Uber gives all 30,000 employees access to OpenAI's ChatGPT Work agent

Uber is giving ChatGPT Work to its entire global workforce, about 30,000 people. ChatGPT Work is the enterprise agent OpenAI introduced in July. Uber and OpenAI announced the wider partnership on Thursday, 22 October 2026, and Uber gave TechCrunch the details. Earlier this year the company ran a pilot with 1,000 employees. It has now extended the tool to everyone.

The size of the rollout is one part of the news. The other is who it is for. Most of the people in the pilot were not software engineers. That matters because enterprise agents were mostly tested inside technical teams, and at Uber one is now standard equipment for staff in finance, operations and product as well.

From pilot to everyone

The pilot included workers from several of Uber's departments, among them finance, operations and product. Sumanth Sukumar, Uber's VP of engineering, told TechCrunch that about 80% of the early users were non-engineers.

Uber has not said how much the company-wide rollout costs. Sukumar said Uber will instead follow how employees use the tool. "We'll look at how employees are using the technology, the kinds of workflows they develop and how it helps them work more effectively," he said. Uber will use the results to decide how it deploys AI across the business. The company says OpenAI worked closely with it on the integration.

Uber has not set fixed uses for the agent. "Our approach is to make powerful AI tools available across Uber while giving teams room to experiment," Sukumar said. In practice, the company is giving employees the tool before it has decided which jobs the tool should do.

What the pilot users did with it

Sukumar described three kinds of work from the pilot:

• analysing complex operational data to find trends and develop recommendations
• localising training materials for different markets and languages
• building personalised interactive dashboards for managing daily work

These are analysis, translation and reporting tasks. Before, they would have needed an analyst's time, a localisation vendor or a request to a data team. With the agent, the person who needs the result can produce a first version of it.

Uber also described a second, more structured experiment. Earlier this year it paired 30 AI engineers with teams across the company and gave each pair two weeks to automate key parts of the team's job. Sukumar said these "AI pods" showed how much opportunity there is when technical expertise is combined with a deep understanding of a particular job.

What ChatGPT Work is

OpenAI launched ChatGPT Work on 9 July 2026, alongside GPT-5.6. According to VentureBeat's report on the launch, the agent takes a stated outcome, breaks it into steps and can keep working on a complex project for hours. It runs on a persistent virtual machine on OpenAI's servers, so it keeps working when the user's own device is off. Through plugins it connects to tools such as Gmail, Google Calendar, Slack and GitHub. Its output is finished work: documents, spreadsheets, presentations, reports and websites.

VentureBeat reported that enterprise accounts have zero data retention and that users can opt out of having their conversations used to train future models. OpenAI product manager Ty Geri told the publication that the agent is "always in the user's control."

TechCrunch places the Uber deal in OpenAI's push to grow its enterprise business, where Anthropic is its main competitor. Anthropic's comparable product, Claude Cowork, launched in January. Signing every employee at a company of Uber's size is the kind of deal both companies are competing for.

What it means for work

At Uber, the agent takes on the first pass of analytical and preparation work: reading operational data, drafting recommendations, adapting training material for other languages and building dashboards. The pilot suggests the main users will be people outside engineering who used to depend on analysts or data teams for this work. Their job moves from producing these outputs to giving instructions and checking what the agent produces. The AI pods point to a second pattern, in which engineers spend short periods inside a team to automate parts of a particular role. Uber has not published any results on time saved, quality or changes to roles. It has said only that it will measure usage and decide next steps from that. Several questions are still open: which tasks the agent actually takes over across 30,000 jobs, who checks its output, and how Uber's teams will be organised once those tasks are automated.