Access is the announcement

Read the announcement for what it actually changes. Uber is expanding ChatGPT Work from an initial 1,000-employee pilot to company-wide availability for more than 30,000 employees, with the rollout beginning on 22 October. Neither company has disclosed how many employees have activated or regularly use it. That makes this a procurement and distribution event: integrations, training and an internal skills library, extended to company scale. It is not a result. Nobody at Uber or OpenAI is claiming, yet, that the work got faster, cheaper or better, because the companies have published no measurement that would support the claim.

That distinction sounds pedantic until you notice how much enterprise AI coverage depends on blurring it. A rollout headline travels as adoption. Adoption implies people chose the tool, kept using it and produced something with it. What has been announced is only the first step, the one the buyer controls completely. The other steps belong to more than 30,000 individual employees, most of whom have not been asked yet. To their credit, the companies have not claimed otherwise: OpenAI's announcement says the two plan to share further insights into adoption, use cases and measurable outcomes as the rollout progresses. Those numbers do not exist yet.

The pilot was the interesting part

The pilot, oddly, contains the most useful information in the story, and it cuts in Uber's favour. A thousand employees is a serious trial, and its composition matters more than its size. According to Uber VP of engineering Sumanth Sukumar, speaking to TechCrunch, approximately 80% of the 1,000-person pilot participants were non-engineers, with reported use cases in functions such as finance, operations and product. The work they handed the agent was the unglamorous middle of a knowledge job — pulling operational data apart to find trends, converting training material across markets and languages, assembling dashboards they would otherwise have maintained by hand.

That is a better test of a workplace agent than another coding benchmark. Code has compilers to check it. A meeting brief, a localised training pack or a trend summary has only a reader to check it, and the reader is usually the person who asked for it. If the agent survives contact with that kind of work at pilot scale, a wider rollout is a defensible bet. But defensible is where the published evidence stops. Uber describes what pilot users did. It does not say how often they did it, how much time it saved, how often the output survived review, or what the pilot cost per active user. The expansion from 1,000 people to more than 30,000 was announced before any of those numbers were.

No price, and a recent history with AI bills

Uber declined to say what the company-wide rollout costs. That omission would matter less if Uber had not already supplied the context itself. Uber's wider rollout comes against a backdrop of reported internal AI spending controls. Earlier reporting described employee-level usage caps for agentic coding tools: in June, TechCrunch, relaying Bloomberg, reported a monthly $1,500 cap per employee for each such tool, imposed after Uber's CTO revealed the company had used up its entire annual AI budget in four months. Later remarks attributed to CTO Praveen Neppalli Naga questioned indiscriminate token spending: in August, Naga — the same executive quoted in this week's announcement — said the era of spending freely on AI, which he called tokenmaxxing, was ending, as TNW reported. Reporting at the time also said Uber had used up its entire 2026 budget for Anthropic's Claude Code by April. Treat the budget detail as reported rather than confirmed by Uber. Those reports concern Uber's broader AI usage, not the undisclosed commercial terms of its ChatGPT Work agreement. But the direction of travel was the company's own: this is an employer publicly trying to get its AI spending under control.

Against that background, an agent rollout to more than 30,000 employees with no figure attached is not a small gap in the press release. It is the figure the whole announcement turns on. Enterprise agent pricing is where the current AI market will be decided: whether these tools are cheap enough to give to everyone, like email, or expensive enough to ration, like the coding budgets Uber spent the summer capping. Uber has placed itself on the give-it-to-everyone side of that line without saying what the line costs.

Experiment first is honest, and it is also a deferral

To its credit, Uber is not pretending to have a plan it does not have. Sukumar told TechCrunch the approach is to make the tools available and give teams room to experiment, pointing to an earlier exercise in which 30 AI engineers were paired with teams across the company for two weeks to automate parts of their jobs. The company says it will watch how employees use ChatGPT Work, what workflows they build and whether it helps them work more effectively, and use that to guide future deployment. OpenAI's announcement says the two companies plan to share further insights into adoption, use cases and measurable outcomes as the rollout progresses.

That is a reasonable method and an incomplete announcement, and it can be both. Experiment-first is how a company finds out which of 30,000 jobs an agent actually fits. But it also means the rollout's justification is scheduled to arrive after the rollout. The next update from either company is the one worth reading closely: weekly active users as a share of those 30,000 employees with access, the workflows that survived, and some measure, any measure, of time or money. If those numbers appear, this announcement becomes the first chapter of a deployment story. If the next update is another expansion, readers will know which kind of story they are in.

Sources

  1. [1] OpenAI, "Uber brings ChatGPT Work to all employees with OpenAI", 22 October 2026 (primary source, read in full first-hand for this article — company-wide rollout starting today, making ChatGPT Work available to more than 30,000 employees across Uber's global workforce; expansion from an initial pilot of 1,000 employees; connections to Google Drive, Gmail, Google Calendar and Slack; GPT-6.1 Sol to be deployed as the default capability within ChatGPT Work; support from Uber's internal AI skills library, approved integrations and dedicated employee training across operations, finance, marketing and product; developers and software engineering teams continuing to use Codex; and the plan to share further insights into adoption, use cases and measurable outcomes as the rollout progresses. No price, active-user count or productivity figure appears anywhere in it)Read source
  2. [2] TechCrunch, "Uber puts all 30,000 employees on ChatGPT Work", 22 October 2026 (read in full — the 1,000-person pilot, the ~80% non-engineer share and the pilot tasks from Uber VP of engineering Sumanth Sukumar, the unsettled usage plan and the AI pods exercise, the undisclosed cost, and Uber's plan to evaluate outcomes from observed usage)Read source
  3. [3] TechCrunch, "Uber caps employee AI spending after blowing through budget in four months", 2 June 2026 (relaying Bloomberg — a monthly $1,500 cap per employee per agentic coding tool, including Claude Code and Cursor, tracked on an internal dashboard and exceedable with permission; Uber's CTO had revealed in April that the company had used up its entire annual AI budget in four months. Cited as reported cost-control context for Uber's broader AI usage, not as terms of the ChatGPT Work agreement)Read source
  4. [4] PYMNTS, "Uber Deploys OpenAI Agents to All 30,000 Employees", 22 October 2026 (account of OpenAI's 22 October announcement — company-wide availability following the pilot, the task list, developers continuing on Codex, the Naga and Brockman quotes, and ChatGPT Work's 9 July 2026 launch description)Read source
  5. [5] TNW, via Bitcomme, "Uber gives more than 30,000 staff OpenAI's ChatGPT Work agent", 22 October 2026 (source for the cost-discipline context: CTO Praveen Neppalli Naga's August "tokenmaxxing" remark and the reported exhaustion of Uber's 2026 Claude Code budget by April — the budget point is framed as reported, not confirmed by Uber)Read source