We build AI features where they do something a database query cannot. We turn down more of these requests than we accept, and that is the useful part of this page.
When AI is the right tool
Genuinely unstructured language
- Summarising free-text notesSite reports, call notes, case records that nobody has time to read.
- Answering questions across your own documentsRetrieval over material that was never indexed.
- Classifying messages that arrive a hundred different waysWhere a rules engine would need endless maintenance.
- Drafting where a human reviews before it goes out
When it is not
The cost shape is different
An ordinary feature is built once and costs hosting. An AI feature carries a per-use cost forever: API usage that scales with success, and a vector database if you use retrieval.
Full published pricing and running costs.
Common questions
Should my app have AI in it?
Only if the input is genuinely unstructured language. If your problem has a schema, a query answers it faster, cheaper and more reliably. We turn down more AI requests than we accept.
What does it cost to run?
LLM API usage of roughly £50 to £300 a month depending on volume, plus hosting and, where retrieval is involved, a vector database. Unlike ordinary features, an AI feature has a per-use cost forever.
Can you add AI to an app we already have?
Often, yes, provided we can get at the code and the data. A retrieval layer over documents you already hold is one of the more contained things to add.