Every App We Ship Now Talks to a Model

Last year, AI mostly changed how our team built software internally. This year, it changed the conversations we have with clients before a single line of code is written. It’s rare now for a new web or app project to scope out without some version of “and it should probably be able to answer questions about our data” showing up in the first call.

From “add a chatbot” to “make this actually useful”

Two years ago, an AI feature request usually meant a generic chat widget bolted onto a homepage that answered questions worse than a search bar. The requests we get now are more specific and more grounded: summarise this year’s support tickets, draft a first-pass reply to this type of enquiry, pull the three relevant clauses out of a hundred-page contract. Narrower asks, and much easier to get genuinely right.

The unglamorous work made this possible, not the model

The actual hard part of these projects has rarely been the AI model itself. It’s been getting a client’s data clean and organised enough that a model can be pointed at it reliably. Document formats, inconsistent naming, duplicate records - that unglamorous data work is 80% of the project, and it’s exactly where a lot of AI initiatives quietly fail before the model ever gets involved.

What we tell clients now

We’ve started explicitly scoping projects into two phases: get the data and systems into a shape where automation is safe, then add the AI layer on top. Skipping straight to phase two is how you end up with a demo that looks great and a production system that confidently gives wrong answers.

Looking at where this is going

The pattern over the last couple of years has been consistent: AI capability keeps arriving faster than most businesses’ data and processes are ready for it. The businesses getting real value aren’t the ones chasing the newest model - they’re the ones who did the boring groundwork first.

If your team is trying to figure out which phase you’re actually in, that’s a conversation worth having before committing budget - reach out.