There is a practice you will not find in a textbook: cursigraphy — the art of building a curriculum around a tool you do not yet understand. That is exactly where many organisations begin their GenAI journey: adopt a tool first, then manufacture all the ways people should use it. The result is guaranteed, but it is often the wrong result. We see far too many deployments whose objective is 'employees are using the assistant', and far too few whose objective is 'our proposal lifecycle is 25 percent faster'. One declares activity; the other delivers a result.
Working backwards turns that sequence upside down. Start with the outcome the business actually wants — faster time-to-proposal, higher accuracy in onboarding, lower cost of first-line triage. Then identify the workflows that feed it and, only then, the handful of capability gaps the tool genuinely closes. When the lesson is designed around the outcome, the tool appears at the point the workflow demands it, and learners acquire the skill in context instead of as a technology appendix detached from their real day.
This discipline is unfamiliar, and that is why it stalls. It is comfortable to buy a platform, turn on licences and run a lunch-and-learn. But the difference between adoption and outcome is the one executives actually fund: adoption is a dashboard trend, while outcome shows up in the operating result. In practice we have helped clients reduce their AI curriculum from nineteen generic modules to four, each explicitly linked to a countable result — and adoption did not fall, it rose, because employees could see precisely why the session existed.
So before authoring another prompt workshop, ask the business leader the question that should drive every GenAI decision: what is the outcome, and what would it take to achieve it? The curriculum writes itself from there. Tool-led learning is a technology fantasy; outcome-led learning is an operating discipline. The organisation that inverts the ordering — outcome first, tool second — is the one with a working, evidence-based transformation case at the end of the year.