# Uber Rolls Out ChatGPT Work to More Than 30,000 Employees

Uber is expanding a 1,000-person pilot across the company, with approved integrations, shared skills and training doing much of the practical work.

By Paras Madan | 2026-10-22


Uber began rolling out ChatGPT Work to more than 30,000 employees on October 22, expanding a pilot that covered 1,000 people. The change puts multi-step AI workflows into the company's general workplace tools, rather than keeping them mainly with developers or a small group of early adopters.

According to [OpenAI's announcement](https://openai.com/index/uber-chatgpt-work-rollout/), this is Uber's first company-wide deployment of an agentic platform for knowledge work. It runs within ChatGPT Enterprise, with GPT-6.1 Sol as the default model. Uber's engineering teams will continue using Codex as well.

## What Uber employees are getting

ChatGPT Work can gather context across connected applications and produce documents, presentations, analysis and other deliverables. The announcement names Google Drive, Gmail, Google Calendar and Slack as examples of tools it connects with. Uber is supporting the rollout with approved integrations, an internal AI skills library and employee training.

Those details matter more than the seat count alone. A useful agent needs access to the right information, instructions that reflect how a team works, and a clear boundary around what it may do. An employee who has to rebuild that context for every request is still doing much of the coordination manually.

[TechCrunch reports](https://techcrunch.com/2026/10/22/uber-puts-all-30000-employees-on-chatgpt-work/) that about 80% of Uber's early adopters were non-engineers, according to engineering VP Sumanth Sukumar.

That makes this a different adoption question from coding assistance. Finance and operations teams need a useful finished result, but they may not have the time or technical background to diagnose an agent's broken workflow themselves.

## How ChatGPT Work differs from a normal chat

OpenAI's [Work and Codex documentation](https://help.openai.com/en/articles/20001275-chatgpt-work-and-codex) describes Work as an agent for longer, multi-step tasks. Chat is the conversational interface; Codex remains focused on software development and technical work.

Work can operate in the cloud, and workspace administrators control access through roles and permissions. Browser use and network access have separate controls. That means giving an employee access to Work does not automatically give every task access to every company system.

Consider a weekly operations review. Summarizing a pasted report is one task. Finding the current data, checking it against earlier results, preparing a presentation and keeping the output aligned with a team's preferred format is a longer workflow. The value of an agent depends on how much of that sequence it can complete reliably, including knowing when to stop for review.

## What the rollout does not establish yet

**Availability is not evidence of productivity.** OpenAI says the companies plan to share adoption details, use cases and measurable outcomes as deployment progresses. The announcement does not provide a company-wide productivity result or disclose the commercial terms.

There is also an important cost distinction. OpenAI's documentation says Work follows Codex's usage structure, with consumption varying by task. A seat count therefore tells us little about the eventual cost of longer-running workflows. Uber's actual agreement and usage pattern will determine that.

For an enterprise evaluating a similar deployment, the useful comparison is the cost of a completed, accepted task. That includes employee review, corrections and the agent's usage, not only the time spent producing a first draft.

## The next test is whether workflows become reusable

The internal skills library is the part worth watching. If one team's successful process can be reused with the right access controls and review steps, the next team has less to build. If each employee has to invent a separate workflow, a large rollout can create a lot of experimentation without a corresponding increase in dependable output.

Uber has made access broad. The more revealing result will be whether a useful workflow remains useful after it leaves the team that first designed it.

I'm Paras. I work at the intersection of AI, engineering and distribution.

I help technical founders turn complex AI products and enterprise deployments into clear stories that developers and buyers can understand.

[Book a 30-minute call](https://cal.com/paras-madan/30min).