Uber takes ChatGPT Work from a 1,000-person pilot to a company-wide rollout

Uber announced a company-wide rollout of ChatGPT Work on 22 October, extending a pilot of 1,000 employees to availability for more than 30,000. The change puts an agent designed to complete multi-step assignments within reach of teams across operations, finance, marketing and product. OpenAI and Uber have not yet reported company-wide adoption or productivity results.[1]

The distinction matters because a traditional chatbot can help draft an answer, while a work agent is presented as able to carry out multi-step assignments: gather material, compare files, prepare a presentation, analyse data or move information through approved tools. Uber's deployment is designed to connect common enterprise applications, including Google Drive, Gmail, Google Calendar and Slack. If those connectors are approved, an employee can request a finished work product instead of copying fragments between systems by hand.[1]

The useful unit is a finished task

The appeal is the handoff between applications. Consider a hypothetical meeting brief: an employee may need to find documents, check recent correspondence and turn that material into a usable account of what has changed. The agent’s value depends on completing that chain accurately enough to reduce work for the employee.[1]

The headline denominator is access: over 30,000 people will be eligible. The second is adoption: how many activate the tool, return to it and use it for work that matters? The third is successful completion: how many assignments are delivered to an acceptable quality, on time and with fewer corrections than the prior workflow? The announcement gives the first figure and the earlier pilot size, but not the latter two. Uber and OpenAI say they intend to share usage and outcome insights as deployment progresses.[1]

Uber is pairing access with approved integrations, employee training and an internal AI skills library. That gives teams a way to share repeatable instructions for their own work, rather than asking every employee to invent a workflow independently. OpenAI says GPT-6.1 Sol will provide the default capability within Work, alongside access to other models and multimodal tools.[1]

One rollout, several kinds of work

The announcement also says Uber developers will continue to use Codex for software engineering while broader teams use ChatGPT Work for cross-application tasks. A coding assistant may be evaluated by tests, code review and deployment incidents. A general work agent might require checks for factual fidelity, spreadsheet arithmetic, document completeness and inappropriate disclosure. One enterprise licence cannot flatten those differences.[1]

The companies say they will share adoption, use-case and outcome findings as deployment progresses. The revealing comparison will be between complete tasks before and after the rollout, including the time people spend checking and correcting the result. Moving from a pilot to company-wide availability creates the opportunity to make that comparison at scale.[1]

Report an error or suggest a correction