A software company, not an AI consultancy

Most of the work that keeps a company running is unglamorous and repetitive: categorising transactions, chasing an invoice, working out why a page slipped off the first results page, filing the thing that is due next week. It is also the work that gets postponed, because the owner is doing it at the end of a day spent on the actual business.

That work is now largely automatable. Not because a model can be trusted to run a business, but because the reading, matching and drafting that fills those hours is exactly what models are good at — provided the parts that must be exact stay exact.

So we draw a hard line through every product. The ledger, the totals, the filings and the audit trail are deterministic code that either balances or does not. The judgement calls around them — is this transaction what it looks like, is this draft right, what changed and why — are where the AI works, with a person in the loop wherever being wrong would be expensive.

We build several products at once because they share more than they differ. Authentication, billing, the data model, the audit trail and the AI layer are built once. Each new product starts from that foundation rather than from nothing, which is why a company this size can ship across bookkeeping, operations, visibility and compliance at the same time.

We would rather say a product is in development than call it live, and rather tell you it does not fit than sell you a pilot. Every stage label on this site means what it says.