Prince Mario-Max Schaumburg-Lippe: Uber Gives ChatGPT Work to All 30,000 Staff

Thirty thousand people is a lot of colleagues to hand a new tool to on the same week. That is what Uber just did. Uber and OpenAI announced on Thursday, October 22 that the company’s entire global workforce, some 30,000 people, will now use ChatGPT Work, the enterprise AI agent OpenAI introduced in July.

This was not a leap in the dark. Uber piloted ChatGPT Work earlier this year with 1,000 employees, then scaled it to everyone. Reporting on the announcement describes it as Uber’s first company-wide deployment of an agentic platform built for knowledge work, running inside ChatGPT Enterprise. The agent takes on multi-step jobs and returns finished work: research, meeting briefs, presentations and data analysis, with connections to tools staff already use, including Google Drive, Gmail, Google Calendar and Slack.

The pilot belonged to the non-engineers

Here is the number that tells the story. About 80% of the early pilot users were non-engineers, Sumanth Sukumar, Uber’s vice president of engineering, told TechCrunch. The pilot drew people from finance, operations and product, not just the coding teams.

And what did they do with it? Sukumar put it plainly: “They used it for a wide range of tasks: analyzing complex data from our operations to identify trends and develop recommendations; localizing training materials for different markets and languages; and building personalized interactive dashboards to manage their daily work.”

Read that list again. None of it is exotic. It is the everyday knowledge work that fills calendars in any large company, done faster and in more languages. Uber is not alone in betting on agents for that kind of work, as our recent look at governed AI agents going to work for banks and insurers showed. The pattern is the same: give the agent bounded, useful jobs, and let people check the results.

AI pods and room to experiment

Uber seeded the pilot with a neat trick. Earlier this year the company paired 30 AI engineers with teams across the business and gave them two weeks to automate key parts of their jobs. Sukumar calls those groups “AI pods,” and he credits them with showing how much opens up when technical expertise sits next to a deep understanding of a particular job.

What comes next is deliberately open. “Our approach is to make powerful AI tools available across Uber while giving teams room to experiment,” Sukumar said. There is no rigid script for how each team must use the agent. Instead, Uber says it will track how employees use the technology, the workflows they develop and how it helps them work more effectively, then use those outcomes to guide how AI is deployed across the business. OpenAI worked closely with the company on the integration, and the rollout is backed by training, approved integrations and an internal library of AI skills.

That watch-and-learn stance feels right for 2026. Even model design is shifting toward teamwork, as we saw when five small Qwen models teamed up to match a frontier giant. Big or small, the wins are coming from systems and people working in combination.

A clear message from the top

Uber’s chief technology officer, Praveen Neppalli Naga, framed the rollout around what employees can build with it. “As AI transforms how we work, we want to give everyone the tools to innovate, solve problems and help shape Uber’s future,” he said in the announcement. Engineers, by the way, keep their specialist tool: Uber’s software teams continue to use Codex for coding work, while ChatGPT Work serves the wider company.

Scale helps explain why Uber is a telling test case. This is a company that runs live operations in city after city, around the clock, in dozens of languages. If an agent can help an operations analyst in one market spot a trend sooner, or help a training team localize material for another market in days instead of weeks, the gains add up across 30,000 desks. The same appetite for practical, at-scale technology runs through our coverage of the Tesla Semi entering volume production in Nevada: new tools earn their place by doing real work, every day.

What other companies can copy

Start small, then widen fast. Uber tested with 1,000 people, learned, and only then went company-wide. Include the non-specialists from day one, because they found most of the uses. Put engineers inside business teams for short, focused sprints. And measure what people actually invent with the tool, not just how often they log in.

Will every workflow stick? Of course not. Some experiments will flop, and that is fine. The point of giving 30,000 people a capable agent, training and permission to try is that the best ideas rarely come from a planning document. They come from someone in finance in one city, or operations in another, who finds a shortcut worth sharing. Uber has just run that experiment at full company scale, and the rest of the business world will be watching what its teams build.

Also published on: