Key facts
- Databricks said the updated Genie can complete multi-step tasks unsupervised, drawing on an ontology built from customer data on its platform.
- CEO Ali Ghodsi demonstrated his own Genie chats, including analysis of an employee satisfaction survey and a poorly received sales pitch.
- Ghodsi said Genie surfaces candid information he is unlikely to hear directly at a company of 12,000 employees.
- Ghodsi said AI costs can be controlled through Databricks' Unity Gateway, which sets usage limits and allows cheaper open-weight models.
Databricks on November 11, 2026 unveiled an updated version of its AI agent Genie, which the company says can now perform multi-step jobs that take extended work to finish. The agent draws on an ontology Databricks built to supply context from information customers store on the Databricks platform, according to TechCrunch.
Databricks CEO Ali Ghodsi demonstrated his own use of the agent in an interview with TechCrunch, showing recent chats and charts rather than a scripted product demo. He said he used it to analyze weak spots in an employee satisfaction survey, to rate and discard fixes proposed by executives, to summarize a sales pitch that went poorly, and to prepare for a podcast interview.
Ghodsi said Genie gives him a source of candid information about problems he is unlikely to hear directly at a company of 12,000 employees. He cited Dunbar's number, the theoretical maximum of about 150 stable relationships, and said anyone with more than 150 employees could get value from AI agents.
Ghodsi said he also uses the agent in his personal life. He said corporate AI costs can be controlled through Databricks' Unity Gateway product, which lets companies set usage limits and swap in less expensive open-weight models.