How it works
Seven stages, about an hour, in as many sittings as you like.
It is a structured interview, not a chat window. Each stage is a form the co-pilot filled in, which you correct — because a transcript implies the output is a conversation, when what you actually want at the end is a document you will follow.
The seven stages
0 · Getting started
What this is, who sees it, and what happens to it if you leave your employer. Consent in plain sentences, not a wall.
1 · What matters to you
Not a personality test. What makes work feel worth doing — and, critically, how many hours a week you actually have. Two honest hours beat six aspirational ones.
2 · Where you are going
Far enough out to matter, close enough to act on this quarter. Vagueness gets pushed on rather than accepted: "grow into leadership" is not something anyone can do on Tuesday.
3 · Where you are now
Your assessment results, with every number saying whether it was measured or self-declared — and where you think they are wrong.
4 · What to work on
Three competencies. Not five. Over-scoping is the most common reason plans get abandoned, so the limit is enforced rather than suggested.
5 · What will be different
Outcomes, never activities. Write "complete the negotiation course" and the co-pilot will explain the difference and offer a version drawn from your actual job — which you can take, edit, or refuse.
6 · How you get there
Roughly 70% practice at work, 20% people to talk to, 10% study. If it does not fit the hours you gave, it gets cut until it does.
7 · Making it real
A check-in rhythm and one thing to do this week.
The part we are most careful about
Why we sometimes say “held steady”.
Every instrument has measurement error. Ask the same person the same questions twice and the number moves a little even when nothing about them has. So before we call anything growth, the change has to be bigger than the noise — and how big that is depends on how long you waited.
90 days
8 points
Short retest spacings have the largest practice effects — people partly get better at the test, not the thing.
6 months
6 points
Long enough that most of the practice effect has washed out.
12 months
5 points
The most reliable interval, and the one we suggest for anything you intend to show a board.
Movement smaller than that is reported as “held steady”, with the reason stated. It is not greyed out and not apologised for — it is a different and equally legitimate result.
What stops the AI making things up
It cannot invent a course
Every recommendation must reference a course that exists in the catalogue. This is enforced in the database, not requested in a prompt — a prompt is a preference, a constraint is a guarantee.
It cannot exceed your hours
The plan is cut to fit the time you said you have, dropping the course first and practice last. A plan that consumes every available hour has no room for the week when everything goes wrong, and every quarter has one.
It stops when it should
If someone writes something that suggests real distress, the interview stops and offers support rather than carrying on collecting career goals.
These are checked in code after the model responds, not merely asked for in the prompt. Each has an automated test that runs on every change.
See it rather than read about it
The demo runs the whole interview with invented data. Try writing “complete the leadership course” as an objective.
Open the demo