What should your AI capability plan actually be?
A bespoke, self-hosted, self-tracked AI fluency plan built on the four competencies of delegation, description, discernment and diligence. One HTML file that lives on your machine. No account, no subscription, and the page makes no network requests at all.
Training attendance is not capability.
Most organisations measure AI capability by counting who attended the session. Attendance is an input. It tells you nothing about whether anyone can now delegate a real task to a model, specify it precisely enough to get a usable answer, judge whether the answer is wrong, and stand behind the result.
The four competencies come from the AI fluency framework: Delegation, deciding whether and when to use AI at all; Description, specifying the goal precisely; Discernment, assessing the quality of what comes back; and Diligence, owning the outcome. They are competencies, which means they are practised rather than attended, and they are measurable if anyone bothers to define the measure.
That is the gap this closes. A plan with sessions, hours, a gate date, and a measures table is a capability commitment. A calendar invitation is not.
Thirty-five sessions, six sprints, one gate date.
You set three things: a plan name, a testable objective in one sentence, and the date you intend to finish. The plan assembles around them, defaulting to roughly 35 sessions and 35 hours across six sprints, and it is yours to prune or extend.
Whether and when
Deciding which work belongs with AI, which belongs with a person, and which should not be attempted at all.
Specifying precisely
Turning an intent into a specification a model can act on, including the constraints and the definition of done.
Judging the output
Assessing quality, spotting confident error, and knowing which claims require verification before they leave your hands.
Owning the result
Accountability for what you sign, including provenance, disclosure, and the professional obligations that do not transfer to a tool.
A file, not a login.
The plan is a single HTML page. It tracks completion, hours done, sessions remaining, days to your gate date, and the weekly pace you need to hit it. It carries a measures table because feelings do not pass gates and measures do, and you replace the default targets with your own.
Progress is stored in your browser and can be exported as JSON so it survives a new machine. The page makes zero network requests: nothing is uploaded, nothing is telemetered, and there is no account to close later. That property is the point, not a feature. It is what makes the plan usable inside an organisation that cannot send its capability data to a third-party platform.
A personal plan, not a programme.
This builds a plan for one person. It is not a learning management system, it does not track a cohort, and it does not report to anyone. If you need capability measured across a department, that is a programme design question and it is a different piece of work.
A Pro tier adds an intake diagnostic and an evidence log for people who want the plan tailored rather than assembled. The free version is complete on its own and is not a trial.
Ten minutes, and you own the output.
Ten minutes to a plan with a gate date on it. Nothing is sent anywhere, including to us.
An indicative self-assessment, not an audit and not professional advice. Where a figure is modelled rather than measured, the instrument says so on screen and shows the assumption.