Key takeaways
- The curriculum splits by role family; narrower scope means greater depth.
- The executive programme is decision-oriented, not tool-oriented.
- Safe-usage training aligns one to one with the governance framework.
- Training is a third of the programme; the rest is practice and measurement.
Define the role families
The first step is deciding who needs to learn what. A typical organisation yields four to six role families: executives, sales and customer communication, technical and field, finance and operations, content and marketing, HR.
Three to five critical capabilities are selected for each family. Choosing more than five makes the programme superficial and leaves participants feeling they have seen something they cannot do.
The executive programme: decision, not tools
Executives are not expected to train models. What is needed is enough understanding of what AI can and cannot do to allocate decision rights: which decision sits with the machine, which with people, and where the machine produces and a person approves.
This programme is also a risk conversation. Which data may enter which tool, who is accountable when something is wrong, how is customer-facing output reviewed?
The specialist programme: their own work
Specialist training is applied, and participants work on assignments brought from their own real work. Transfer rates are low when training uses generic examples; a specialist who works on their own proposal document uses the same method again the next day.
Prompt and workflow training happens here. But what is really taught is verification: how to check the output, which errors are typical, and when to stop.
Measurement and certification
Measurement uses usage and output rather than satisfaction surveys: weekly active use, hours saved, consistency of output quality and the fall in off-policy tool use.
Certification is an optional layer. Badges and levels raise participation but are not on their own an indicator of capability.