Stop rewriting the answer yourself.

Most people blame the model. Usually the prompt left out the one thing the answer depended on. This course teaches the small number of moves that fix that — practised on the work already sitting in your inbox.

See what's inside

3
Modules: 9
Practical lessons: Free
No card required: 1-to-1
A coach reviews your work

You typed something in, and the answer was almost useful

You asked for a summary and got a summary of the wrong things. You asked for a first draft and got something so generic you rewrote it from scratch. The tool worked. The output did not, and doing it by hand was quicker — so you did, and the tab stayed open, doing nothing.

The advice on offer does not help much either. Prompt libraries hand you a hundred templates for tasks you do not have. "Act as a world-class expert" does not survive contact with a real quarterly report. And nobody mentions the boring reason your prompt failed: the model never had the information it needed, and answered anyway.

That part is fixable. It is a skill rather than a trick, it takes a few hours, and it keeps working whichever assistant your company ends up standardising on.

What you will be able to do

Get a usable answer early instead of rewriting the same prompt until it gives up something you can use.
Give a model the context it is missing — the audience, the constraints, and an example of what "good" looks like that it cannot guess.
Ask for a specific format and get it back every time, so the output drops straight into the document you are filling in.
Recognise a confident, well-written answer that is wrong, and check the parts you cannot verify yourself.
Turn a prompt that worked once into a template a colleague can run without you.

What you'll cover

1
Structure: the four parts of a prompt
Nearly every disappointing answer came from a prompt missing one of four parts. You learn what they are, which one you are almost certainly leaving out, and a template you can paste today.
2
Iteration: from mediocre to usable
The first answer is a draft and you should read it like one. Four trainings on the moves that turn a disappointing first answer into something you would put your name on — usually by the third try, not the thirtieth.
3
Capstone: build your prompt library
A prompt you wrote once and cannot find again saved you nothing. The capstone turns the last two modules into five reusable templates for the tasks that actually eat your week — and one you have already used for real.

This is for you if

You already use an assistant most weeks and suspect you are getting a fraction of what it can do.
Your work is words and numbers, not code — you are an analyst, a marketer, an operations lead, or a founder doing six jobs.
You have collected prompt templates and never reused the same one twice.
The output goes to somebody else, so "close enough" is not good enough.
You would rather learn three moves you keep than thirty you forget by Friday.

Why this course is different

Three moves, not thirty templates

The whole course rests on a short list of things that transfer: supply the context, state the constraint, show an example of good. You will use them on every tool and every task, long after any particular template has gone stale.

You practise on your own work

Every exercise asks for a real prompt about something from your actual week — the report, the messy thread, the spreadsheet nobody wants. Nothing is practised on a tidy sample that behaves better than your work does.

A coach reads what you submit

Written exercises are read by a person, not scored by a script. You get told what your prompt left out, which is the one piece of feedback a model cannot give you: it will answer a bad prompt cheerfully and never mention the problem.

Questions

How much time does it take?

Around three to four hours of trainings and exercises in total, and you can spread that over as many weeks as suits you. Nothing expires, so a week where you do none of it costs you nothing.

Do I need a technical background?

No. No coding, no maths, no machine learning. If you can write a clear brief for a colleague who was not in the meeting, you already have the underlying skill — most of this course is about making that brief explicit.

Which AI tool do I need to have?

Any general assistant you can already get to: ChatGPT, Claude, Gemini or Copilot all work. The free tiers are enough for every exercise. You do not need to buy a subscription to take the course.

Should I start here, or with the free foundations course?

If you have barely used an assistant, start with the free foundations course — five short trainings covering what the model is doing and what you must never paste into an unapproved tool. If you already use AI most weeks and the answers are just mediocre, start here. This course is $149, one payment, and you keep access to it and its updates.

Do I get feedback on what I submit?

On the written exercises, yes. You submit the prompt you wrote and the answer you got back, and a coach reads both and replies. Quizzes are graded the moment you answer them, so you do not wait on those.

Is this about building AI products?

No, and we would rather say so before you pay. There is no fine-tuning, no retrieval pipeline, no API work. This is about using an assistant well on ordinary office work. If you want to ship an AI feature, this is the wrong course.

Bring one task you would like back

Pick the thing you rewrite by hand every week. By the end of the course you will have a prompt for it that you actually keep — and a clear-eyed list of the tasks where the honest answer is still to do it yourself. $149, one payment.