ChatGPT Work Enterprise Rollout at Uber: What Happened and Why

Uber is rolling out ChatGPT Work, OpenAI’s agent mode that runs inside ChatGPT Enterprise, to its whole global workforce of more than 30,000 employees, up from a pilot of about 1,000 people. OpenAI announced the deal on October 22, 2026. The agent goes to operations, finance, marketing and product teams, and Uber’s software engineers keep using Codex.

Here’s what has actually been confirmed, what’s still unknown, and what other companies planning a similar rollout can take from it. Everything below is as of October 23, 2026. This product is only a few months old and its details are still changing.

What Uber announced

According to TechCrunch’s report and OpenAI’s announcement, the main facts are:

• Scale: ChatGPT Work is expanding from a pilot of about 1,000 employees to all of Uber’s 30,000+ staff. OpenAI calls it Uber’s first company-wide deployment of an agentic platform built for knowledge work.
• Who gets it: Operations, finance, marketing and product teams. Engineering stays on Codex.
• Pilot makeup: Sumanth Sukumar, Uber’s VP of engineering, told TechCrunch that about 80% of the early adopters were non-engineers.
• Integrations: Coverage says it connects to Google Drive, Gmail, Google Calendar and Slack.
• Support: Uber is backing the rollout with an internal library of AI “skills,” a set of approved integrations, and dedicated employee training.
• Money and results: Uber hasn’t said what it’s spending. Sukumar said the company will track how employees use the tool, and OpenAI’s announcement included no productivity or time-saving figures.

A few smaller details vary between outlets. Some date the announcement to October 23 rather than October 22, and one outlet named a specific default model that no other source confirmed. I’ve left that claim out until OpenAI or Uber confirm it.

What ChatGPT Work actually is

ChatGPT Work isn’t a separate product from ChatGPT. It’s an agent mode inside ChatGPT that launched on July 9, 2026. A regular chat answers a question. Work is built to take a goal and finish it. It can act across your connected apps and files, keep working on a project for a long time, and hand back finished files such as spreadsheets, slide decks, documents and shareable web apps. Launch coverage also describes scheduled tasks, so it can check for updates and prepare reports on a regular schedule.

For enterprises, the important part is that Work sits on top of the existing ChatGPT Enterprise workspace. OpenAI’s enterprise security paper for Work says access goes through the same identity management, workspace roles and, on eligible plans, role-based access control (RBAC) that admins already use. OpenAI’s documentation also says availability depends on role, plan, workspace and region.

How it fits next to Uber’s other OpenAI tools

What employees used it for during the pilot

The pilot examples Sukumar described to TechCrunch were ordinary office tasks, not experiments:

1. Analyzing complex operations data to spot trends and prepare recommendations.
2. Localizing training materials for different markets and languages.
3. Building personalized interactive dashboards for day-to-day work.

TechCrunch also reported that earlier in the year, Uber paired about 30 AI engineers with teams across the company and gave them two weeks to automate key parts of those teams’ work. That groundwork likely explains why Uber had a skills library and approved integrations ready before going company-wide.

The context you shouldn’t ignore

The rollout comes weeks after Uber announced about 3,300 corporate job cuts in September 2026. MarketScale and others pointed out the timing. Neither Uber nor OpenAI has said the agent is replacing specific roles, and nothing we found shows that. It’s still reasonable for employees to ask what “tracking usage” will be used for.

It’s also worth being honest about the evidence. A 1,000-person pilot moving to 30,000 users shows Uber is confident in the tool. It doesn’t measure productivity, though. Until Uber publishes results, the claim that ChatGPT Work makes knowledge workers faster at Uber is unproven.

Lessons for your own ChatGPT Work rollout

If you run IT or AI adoption at another company, Uber’s approach is a useful model. Here’s a rollout sequence based on what Uber did and on OpenAI’s official admin controls:

1. Pilot with non-engineers first. Uber’s pilot was mostly non-engineers, and that’s who Work is built for. Pick teams with repetitive analysis, reporting or document work.
2. Decide who gets access by role. Use your workspace roles and, where your plan supports it, RBAC in ChatGPT to choose which groups can use Work. OpenAI’s docs say permissions add up across roles: if any role grants a feature, the member gets it. Review each person’s effective permissions carefully.
3. Check the defaults. OpenAI’s security documentation says the default workspace role in Enterprise and Edu includes Work unless an admin turns it off. If you’re not ready to roll out, turn it off on purpose.
4. Approve integrations deliberately. Connectors in Enterprise workspaces start disabled and only owners and admins can turn them on. Uber started with a short approved list (Google Workspace and Slack). Keep in mind that enabling an app doesn’t limit what data that app returns, so start with the systems you’re comfortable exposing.
5. Build a shared skills library. Uber’s internal library of reusable “skills” means employees don’t each have to work out the same workflows from scratch.
6. Train people, then measure. Uber paired the rollout with dedicated training and says it will track usage. Decide on your own success metrics before the wider launch, not after.

Bottom line

The Uber news is about scale, not a new feature. A company-wide deployment at this size shows that ChatGPT Work is now being bought as a standard tool for everyone, not just as an experiment. What’s missing is any public data on cost or results. If you’re weighing a similar move, copy Uber’s structure (pilot, approved integrations, skills library, training) and wait for real outcome numbers before believing claims about how much time it saves. Plan availability and admin options for Work are still changing, so check OpenAI’s current admin documentation before you commit.