KorvaJoin the waitlist
Startups

An AI Read 10,000 Employee Surveys and Found a Work-Life Theory Didn't Hold Up

Genie compared scores across groups, offered a counterexample, and traced complaints about inaction to different causes by department.

AI image
All-In with Chamath, Jason, Sacks & Friedberg · Databricks CEO: AGI Is Already Here, Why Companies Are Behind, and the Superintelligence Debate

Ali Ghodsi, the chief executive of data and AI company Databricks, opened his laptop to a morning inbox that had already been working all night.

His AI assistant, called Genie, runs around the clock with access to his calendar, emails, meetings, Slack messages, and company documents. It assembles those signals into what he calls an organizational map.

The other day, he said, he had turned it to a familiar headache: an employee survey answered by more than 10,000 workers.

He asked Genie to make a dashboard of the results. In seconds, it showed 88% favorability for executive leadership and 60% for work-life blend, alongside a stubbornly low score on what Databricks calls action, whether employees feel leaders act on their feedback.

That low action score bothers him, he said, because it stays low every year even though he promises the company it will be fixed.

As managers discussed the results, his finance chief offered a theory: poor work-life blend was causing the low action score.

Ghodsi said he typed the idea into Genie while the team was discussing it. Genie pulled up the correlation between the two scores and argued that the link was not causal.

As one illustration, it surfaced a manager whose team rated work-life blend at 44 but rated action at 77%. People were working hard, in other words, but still felt heard, a comparison that undercut a simple cause-and-effect story.

Pressed for the real reasons, and asked to slice the results by department, Genie gave a different answer for each group. Product managers pointed to slow decisions and too many stakeholders, engineers pointed to tools and systems, and sales operations pointed to career growth.

Ghodsi said the work would have taken his data team about a week. Genie drew on 19 relevant facts from five contributors that it had already assembled, without a live search.

Questions people ask

What exactly is Genie?

Genie is Databricks’ assistant for querying company data in plain language. In this case, it sat on top of governed survey results plus related people, team, and document data, so it could chart scores, compare groups, and read written comments together.

How did it push back on the work-life theory?

It compared work-life and action scores across groups rather than looking only at averages. Finding a team with a low work-life score but a high action score argued against simple causation, though the demo presented it as an illustration rather than a formal proof.

What does organizational ontology mean in plain terms?

An ontology is a map of how things relate, here people, bosses, teams, projects, meetings, messages, and metrics. Genie maintains that map continuously, so when a question arrives it already knows which survey response, manager, and department definition to combine.

More in Startups