Uber expands ChatGPT Work to over 30,000 employees
Uber, a company that offers ride-hailing and delivery services, is giving more than 30,000 employees access to ChatGPT Work after testing it with 1,000 staff members. Uber says this is the first time it has introduced an AI agent for knowledge work across the whole company.
Uber said approved connections to other tools, employee training and an internal library of AI skills will support the rollout across its operations, finance, marketing and product teams. The company also said GPT-6.1 Sol would be the main model used in ChatGPT Work, while its software engineering teams would continue using Codex.
The agent works inside ChatGPT Enterprise. It can use connected services such as Google Drive, Gmail, Google Calendar and Slack to collect information, study data and prepare summaries or presentations. OpenAI says users can follow the agent's progress, give it more instructions and approve important actions.
Early use in the pilot
Sumanth Sukumar, Uber's vice president of engineering, told TechCrunch that about 80% of the people who started using the agent early in the pilot were not engineers. He said they used it to study complex business data, adapt training materials for different markets and create interactive dashboards tailored to individual needs.
Sukumar also said Uber is letting teams try ChatGPT Work in different ways instead of telling them exactly how to use it. The company plans to review the work methods employees develop and find out whether they help people work more effectively.
Neither Uber's announcement nor Sukumar's interview explains how many of the 1,000 people in the pilot used the agent regularly, how much time they saved or what evidence Uber used to decide to expand the rollout.
Access controls and costs
OpenAI says company administrators can decide which connected tools employees may use and limit what ChatGPT Work is allowed to do. Its Compliance API lets companies see conversations and agent activity. Uber has not explained which permissions or approval rules it has set up.
OpenAI says the amount of usage depends on how much work a task requires. More difficult tasks may use more of a plan's allowance. Companies can set spending limits for the whole organization, groups and individual users. Uber did not tell TechCrunch how much the rollout would cost or say which spending limits it had turned on.
Uber and OpenAI plan to share more information about how many people use the agent, how they use it and what measurable results they achieve as the rollout continues.