Uber began rolling out ChatGPT Work to its global workforce on October 22, 2026, following a pilot involving 1,000 employees. According to OpenAI's announcement, the rollout targets more than 30,000 employees and provides an agent for completing tasks inside ChatGPT Enterprise. This expands access for company employees; it is not an announcement of a new tool for drivers, couriers or passengers using Uber services.
The announcement describes ChatGPT Work as Uber's first company-wide agentic platform tailored to knowledge work. It is designed to handle complex, multistep tasks and produce finished deliverables, such as presentations, data analyses or meeting preparation. For employees, the practical change is assistance throughout a workflow, rather than using a chat window only for questions and answers.
From a limited pilot to broad organizational access
Moving from 1,000 pilot participants to more than 30,000 eligible employees changes the scope of access. It does not establish that everyone has activated the product, uses it daily or completes the same tasks with it. OpenAI and Uber said they plan to share more information about adoption, use cases and measurable outcomes as the rollout progresses. The announcement provides no figures for time saved, active usage or financial returns.
Uber will support deployment with an internal AI skills library, approved integrations and dedicated employee training. Functions mentioned include operations, finance, marketing and product. This goes beyond distributing accounts: the plan includes an organizational framework intended to help people apply the tools to their jobs. The announcement does not detail the curriculum, library contents or each department's access conditions.
A skills library can provide teams with a shared starting point for recurring workflows. Defining instructions, background material and desired outputs for regularly prepared documents or analyses may reduce the need to start over each time. This is a possible operational role for such a library, not a description of a specific skill Uber has published. Its actual value depends on instruction quality, relevance to the job and updates when processes change.
The agent reaches information inside work tools
OpenAI says ChatGPT Work connects to everyday tools including Google Drive, Gmail, Google Calendar and Slack. Integrations are intended to support finding information, preparing meeting briefs, creating presentations, analyzing data and completing workflows across applications. The announcement does not list every integration actually enabled at Uber or the permissions granted to employees. It identifies approved integrations as part of the rollout framework.
Connecting applications matters because information needed for a business task rarely sits in just one place. A document may live in a drive, a changed plan in a message, and the next discussion in a calendar. With appropriate access, an agent may gather that context into a coherent deliverable. This example illustrates the product's purpose; it is not an account of a particular Uber employee's workflow.
Information access also raises permission questions. An organization must determine which data an employee and their agent can access, what they may read or change, and when approval is needed before acting. Connecting an application does not justify exposing every document to every user. The announcement does not detail a complete permissions policy, so it cannot support assumptions that all employees receive identical access or that agents may perform every action.
GPT-6.1 Sol as the default, with Codex continuing separately
OpenAI says GPT-6.1 Sol will serve as the default capability within ChatGPT Work in the rollout. It presents this as balancing advanced capability with cost efficiency, but provides no contract prices, per-employee cost or Uber-specific performance measurements. It is a reported model choice within a commercial partnership, not comparative evidence that the model is best for every task at the company.
Employees will also have OpenAI multimodal capabilities, including voice and image generation, to suit different needs. Uber's developers and software engineers will continue using Codex for software development. The expanded ChatGPT Work deployment is therefore not presented as replacing every existing AI tool. The announcement describes a combination of tools for different users and tasks, without specifying how each department will manage transitions between them.
Choosing a model and tool is only part of the process. A seemingly finished output may still need checks of figures, wording, permission to use information or suitability for a business decision. Responsibility for approval remains important even when an agent performs preparation. The announcement does not specify which Uber outputs require review, so it does not establish a uniform approval procedure.
How deployment can be assessed after access opens
The scale indicates a substantial procurement and deployment decision, but it is not a sufficient measure of benefit. Assessing change requires distinguishing employees granted access, those who try the product and those who return to it for regular workflows. Draft preparation time also differs from the time needed to approve a deliverable. These are useful assessment distinctions, not figures Uber has already published.
Further evaluation could examine which recurring tasks move to agents, how many corrections are needed and whether quality persists. Departmental differences matter: a marketing presentation, financial analysis and operations summary have different requirements. The partnership has not published metrics for each area. The announcement therefore marks the start of broader organizational deployment, with promised outcome reporting still to follow.
The announcement gives the rollout start, without a deadline for activating every employee. Its publication date should therefore be distinguished from completion of access across the organization.
The development is a change in intended agent access: from a defined pilot to a service meant for tens of thousands of employees, supported by training and approved integrations. It may expand the number of people able to use advanced tools, but results depend on actual usage and deliverables the organization accepts. Access is a starting point; the information the companies plan to publish should clarify what changed in the work itself.