Uber rolls out ChatGPT Work to 30,000 employees

Uber is putting ChatGPT Work in the hands of 30,000 employees, extending its use of AI beyond the teams that build software. The company has already tested the tool with a smaller group of workers, including people in finance, operations and product development.

The uses have been fairly practical. Employees have used it to examine business data, prepare training material for different markets and build dashboards. These are the kinds of jobs that can keep staff busy for hours, particularly when information has to be collected from different sources before anyone can make sense of it. Uber is now taking the experiment to a much larger workforce. What remains unclear is whether the wider rollout will save the company enough time and money to justify the cost.

Employees are finding their own uses

During the initial trial, about 80% of participants were not engineers, according to Sumanth Sukumar, Uber's vice-president of engineering. That matters because much of the discussion around AI at work has focused on software development. Uber's experience suggests there are other areas to explore, from analysing operational figures to preparing material for employees in different markets.

The company has also tried pairing AI engineers with other teams to identify work that could be automated. In one exercise in 2026, 30 AI engineers worked with teams across the business for two weeks. Instead of expecting every department to use the technology in the same way, the exercise focused on the tasks employees were already doing.

That approach is relevant to smaller businesses, where staff often handle several responsibilities at once. An employee might spend part of the morning responding to customers, then switch to preparing quotations and updating sales records. If AI can take some of that administrative work off their hands, it could free up time for serving customers or finding new business. But the tool has to work in practice. A quotation containing incorrect prices or a report based on faulty figures can create more work than it saves.

The price of the rollout is still unknown

Uber has not disclosed the total cost of extending ChatGPT Work to its workforce. It has also not published figures showing how much time the initial trial saved or whether it reduced operating expenses. Without those numbers, it is difficult to judge the financial case for the rollout.

The subscription is only one part of the cost. Employees need to learn how to use the tool, and somebody has to check the information it produces. Businesses also need clear rules about what staff can share with an AI system, particularly when customer details or confidential company information are involved. These considerations carry weight for smaller companies. An SME paying for several AI subscriptions needs to know whether the tools are helping employees finish work sooner, handle more customers or reduce the need for outside services. If staff still spend the same amount of time completing a task, the business needs to ask what it is getting for the additional expense.

A useful test for South African SMEs

A business owner does not need to introduce AI to every department to find out whether it is worth using. A better place to start might be a job that comes around every week. A wholesaler could test whether AI helps staff prepare customer quotations. An accounting practice could use it to organise information before compiling reports. A retailer might try it on routine customer enquiries.

The business can compare the time taken before and after the change, check for errors and work out whether the savings cover the cost. Uber has the resources to test AI across thousands of employees at once. Smaller businesses have less room to spend money on tools that do not deliver, so they need to be more selective about where they start. For now, Uber's decision shows the scale at which a major company is prepared to deploy AI. Whether that decision pays off will depend on results the company has not yet made public.