From AI experiments to a team workflow

Turn AI experiments into a team workflow by moving the useful habit out of one person's chat and into the repository. A workflow is shared when a colleague can run the same task, see the same instructions, and review the result without asking the person who tried it first.

The homepage says this directly: we help teams move from individual experiments to a shared way of working with AI. Training is how we practise that on your codebase. The steps are Delegate, Review, Own.

Name the experiment that is worth keeping

Look at what people already do. One engineer may delegate a bug fix. Another may only autocomplete a line. Keep the experiment that produced a change a second person could review. Drop the one that only made sense inside a long chat.

Write down the task, the files it touched, the checks that ran, and the decision a person made. That note is the start of a receipt the whole team can reuse.

Bring the habit up to the team's level

The team proficiency matrix has five levels. Assisted means people still write the code and AI completes lines. Delegated means an agent does a whole task and a person reviews every change. Systematic means the team shares one way of working: the same instructions, context files and review checks. Anyone can pick up anyone else's work.

A single engineer can work at Delegated inside a team that is still Assisted. That gap is what the measurement shows. Training starts at the level three in four people have reached, then aims one level up. See AI coding maturity and the free team check.

Put the workflow where the next person will find it

Save the instructions next to the code. Say which tasks may be delegated, what evidence comes back, and who merges. Pilot it on one repository. When a second person can run it without a private explanation, you have a workflow.

Agentic coding, at the Delegated level, still ends with a person. The context window is only the text the agent can hold in one turn, so the durable rules belong in the repository, not in the chat that will be forgotten.

Services continues this after the workshop, with a briefing every two weeks and a monthly live session. Book a 30-minute call to pick the first workflow.

Frequently asked questions

What is the difference between an experiment and a team workflow?

An experiment lives in one chat. A team workflow lives in the repository, with the same instructions, context files and review checks.

Which level is a team workflow?

The Systematic level of the team proficiency matrix. Anyone can continue the work because the way of working is shared.

What is the first step?

Measure the team, pick one repository, and write the task boundary and the review receipt where the agent will read them.