AI where it earns its place
Language models pointed at the specific jobs that eat your week — reading documents, sorting enquiries, answering from your own files.
What changes
- Invoices, receipts, and forms get read automatically instead of typed in by hand
- Inbound enquiries arrive sorted, with a draft reply already written
- Staff can ask a question of your own documents and get a cited answer
The interesting thing about AI for a small business is not the technology. It is that a handful of jobs which used to require a person reading carefully can now be done in seconds — and most of those jobs are the ones nobody wanted.
Concretely, the work usually looks like one of these:
Reading documents. Supplier invoices, delivery notes, timesheets, application forms. The information comes out and goes straight into your system, with the original attached so anyone can check it.
Sorting what comes in. Enquiries get categorised, routed to the right person, and arrive with a draft response already written. A person still sends it.
Answering from your own material. Staff ask a question in plain English and get an answer drawn from your policies, manuals, or past jobs, with a link to the document it came from.
Where we will tell you not to
AI is confidently wrong sometimes, and the cost of that varies enormously with the task. Sorting an email badly wastes a minute. Getting a dosage, a price, or a legal deadline wrong is a different category of problem.
So the rule we work to is straightforward: AI drafts, a person approves, wherever a mistake is expensive. If a job cannot tolerate an occasional wrong answer and cannot practically be checked, we will tell you it is the wrong tool and build it the ordinary way instead.
If this sounds like the problem you have, describe it in an email and we'll tell you what it would take.