Skip to content

Technology

Putting order into how a development team uses AI

Rolling out AI-assisted development workflows across a whole team: tool selection, project-level instruction files, review conventions and training.

Client
Digital product team
Year
2024 — present
Abstract illustration of scattered inputs funnelled into a single ordered lane
Whole team
on shared criteria, not just the curious ones
Versioned
AI instructions live in the repository
Review
explicit conventions on what gets accepted

The situation

The team was already using AI tools, but everyone in their own way: different tools, different criteria about what to accept and no shared conventions. The results were uneven — very good in some cases, counterproductive in others — and hard to review.

What was done

  • Tool selection based on where the team’s bottleneck actually was.
  • Project-level instruction files versioned in the repository, so project context and conventions travel with the code instead of living in each person’s head.
  • Review conventions: what a human must always verify, what can be accepted on a quick read, and what should never be generated automatically.
  • Hands-on training on the team’s own repository, with written agreements at the end of each block.

What changed

The measurable effect was not "writing code faster", which is the easy promise, but shorter delivery cycles from removing time lost to project set-up, repetitive work and arguments about conventions that were already written down.

Next step

Half an hour well spent

Walk us through the problem on a short call. You leave with a first read on how we'd approach it and what it would involve — no commitment, no sales deck.