Where AI Should Sit Before You Bring It Into a Dev Team
Many dev teams bought AI coding tools, only some people use them, and the speedup comes with doubts about quality. What needs deciding first is where AI sits in the work.
There is a story we hear often from dev teams that brought in AI coding tools. Everyone got an account, only some of them actually use it, and those who do got faster while nobody is quite sure the quality kept up.
It is rarely the tool's fault. Usually the tool arrived before anyone decided where AI fits in the development flow.
Decide the seat before the tool
Development starts with planning, moves through design, building, debugging and verification, and ends with a release. AI can help at every step, but it does not enter every step the same way.
That is why the development track of MVPIT's AI Advisory starts with repositories, deployment, and code review. We look at how code gets written today, who reviews it and how, and which path it takes to production. Only then do we decide where AI goes.
Three things to settle along with it
When you decide where AI sits, a few other things have to be decided with it.
First, rules. Code AI writes still has to follow the team's conventions. If those conventions live only in people's heads, AI produces something different every time. Write the rules down in files and AI reads and follows the same rules.
Second, review. Faster output means more to review. Reading every line by hand cannot keep up. Automated checks filter first, and people look where judgment is needed, so the job of review has to be split again.
Third, security settings. Before rollout, decide what code and data the AI tools can reach and what must never leave.
How MVPIT works
MVPIT works this way itself. Fixes go onto a request board, AI changes the code and submits it, and it ships after automated checks. Rules live in files, and hooks and automated tests block what must not break.
The standard fits in one sentence. Let AI do the work, and keep judgment and verification with people.
Measure it or lose it
Once the seat is set, define the metrics that show the effect. If you don't measure what got faster and what got better, a few months later the tools are still there and nobody can explain why they were brought in.
You can see how we spend three weeks deciding where AI sits in your dev team on the AI Advisory page.
Get notified of new posts
We'll email you when a new blog post is published.