Uber Rolls ChatGPT Work Out to 30,000 Employees, Turning a 1,000-Person Pilot Into Company-Wide Knowledge Work

ChatGPT Work moves from a 1,000-employee pilot at Uber to a company-wide deployment covering more than 30,000 employees, Uber's first company-wide agentic platform for knowledge work. The agent had previously been used only within a pilot of 1,000 employees, with limited scale and departmental reach. Uber states the rollout reaches more than 30,000 employees across functions including operations, finance, marketing and product, supported by an internal AI skills library, approved integrations and dedicated training.

Employees can have the agent carry multi-step tasks across applications to finished work inside ChatGPT Enterprise, rather than receiving only a conversational reply. The connected everyday tools are Google Drive, Gmail, Google Calendar and Slack, so tasks can move across those applications. The stated task types include researching information, preparing meeting briefs, creating presentations, analyzing data and completing workflows across applications, while software engineering teams continue to use Codex.

GPT-6.1 Sol serves as the default capability, balancing frontier performance and cost efficiency, alongside multimodal options including voice and image generation. Uber gains access to OpenAI's portfolio of frontier models and can choose models and tools for different needs. GPT-6.1 Sol is set as the default capability within ChatGPT Work, with multimodal capabilities offered as additional options.

The report rests on the deployment announcement from Uber and OpenAI, covering more than 30,000 Uber employees worldwide across functions including operations, finance, marketing and product, with an internal AI skills library, approved integrations and employee training. The announcement says the companies plan to share further insights into adoption, use cases and measurable outcomes as the rollout progresses; those have not been published.

The announcement gives no adoption rate, task-quality or cost data after deployment, and says only that the companies plan to share adoption, use cases and measurable outcomes as the rollout progresses; how GPT-6.1 Sol as the default balances frontier performance against cost efficiency is also not quantified.