Uber announced on October 22 that ChatGPT Work would reach more than 30,000 employees, expanding a pilot involving 1,000 people.
TL;DR: Uber is extending employee access. Regular usage, time savings and financial returns have not yet been established.
Deployment extends beyond software development
OpenAI’s announcement says GPT-6.1 Sol will be the default capability, with Uber supporting the rollout through an internal skills library, approved integrations and employee training. Developers will continue using Codex.
The difference between granting access and changing how a company works is consequential. An available account does not show which tasks employees delegate, how often they use the system or whether reviewing its output takes less time than doing the work themselves.
Those are questions a broader deployment can help answer. A team might find an agent useful for gathering source material but retain close control over interpretation, approvals and communication. Another might identify a repeatable process that can be delegated with clearly defined checks.
OpenAI’s July 9 product introduction describes ChatGPT Work as an agent that can use connected apps and files to produce documents, presentations and analyses. It also explains central controls over access, company context, connected tools and permitted actions.
That governance is part of evaluating the product. Allowing an assistant to read a document is a different decision from authorising it to change information or send material elsewhere. Useful deployment design needs to make those boundaries understandable to both staff and administrators.
Earlier AI work provides context, not proof of results
Uber’s relationship with OpenAI already extends beyond employee productivity. A May 6 account of the collaboration described customer-facing work, including an assistant intended to help drivers understand marketplace conditions and voice features intended to make interactions easier.
That earlier work illustrates a separate application of the technology. A driver-facing service must fit into a product used in the marketplace; an internal agent must fit the information, permissions and review processes of a particular team. Experience in one setting does not automatically demonstrate success in the other.
For the new rollout, a meaningful assessment would compare completed tasks, corrections required and repeat usage across departments. Counting accounts alone would miss cases where an apparently useful demonstration fails to become a dependable part of daily work.
The companies say they intend to share adoption insights and measurable outcomes as deployment progresses. Until then, the announcement establishes a substantial expansion in availability, rather than a quantified productivity improvement or a conclusion about how many jobs any workflow might replace.
Written by the News Desk and edited by Samuel Rae.
Leave a Reply