It will do the typing. The judgement is still yours.

Nobody here is going to tell you this transforms your business or halves your payroll. What it does, used properly, is take the repetitive typing out of a job so the person doing it can spend their day on the part that needed them.

We are not an AI consultancy. We are engineers who use it where it earns its place, and who can tell you where it does not.

A panda reading a draft and marking it up by hand
Shown, rather than claimed

Automate the volume. Keep a human on the decisions.

Three things we run. Same rule in every one, and in each case the interesting part is where it deliberately stops.

In the PIM

Writing the product copy a supplier left blank

Thousands of rows arrive with an empty description field. Writing them by hand is a job nobody has ever had the time for, which is why catalogues sit half-finished for years.

Where it stops

Nothing publishes until somebody has looked. A catalogue full of confident nonsense is worse than one with gaps in it.

In the connector

Moving prices and stock without anybody touching them

Thousands of updates a week between an inventory system and a storefront, which is exactly the sort of work a person should never be doing.

Where it stops

Anything moving more than 5% is held and flagged. One bad supplier file should not sell your stock at a loss all weekend.

At the sales desk

Drafting the reply to every inbound enquiry

Using your prices, your lead times and your tone, so a quote goes back in an hour rather than three days.

Where it stops

A person reads every reply before it sends. Your customers are not talking to a machine, and they should not have to wonder.

The question you are not asking out loud

"Does this mean I can let someone go?"

Probably not, and if that is the reason you are buying it you will be disappointed. What it removes is the part of somebody's week that was never really their job: the re-keying, the copying between screens, the sitting down on a Monday to type what the system already knew.

In every case we have run, the person stayed and the work got better, because they were finally doing the thing you hired them for. In a small business the constraint is almost never headcount. It is that your good people are buried.

If you do want to cut staff, say so at the start and we will tell you honestly whether this gets you there. Usually it does not, and finding that out after the project is an expensive way to learn it.

Good places to start
  • › High-volume text somebody drafts the same way every time
  • › Classifying or tidying messy data where a human checks the odd ones
  • › Filling gaps nobody was ever going to get round to by hand
  • › First drafts, where being 80% right saves an hour and costs nothing
Where we would say no
  • × Anything where being wrong costs real money and nobody is checking
  • × Technical specification and fitment. Confident and wrong is the worst outcome
  • × Anything customer-facing with nobody between it and the customer
  • × Putting your data somewhere you have not read the terms of
How we would start

Find the typing first. The technology is the easy part.

Most AI projects fail because somebody picked the tool before they picked the problem. We go the other way, and quite often the answer turns out to be that you did not need AI, you needed two systems to talk to each other.

01
Where is a person typing something a system already knows
Usually there are three or four of these and everyone has stopped noticing them.
02
What does being wrong cost, and who would notice
This decides where the human gate goes, and whether it is worth doing at all.
03
One narrow thing, running properly, before anything else
With the review step built in from the start rather than promised for later.
A fair warning

If you have been sold an AI strategy that is a slide deck and a list of tools, we are probably not what you are after. We build one working thing and then the next one.

Tell us where the typing is