Uber rolls out ChatGPT Work agent to more than 30,000 employees after 1,000-person pilot
Uber is opening the ChatGPT Work agent to the whole company: more than 30,000 employees will get access, up from a pilot that started with 1,000. The agent can work directly in Google Drive, Gmail, Calendar and Slack, for example pulling together a meeting brief from emails and documents. It runs on GPT-6.1 Sol by default, which OpenAI positions as a balance of quality and cost. The takeaway: office agents are moving from pilots into production, but there are no performance numbers yet.
AI-processed from OpenAI Blog; edited by Hamidun News
Most corporate AI agent stories end at the pilot stage: a couple hundred enthusiasts try something, and then the project quietly stalls. Uber went further. The company announced that ChatGPT Work, an agent inside ChatGPT Enterprise, is now available to more than 30,000 employees worldwide. The rollout started with a group of 1,000 people. According to OpenAI, this is the first large-scale deployment of an agentic platform for office work at Uber.
What employees actually get
A regular chat works simply: you paste in a piece of text and get a response. ChatGPT Work works differently. The agent takes on a task as a whole and goes through several steps to produce a finished result. To do this, it connects to work tools: Google Drive, Gmail, Google Calendar and Slack.
Imagine an operations manager who has a meeting with a contractor in an hour. Previously, they would have manually searched for correspondence in email, opened spreadsheets on the drive, checked the calendar and compiled notes. Now they can give the agent a task: prepare a brief for the meeting. The agent gathers the relevant emails, documents and context on its own. The other scenarios in the announcement follow the same principle: researching information, building presentations, analyzing data and handling processes that span several applications.
The model layer is also notable. By default, ChatGPT Work will run on GPT-6.1 Sol. OpenAI describes it as a balance between frontier-level capability and cost. For a company of 30,000 people, this is an important detail: if every agent run used the most expensive model, the token bill would grow fast. A sensible default model, plus the option to choose other tools, including voice and image generation, helps keep spending under control. The announcement does not include specific pricing, which is worth keeping in mind.
How Uber is preparing the ground
Access to the agent alone doesn't solve much: people often don't know what to ask it to do, and security requires boundaries. That's why the rollout comes with three things. The first is an internal library of AI skills, meaning ready-made workflows already tested inside the company. The second is an approved list of integrations, so the agent doesn't get access to just anything. The third is employee training. Operations, finance, marketing and product teams are involved.
Developers are affected only partly. Uber's engineers continue to use Codex for development and code work. The result is a division of labor: programmers get one tool, and other teams get a general-purpose agent.
What this means in practice
First, it's a signal to the market. When a company of Uber's scale rolls out an agent to tens of thousands of people, the question shifts from "should we try agents" to "how do we organize access, permissions and training." Uber built its rollout on exactly these three things, and any team that plans to follow the same path will have to think them through.
Second, the link to work services is telling. An agent's value grows with the number of connected sources: without email, drive and messenger it remains a conversational partner, but with them it becomes an executor. Product managers and founders building enterprise AI products should look at this closely. Quality integrations and access control deliver more value than yet another polished interface.
Third, caution is warranted. OpenAI and Uber have not yet disclosed measurable results: how much time is saved, which tasks the agent completes reliably, and what it costs per person. The companies promise to share data on the deployment and its use cases as the rollout progresses. For now, this story is about scale and intent, not demonstrated returns. If you are planning a similar deployment, wait for those numbers or set up your own metrics from day one: the share of tasks completed without human edits, the time spent preparing routine materials, and token spend per employee.
For OpenAI, the deal is also instructive: the company is selling a bundle of agent, models, multimodal features and a coding tool. Uber becomes a showcase for how an enterprise customer can cover several types of work with a single vendor.
- Nov 21, 2026FDE job postings grew 4.6x in a year: what a forward deployed engineer actually does at the client's site
- Nov 21, 2026A swimming team captain handed ChatGPT three scheduled tasks and cut her routine work by a multiple
- Nov 21, 2026OpenAI GPT-6 Cyber Found About 35 Flaws in iPhone, Mac and Other Apple Devices
- Nov 21, 202622 Claude Agents Spent Six Weeks on Black Hole Problems — and Solved Two
Key Questions Answered
What employees actually get?
A regular chat works simply: you paste in a piece of text and get a response. ChatGPT Work works differently. The agent takes on a task as a whole and goes through several steps to produce a finished result. To do this, it connects to work tools: Google Drive, Gmail, Google Calendar and Slack. Imagine an operations manager who has a meeting with a contractor in an hour.
How Uber is preparing the ground?
Access to the agent alone doesn't solve much: people often don't know what to ask it to do, and security requires boundaries. That's why the rollout comes with three things. The first is an internal library of AI skills, meaning ready-made workflows already tested inside the company. The second is an approved list of integrations, so the agent doesn't get access to just anything. The third is employee training. Operations, finance, marketing and product teams are involved. Developers are affected only partly.
What this means in practice?
First, it's a signal to the market. When a company of Uber's scale rolls out an agent to tens of thousands of people, the question shifts from "should we try agents" to "how do we organize access, permissions and training." Uber built its rollout on exactly these three things, and any team that plans to follow the same path will have to think them through. Second, the link to work services is telling.
Build & Ship Production AI Agents Without Boilerplate Code
Start your free 72-hour trial at Hamidun Academy (academy.hamidun.com) — interactive lessons on Claude Code & Cursor, live Tech Radar, and production agent templates.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.