This course is for the people who decide whether to build an AI feature: startup founders, product managers, executives and support-function leaders in large companies. It needs no technical skills. It needs a real task from your own work, to apply each lesson to.
A one-page scoping sheet fills itself, one line per lesson: the problem, the tool, the journey, the criteria, the budget, the risks and the pilot. The choice of tool stops on the decision model, and the errors are re-sorted by harm. The budget line stays empty, and the stamp lands on "not yet": an empty line is already an answer.
The path
Seven lessons, in the order a good decision gets made:
- Start from a task done by hand, not from the wish to "add AI".
- Choose the right tool: build nothing, a written rule, a decision model or an LLM.
- Place AI in the user journey: what goes in, what comes out, and who checks it.
- Define "good enough" before building, with success criteria and test cases.
- Estimate the cost and response time of each use.
- Plan for wrong answers, with guardrails sized to the harm they could cause.
- Go from demo to pilot, with a kill criterion and a review date.
Each lesson builds on the FSH Learn concept pages, which are free to read, such as the LLM, hallucinations or evals.
How long it takes
Each lesson takes six to seven minutes to read, and six of them end with an exercise of about ten minutes with the Coach. Each lesson also offers an experiment in two parts: one to do straight away, with what you already know, then one to do during the week, in the field. You can take the course in one sitting, relying on estimates, or over several weeks, coming back with your measurements. In its feedback, the Coach gives more weight to a measured figure than to an estimate.
What you leave with
By the end of the course you have a one-page scoping sheet for a feature of your choice: the problem and its cost, the chosen tool and why, the success criteria, the budget per use, the risks and the boundaries of the pilot. It is the document that lets you answer yes, no or not yet.
How an exercise works
Six of the seven lessons end with an exercise with the Coach, the FSH Learn AI built on my method. You describe your real situation. The Coach asks a few questions and has you apply the lesson to your case, without asking you to recite it. After a few exchanges it gives you a feedback card: what you did well, what to work on, a next step, and a level for each skill, from "Not yet" to "Strong".
What an account gives you
Concept pages are public. Lesson 1 and its Coach exercise need a free account, which also gives you a monthly quota of Coach exercises, progress tracking and a radar of your skills. When you finish the course, you can create a completion card to share.
Around this topic
Go deeperRolling AI out in a team
More courses are in preparation, including one on rolling AI out to teams, treated as change management: training, tooling, measuring use and tracking costs. This course lays the groundwork: a team adopts a tool that takes a familiar task off its hands more readily than one imposed from above.
Go deeperArchitecting with LLMs
Another, for technical leads, will cover the context window, document retrieval, agents, guardrails and evals in a real architecture. It picks up where this one stops, once the pilot is decided and you still have to build a system that holds up in production.
