Uber Gives 30,000+ Staff OpenAI's ChatGPT Work

Every Uber employee is getting an AI agent. OpenAI said on 22 October that Uber has begun a company-wide rollout of ChatGPT Work, its agent for multi-step office tasks, making it available to “more than 30,000 employees across Uber’s global workforce” (OpenAI). The rollout follows a pilot with 1,000 staff.

The move matters less for what the software does than for who gets it. Uber is not limiting the agent to engineers or a single team: OpenAI names operations, finance, marketing and product among the functions that will use it. OpenAI describes the expansion as Uber’s “first company-wide deployment of an agentic platform tailored for knowledge work”. That is the two companies’ own description, and neither has yet published any figures on what the pilot achieved.

In brief

Uber is rolling out OpenAI’s ChatGPT Work agent to more than 30,000 employees, up from a 1,000-person pilot, with GPT-6.1 Sol as the default model.
It is a whole-company deployment rather than a team trial, and about 80% of pilot users were not engineers, an Uber executive told TechCrunch.
No cost, contract terms, productivity gains or effect on jobs have been disclosed. Both companies say results will be shared later.

Uber’s pilot was mostly non-engineers

One Uber executive gave TechCrunch the clearest picture of how the pilot went. The 1,000-person trial ran “earlier this year” across finance, operations and product, and “some 80%” of participants “were non-engineers”, Sumanth Sukumar, Uber’s vice-president of engineering, told the outlet (TechCrunch). He said staff used it for “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”.

Uber also tested a more hands-on model. “Earlier this year we paired 30 AI engineers with teams across Uber and gave them two weeks to automate key parts of their job,” Sukumar told TechCrunch. He said those “AI pods” showed “how much opportunity there is when you combine technical expertise with a deep understanding of a particular job”.

According to OpenAI, ChatGPT Work runs inside ChatGPT Enterprise and connects to “Google Drive, Gmail, Google Calendar, and Slack”. Employees can use it to research, prepare meeting briefs, build presentations and analyse data. GPT-6.1 Sol will be the default model, which OpenAI says balances “frontier performance and cost efficiency”. Voice and image generation are also included. Uber’s software engineers will “continue using Codex”, OpenAI’s coding agent. The wider rollout comes with Uber’s “internal AI skills library, approved integrations and dedicated employee training”.

Praveen Neppalli Naga, Uber’s chief technology officer, called Uber “a company of builders”. He said the company wants “to give everyone the tools to innovate, solve problems and help shape Uber’s future”. OpenAI’s president, Greg Brockman, called it “a big step toward expanding access to frontier AI capabilities and tools”.

What ChatGPT Work is, and what admins control

OpenAI launched ChatGPT Work on 9 July. It described an agent that can “take action across your apps and files” and “stay with a project for hours if needed”. At launch it ran on GPT-5.6 (OpenAI). Uber’s default is a newer model. OpenAI’s help centre says GPT-6.1 Sol is one of three models reserved for Work and Codex, “not available in regular ChatGPT conversations”, with a default context window of 500,000 tokens (OpenAI Help Center). The same model family reached Google Cloud’s catalogue this week, as TSN reported.

For a company of Uber’s size, the controls matter as much as the model. OpenAI says Enterprise admins can manage “who has access, what company context ChatGPT can use, which tools it can connect to, and what actions it can take”. It also says its Compliance API gives “visibility into ChatGPT Work conversations and actions at scale”. An Auto-review feature checks important actions in connected tools “before they happen”. Usage is also metered differently from ordinary chat: OpenAI says “more complex tasks may use more of your plan’s included usage”, and admins can set spending limits by group or person.

TechCrunch reports that Uber did not say how much it is spending. Uber told the outlet that OpenAI “worked closely with the company on integrating the chatbot”.

Uber is betting on access before evidence

The most striking thing about the announcement is what it leaves out. Uber and OpenAI have given a headcount, a pilot size and a list of tasks. They have given no measure of time saved, error rates, cost per employee or any change in hiring. OpenAI’s post says only that the companies “plan to share further insights into adoption, use cases and measurable outcomes as the rollout progresses”. Sukumar framed the next stage as an evaluation: “We’ll evaluate those outcomes and use what we learn to guide how we deploy AI across the business.”

That order is telling. Uber is choosing to put the agent in everyone’s hands first and measure afterwards. The pilot evidence it has shared is about who used the tool and for what, not whether the work got better or cheaper. The 80% non-engineer figure suggests demand reaches well beyond software teams. On its own, though, it says nothing about return on the spending.

The connector list is also narrow. All four named integrations are Google Workspace apps or Slack. That suggests the first job is pulling together documents, email and chat, rather than touching Uber’s ride or delivery systems. Agents with access to company mail and files raise the same questions that infrastructure such as Google’s Agent Compute Engine is trying to answer: what an agent can reach, and who checks its actions. For compliance teams, the audit trail through the Compliance API and the approval steps will matter more than the model. Neither company has said how Uber has configured them.

None of the sources breaks the rollout down by country. Uber’s UK work so far has centred on robotaxi trials, with Pony.ai in London and with Verne and Pony.ai in Manchester.

The next test is the outcome data both companies have promised. Until it arrives, the 30,000 figure measures access, not results.