Uber and OpenAI announced on Thursday, October 22, that Uber’s entire global workforce, about 30,000 people, will now use ChatGPT Work. According to TechCrunch, Uber started with a pilot of 1,000 employees earlier this year and has now scaled to the whole company. ChatGPT Work is an enterprise-focused AI agent that OpenAI introduced in July for corporate teams to deploy across workflows. The announcement is a useful case study: it shows how a large employer moves from pilot to rollout, and it raises the question that surveys keep asking, which is whether any of this pays off.
What Uber said
Uber’s VP of engineering, Sumanth Sukumar, told TechCrunch that about 80 percent of the early adopters in the pilot were non-engineers, drawn from departments including finance, operations and product. Uses he described included:
- analyzing complex operational data to identify trends and develop recommendations;
- localizing training materials for different markets and languages;
- building personalized interactive dashboards to manage daily work.
Uber did not disclose how much it is spending on the rollout. It also has not settled on a single prescribed way of using the tool. “Our approach is to make powerful AI tools available across Uber while giving teams room to experiment,” Sukumar said. Earlier this year, he added, Uber paired 30 AI engineers with teams across the company and gave them two weeks to automate key parts of their jobs; those “AI pods” showed how much opportunity exists when technical expertise is combined with deep knowledge of a particular job.
OpenAI’s own announcement describes the rollout as a partnership expansion, and says it worked closely with Uber on integrating the product.
The anatomy of a company-wide rollout
Though the details of Uber’s plan are limited, the pattern it describes matches how large organizations typically scale a new tool:
- Pilot with a mixed group. A pilot made up mostly of non-engineers tells you whether the tool helps people who do not write code, which is where most of a company’s headcount sits.
- Embedded help. Pairing AI engineers with business teams, as in the AI pods, turns a general tool into specific workflows.
- Broad access with room to experiment. Rather than mandating one use, the company lets teams find their own, then looks at what spreads.
- Measurement afterward. Sukumar said Uber will look at how employees use the technology, the workflows they build and how it helps them work more effectively, then use that to guide deployment.
For a general framework, see our guide to an enterprise AI rollout.
The measurement gap
Notice what Uber has not published: a figure for time saved, cost, or output. That is not unusual. A Gartner survey of 2,002 CIOs and other technology leaders, presented at the Gartner IT Symposium in Orlando and reported by The Decoder, found that only 13 percent of respondents report major or transformative returns from AI. In the same survey, 58 percent of CIOs are tasked with delivering cost savings through AI, yet 51 percent expect AI to raise the total operating costs of their tech stack. We chart these numbers separately in today’s Chart of the Week.
The lesson is not that Uber’s rollout will fail. It is that access is easy to count, and value is hard. Seat counts and weekly active users measure adoption. They do not tell you whether decisions got better, work got faster or customers noticed.
What employees and managers can do with this
If your company is considering a similar step, a few practices follow from the Uber example and from the research on workplace AI:
- Pick three to five concrete tasks per team before launch, and agree how you will judge them: time to complete, error rate, quality as rated by a reviewer.
- Train for the work, not the tool. A session on “how to analyze our weekly data with the assistant” beats a generic introduction to prompts.
- Set data rules early. Decide what employees may paste in, which connections the agent may use, and who reviews outputs that go to customers.
- Share working examples. Internal libraries of tested prompts and workflows spread faster than policies.
- Review after 90 days. Compare the pilot hypotheses with what happened, and cut what is not working.
Our overview of what economists say about AI and productivity is a good companion for setting expectations.
What this means for the market
OpenAI has moved to scale its enterprise business as it competes with Anthropic, TechCrunch notes. A flagship customer announcing company-wide use is useful marketing for OpenAI, and it puts pressure on competitors to show comparable deployments. For buyers, the practical point is that vendors will increasingly compete on integration and support, not only on model quality.
What to watch next
- Whether Uber or OpenAI publishes measured outcomes from the rollout.
- Pricing details for ChatGPT Work at large scale.
- Other large employers announcing company-wide deployments of agents.
- Survey data on whether the share of companies reporting strong AI returns grows beyond 13 percent.
- Governance practices that emerge around agents with access to internal data.
Key takeaways
- Uber is giving ChatGPT Work to about 30,000 employees after a 1,000-person pilot.
- About 80 percent of the pilot users were non-engineers, working on data analysis, localization and dashboards.
- Uber has not said how much it is spending or what results it has measured.
- A Gartner survey found only 13 percent of respondents see major or transformative AI returns.
- Define tasks and metrics before rollout so you can tell whether the tool helps.
Frequently asked questions
What is ChatGPT Work?
It is an enterprise-focused AI agent OpenAI introduced in July for corporate teams to deploy across workflows.
Did Uber report productivity gains?
Not in the announcement. Uber says it will track usage and outcomes and use what it learns to guide future deployment.
Why does measurement matter?
Because adoption numbers do not show value. Surveys such as Gartner’s suggest most organizations struggle to demonstrate returns from AI spending.
