AI Co-Working for Teams

You’ve seen your own results.
This is for your team.

Your result placed you on one map. Your team is on the same map but in in different places. When a team can see that shared picture, it can turn a myriad of private ways of working with AI into one everyone can agree on and, most importantly, benefit from.

The problem it solves

AI made producing work cheap. It made checking it expensive.

Most teams don’t have an adoption problem. They have coordination and standards problems. This puts the brakes on using AI effectively in ways that no amount of skills training can solve.

Everyone has settled on their own rules about when to trust AI, what “done” means, what’s worth checking and how carefully. The general idea is usually shared, but the behaviour isn’t. So the demand scales, and the burden of checking it lands unevenly.

Good intentions are shared. Consistent behaviour is not. This is where standards come in.

Four ways to work with AI

The read places each person on two axes: how far they keep the wheel (sovereignty), and how far they trust the tool. Four postures fall out of it. None is “bad” — each is a contribution a team needs. But only one is Partnership, and a team usually spreads across several.

Hypervigilance careful, slow to trust Partnership trusts, keeps the wheel Compliance using it, not owning it Overreliance fast, under-checking Trust in the AI → Sovereignty held →
The same map your own result used — now read across the whole team.
How it works

One 90-minute session. The whole team. From private results to one shared standard.

  1. Everyone reads their own result first

    Private, individual, before the room. Yours is yours; no one is asked to reveal it.

  2. We show the team map anonymously

    Where the team sits, as a shape, not a scoreboard. It is a system map, not a performance review.

  3. We name the shared cost and the real differences

    Usually not adoption — it’s the load of checking, and the quiet disagreement about when the AI has earned trust.

  4. The team turns its recommendations into a few things it builds once

    Dozens of individual moves collapse into a handful of shared assets — a standard, a context pack, a request pattern, a check.

  5. It leaves with named projects, owners, and a 30-day first cut

    Not “use AI more.” Concrete projects the team chose and owns, with a first review at day 15 and 30.

What a team leaves with

A few things built once, that serve everyone

Standard

A definition of done

One agreed sentence for “finished and checkable,” so people and AI aim at the same bar.

Documentation

A team context pack

The principles, standards and examples everyone loads — so no one re-explains the team each morning.

Prompt

A plan-first request pattern

Agree the approach before the work starts, so a wrong plan is caught early.

Check

A shared way to verify

An agreed check for work that matters — matched to what’s at stake, not applied to everything.

Checking matched to consequence — so speed survives where risk is low

GREEN

Low-consequence, reversible, internal. Normal owner review. Keep the fast path.

AMBER

Material analysis, customer claims, strategic calls. An independent challenge, with sources — a different kind of check, not just a second opinion.

RED

Legal commitments, regulated data, production changes. A named human approver and an evidence trail.

One rule holds it together: no new asset without naming what it replaces. Every build must remove work somewhere.
Why you can trust the read

It is built to start a conversation, not to score people.

  • Anonymous. No one is asked to identify their posture, and nothing is a verdict on anyone.
  • Every team number is computed from each person’s own answers — the team view can’t contradict your own page.
  • It reads behaviour, not just belief: a separate index checks whether the checking matches the confidence.

Curious what your team’s map looks like?

The read is a ten-minute survey your whole team can take. The session turns the result into one shared way of working — in 90 minutes.

Talk about a team session See a sample
Not ready yet? I write about the cognitive cost of AI — follow along on Substack or LinkedIn, and come back when the time is right.