Mirror
Self-serve, and worth running again as the organization changes.
How we work
Read the structure. Redesign it. Coach the people who run it.
Nothing here is a maturity model. Each dimension is a state we can describe and change.
Where the work is
Most of what AI could do is not going to happen.
40%
Expected in use
Possible, and actually in use.
60%
Possible, and not happening
Nothing technical is stopping this. It is the part we work on.
What you get
Mirror. Method. Mandate.
Always in this order, and not three stages of maturity.
Method
What to sequence, what to redesign, what to stop.
Mandate
The same finding as a fundable case. The part a champion carries upstairs.
Mirror · 1
Seven dimensions. Where the organization actually is.
Not where leadership believes it is.
how things work here
Dominant Culture
Where the organization is, not where leadership thinks it is.
who really decides
Decision Rights
Who allocates the resources and owns the outcome.
how people think
Cognitive Profile
System thinkers, executors and experimenters, and in what proportion.
where freed time goes
Capacity Flow
Where freed capacity goes, and where it quietly does not.
what gets rewarded
Incentive Alignment
Whether the rewards support the target model or contradict it.
people readiness
Cultural Readiness
Which teams are ready for which model.
do leaders walk it
Leadership Behavior
Whether what leaders do matches what they are asking for.
The mark on each card is where that dimension sits on the diagram below, clockwise from the top.
What the mirror is read against
Three kinds of AI. A choice, not a ladder.
Not three levels of sophistication. The order is how much of the deciding the AI takes over: the more a project decides on its own, the more has to already be true around it.
AI assistance
Takes volume off people.
The work happens the way it always did, there is just less of it for a person to do. Nobody’s authority changes.
1 condition has to hold
- Somebody has decided where the freed-up hours go. If nobody has, the saving never actually appears.
Fits regulated industries and operational excellence.
AI agents
Supports expert judgment.
A person still decides, now with a machine’s suggestion in front of them. Some authority is shared.
2 conditions have to hold
- One named person is allowed to overrule it.
- The careful call that turns out wrong is not punished.
Fits professional services and expertise-driven cultures.
Autonomous AI
Runs whole processes.
The machine decides and acts, and a person looks at the exceptions. Most of the authority has moved.
4 conditions have to hold
- Decisions sit where the work is.
- There is real slack in the calendar.
- The targets do not fight it.
- Leadership is visibly behind it.
Fits technology and high-agility cultures.
Choose a kind of AI
Choose a dimension to see what it reads, and whether it is make-or-break for AI assistance.
AI assistance: 1 condition has to hold
Method · 2
Four calls, and the cheapest one is stop.
Every project in front of you is one of these four. Saying which is the work.
Sequence
What goes first, and what waits.
Redesign
What cannot work in its current shape.
Stop
What will not clear, said early enough to be cheap.
Incentives
Which moves first, because the rest follow.
Mostly through three of the seven above: Decision rights, Incentives, Capacity flow.
Mandate · 3
A finding nobody has funded changes nothing
It starts by separating what you answer for from what you can actually call.
Accountable for
- Adoption across the business
- What the portfolio delivers
- Reporting progress upward
Can decide
- What your own team tries
- Where your own time goes
- Which questions get asked
Accountable for, cannot decide
- Where the time it saves goes
- Who can stop it, and how fast
- Whether the yes from above is real
The ask
One named person, one decision, one date.
Alongside all three
Someone still has to run it on Monday
We redesign the system. We coach the leaders inside it. Two different verbs on purpose: a system changed around a leader who was never supported to run it reverts, usually within a quarter.
Not a fourth part. It runs through all three.
Self
What they are actually optimizing for, including the parts nobody says out loud.
Alignment
Whether their stated priorities and their calendar agree.
Frames
The story they use to explain the change, and whether it survives contact.
Expectations
What they asked for, what people heard, and the gap.
How it runs
When we are in the room, and when we leave
Diagnose
Name what is structural and what is only difficult.
Design
Redesign them with the people who will have to operate the result.
Build
Run the first cycle under the new rules, with us in the room.
Sustain
Coach the leaders until it holds without us, and stop.
Two short reads
Arithmetic on your own answers, in about five minutes
Neither scores your organization and neither issues a verdict. Both do sums on what you tell them. Nothing you type leaves your browser.
Rate seven behaviors twice: what leadership says matters, and what actually happens. The finding is the distance.
Write down what actually happens, then say what it costs and whether it can be named.
Next
Pick one thing, and let’s get it moving
Tell us about one decision that keeps getting stuck, and who would need to agree to it. That is enough to start.