AI training on your own processes and data.

Enough that everybody can still share a screen and build something. Past roughly a dozen it turns into a webinar, and a webinar changes nobody's Tuesday. For larger teams I split the day by role.

On the ones you have or intend to have. If the choice has not been made, part of the boundaries block goes on picking a tool against your data and your legal requirements. A session on somebody's demo account teaches clicking.

There is plenty of free material and it is worth using: provider courses, public programmes, vendor tutorials. They teach the tool in general, on examples your company does not have. This day sits on the other side of that: your material, your process, so the team leaves with work done rather than knowledge to translate later.

Using it privately usually means three things at once: personal accounts, company data pasted without much thought, and text that reads as generated. The day sorts all three, and adds the applications people do not arrive at on their own because they need the model connected to company data.

In what I teach, the model writes a first version and a person checks and sends it. The mix of work changes: less writing from scratch, more verifying. Anything reaching a customer without a person is a separate decision, taken much later.

I agree the scope of material before the day and keep it to what is needed. In regulated industries I start by checking whether the provider signs a data processing agreement; without one, that process does not go on the screen, however simple the technical answer looks.

Sometimes. Work with a model on a document stays with the team after a day. An application that has to cross a whole database, or run without a person, has to be built, and then the training comes after it.