where does our data go?

That depends on which provider does the reading, and we tell you which one before anything is built. We can usually keep the data inside a region you are comfortable with, and we turn off training on your content where the provider allows it. If your information is sensitive enough that none of that is good enough, we will tell you the honest answer is not to do it.

What happens when it gets something wrong?

It will, so the design assumes it. Anything the system is not confident about goes to a person instead of through. Every decision it makes is recorded with the reasoning, so you can look back at a bad one and see what happened. And you can always switch a piece of it off and go back to doing that part by hand.

START HERE

Tell us what's broken.

Two or three sentences is plenty. You will hear back from a person, and the first conversation costs nothing.