Databricks showed teams how to turn repeat work into Genie One skills with a $50,000 test case.
Genie Ontology holds what terms mean while skills hold how work gets done.
Finance and sales operations staff must write the method once instead of fixing Genie One output each week.
Genie One calls a skill directly or picks it when a request matches its trigger.
Ontology now acts as a shared context layer for Genie One, Genie Code and outside agents built on other platforms per Databricks docs.
Metric views define numbers like net sales while Pages define terms like active customer.
A Genie Agent brings domain tables and joins while a skill brings steps, checks and format.
Builders map triggers, tools, order, checks, output and review points before asking Genie One to draft the skill.
Strict rules fit repeated figures, thresholds and sign offs while judgment fits picking the three changes leaders need.
Small skills chain better because Genie One reads only what it needs for the task.
Skills publish through Unity Gateway for any MCP agent including coding agents teams already use.
Testing uses real phrasing, a saved golden reference and a fresh chat to confirm the fix holds.
Genie One can run the saved skill on schedule after testing clears.





