Most organizations adopting Databricks hit the same wall: the platform is ready well before the people who have to use it. So business users get trained the way you'd train data engineers, menu by menu. The result is predictable: intimidated teams and an underused platform.
My conviction is simple: a business user doesn't need to know everything about the platform. They need to succeed at their first real task.
You don't learn Databricks by touring every screen. You learn it by solving a problem the team already knows. So I always start by picking a real deliverable: a report someone still produces by hand, a recurring question from leadership, a month-end reconciliation. That's the thread running through the whole session.
"Trust is built on the ability to doubt, not on magic."
With Genie, you can ask a question in plain English and get an answer in seconds. It's a huge accelerator for non-technical profiles, as long as you also teach them to recognize a doubtful answer and to check a definition. Trust is built on the ability to doubt, not on magic.
- Start with a deliverable the team already knows, never an abstract example.
- Show, have them do it, then let them do it, all within the same session.
- Leave behind templates and sample queries, so the autonomy survives my departure.
A successful training session isn't measured at the end of the session, but a week later: did someone redo the task on their own? That's the only indicator that counts.
