
Databricks shows how to teach Genie One your team's workflows with skills
Databricks published a guide on November 16 to skills in Genie One, its AI assistant for business users. A skill is a written recipe for a task your team does again and again, such as a month-end pack or a weekly forecast memo. It holds the steps, the checks and the format of the result.
You build a skill in one conversation. You explain the workflow in your own words, and Genie One asks you questions before it writes the skill. Databricks suggests that you first answer a few questions yourself: when the skill applies, which data and tools it uses, the order of the steps, what is checked every time, what the result looks like, and where you want to review it.
Genie One uses a skill when you call it, or when it decides the skill fits your request. Each skill is saved in your workspace in the open Agent Skills format, under Customizations and then Skills. You can share it across your organization, and other agents that support MCP can use it too.
Databricks says a skill is a good idea when you explain the same thing to Genie One more than once.
What it means for your work
- For finance and operations teams: Start with one regular report and write down its steps and checks first.
- For team leads: Test a new skill against a report your team already made, and fix it in the conversation.
- For data teams: Keep business definitions in Genie Ontology and the method in the skill.
For more on Genie, see Databricks Genie in Slack.