Skip to content
Frontier Leaf FL markFLFrontierLeaf

A field journal of new AI developments.

Agent workflows

Databricks Genie agent executes unsupervised multi-step tasks using platform ontology

An updated AI agent performs extended, unsupervised jobs by leveraging customer data context to provide leaders with direct operational insights.

Depth
1 outlet · newest source 116h ago
Databricks Genie agent executes unsupervised multi-step tasks using platform ontology

Key points

  • Genie now completes multi-step jobs requiring extended work without human supervision.
  • The agent uses a proprietary ontology to derive context from customer-stored platform data.
  • CEO Ali Ghodsi demonstrated the tool using his actual internal instance, not a demo.

Databricks updated its Genie AI agent on Wednesday to perform multi-step jobs that require extended completion times. The company says corporate leaders can now assign a wide range of tasks for the agent to finish unsupervised.Databricks updated Genie to let leaders assign complex tasks that the agent finishes on its own.The updated Genie leverages a platform ontology to execute unsupervised, multi-step workflows with extended context.

Evidence: company claimDatabricks’ new AI agent tells CEO Ali Ghodsi the unvarnished truth

aifleet-techcrunch

The evidence

TechCrunch AI reported that Databricks CEO Ali Ghodsi demonstrated the updated Genie agent using his actual internal instance rather than a product demo. He showed how the agent analyzed employee satisfaction survey sections where the company scored poorly. The agent also helped him rate and discard suggested fixes for those low scores that some executives had proposed.

Evidence: reportedDatabricks’ new AI agent tells CEO Ali Ghodsi the unvarnished truth

aifleet-techcrunch

Ghodsi further demonstrated how Genie summarized a large sales pitch to a prospect that went poorly. This summary included specific features the prospect stated it needed before purchasing. The agent also prepared him for a podcast interview by reviewing his previous appearances and developing likely questions. While talking to the reporter, he had the agent tell him what he should say.

The decision

Ghodsi explained that with 12,000 employees, he cannot maintain direct relationships with everyone, referencing Dunbar’s number of 150. He noted that in smaller startups, CEOs know everyone and staff feel safe sharing negative information directly. At Databricks' current size, a salesperson is unlikely to confess a disastrous call to the CEO. An HR manager is also unlikely to admit that an employee satisfaction idea will probably fail.

Evidence: reportedDatabricks’ new AI agent tells CEO Ali Ghodsi the unvarnished truth

aifleet-techcrunch

The takeaway

The agent serves as a source of unvarnished truth about operational issues that leaders are unlikely to hear otherwise. Ghodsi stated that anyone with more than 150 employees can get value from this type of AI agent. He described his own use of Genie as dogfooding, which is one reason Databricks is now pitching it as an always-on agent for business leaders.

Evidence: reportedDatabricks’ new AI agent tells CEO Ali Ghodsi the unvarnished truth

aifleet-techcrunch

Building on this

Developers can build agents that operate unsupervised on multi-step tasks within the Databricks platform. The documented capability relies on an ontology that provides Genie with necessary context from customer-stored information. Builders must account for the agent's ability to analyze surveys, summarize pitches, and review historical data without direct human intervention.

Why it matters

  • Enterprise buyers can use Genie to bypass communication barriers in organizations larger than 150 people.
  • Developers can leverage the built-in ontology to give agents context from existing platform data.
  • Leaders can deploy unsupervised agents to analyze sensitive operational data like failed sales pitches.
  • The agent can prepare executives for interviews by reviewing past appearances and generating questions.

Topics: Databricks, Genie, AI agents, enterprise AI, ontology

Updated

Then / now / next

Then

Genie required more direct human oversight for complex tasks.

Now

The updated agent performs multi-step jobs with extended context unsupervised.

Next

Watch for wider adoption of always-on agents for business leaders beyond Databricks.

Story timeline

  1. Then Genie required more direct human oversight for complex tasks.
  2. 11 Nov 2026 Databricks’ new AI agent tells CEO Ali Ghodsi the unvarnished truth
  3. Now The updated agent performs multi-step jobs with extended context unsupervised.
  4. Next Watch for wider adoption of always-on agents for business leaders beyond Databricks.

The strongest doubt

Counterpoint from the sources

Ghodsi admitted he cannot know everyone at 12,000 employees and that staff are unlikely to confess failures directly to him.

Ask the article

Local keyword search of this article; no generated answers.

Did this article give you the evidence you needed?

No vote saved on this device.
Your choice stays on this device. No public tally.

Terms in this story

ontology

A structure that gives Genie context from customer-stored information on the Databricks platform.

dogfooding

Using one's own product internally, as Ghodsi did with Genie.

Dunbar’s number

The theoretical maximum of 150 people with whom humans can maintain stable relationships.

Questions readers ask

Can Genie operate without human supervision?

Yes, Databricks says corporate leaders can assign a wide range of tasks for the agent to complete unsupervised.

How does Genie get context for its tasks?

It uses an ontology that Databricks built to give Genie the context it needs from information stored on the platform.

Who demonstrated the new Genie capabilities?

Databricks CEO Ali Ghodsi demonstrated the agent using his actual internal instance rather than a product demo.

Connected stories

This storyNiCE Cognigy update prioritises agent reliability and operational visibilityNiCE Cognigy update prioritises Dartmouth provost resigns over undisclosed AI use in academic writingDartmouth provost resigns over uAnthropic IPO filing details Mythos agent cutting chip integration to three daysAnthropic IPO filing details MytAWS VP Colleen Aubrey advises using existing business metrics to measure AI successAWS VP Colleen Aubrey advises us

Reading receipt

  • Genie now completes multi-step jobs requiring extended work without human supervision.
  • The agent uses a proprietary ontology to derive context from customer-stored platform data.
  • CEO Ali Ghodsi demonstrated the tool using his actual internal instance, not a demo.

Still open: How does the ontology handle data privacy for sensitive employee surveys?

Still open: What are the specific limits of the unsupervised tasks Genie can perform?

Time spent: 0:00 · Depth: Full

Sources

  1. TechCrunch AI

Related stories

Background reading