Databricks is pitching the new Genie as an “always-on, autonomous AI coworker.” In posts on November 11, 2026, @Yuchenj_UW announced the launch and Ali Ghodsi described using it for stalled projects, internal project tracking and culture surveys. That’s a bigger brief than drafting a polite follow-up email. Source Source
“Today, I just wake up and use Genie, which works for me in the background,” Ghodsi wrote. He says he often uses the phone app to queue work. The tasks he describes used to take him “hours and many meetings, often waiting for the data science team.” His post gives no measured comparison of how long those tasks take with Genie. Source
What’s on the CEO’s list?
Ghodsi offered three examples from the week:
- Organizational bottlenecks: While considering how to flatten the organization, he says Genie found hotspots where projects had become stuck, using information already in its ontology graph. Source
- An internal project called Keanu: He says Genie monitored its Slack channel, correlated insights with source code and produced a dashboard. The post doesn’t explain what Keanu does. Source
- Culture-survey analysis: Ghodsi says Genie used org charts, work output and survey scores to investigate low scores, including work-life balance, and assess which leaders were “AI pilled.” Source
That last assignment deserves more scrutiny than a calendar summary: it involves interpreting information about employees and managers. Ghodsi’s post doesn’t describe how those judgments were checked. Source
Business context is the pitch
@Yuchenj_UW describes Genie Ontology as a context layer that automatically learns how a business works from its data, queries, dashboards and notebooks. Source
For ongoing work, BAD SIGNAL reports that Databricks says Genie tracks progress, follows up on missing inputs and acts as new information arrives. Jobs can run on schedules or triggers, while people set limits, supervise the work and approve steps requiring sign-off. Source
BAD SIGNAL also reports Databricks’ claim that Unity Gateway lets administrators restrict tools, trace and audit actions, and set workspace and per-user budgets with spending caps and alerts. The report doesn’t describe independent testing of those controls. An always-on coworker apparently still needs an expense policy. Source
Comments