The Uber Experiment: Can Agentic AI Scale from Pilot to OpEx Engine?

In the current enterprise software cycle, the divide between 'AI potential' and 'AI ROI' is the primary friction point for COOs and CFOs. Many organizations are trapped in a cycle of small-scale pilots that fail to translate into meaningful operating expense (OpEx) reductions. However, Uber is attempting to break this pattern by moving from a controlled experiment to a total company-wide rollout.

As first reported by TechCrunch, Uber has expanded its partnership with OpenAI to put its entire global workforce—approximately 30,000 employees—on ChatGPT Work. This move represents a massive scaling effort; Uber previously piloted the tool with a smaller group of 1,000 employees earlier this year.

From an operational standpoint, the most telling detail of the pilot phase is the user profile. Sumanth Sukumar, Uber's VP of engineering, told TechCrunch that roughly 80% of the early adopters were non-engineers. This suggests that the value proposition of agentic AI is not limited to accelerating code production, but is instead being tested as a productivity layer for the broader corporate machinery. Sukumar noted that these users applied the tool to a variety of high-impact tasks, including the analysis of complex operational data to spot trends and create recommendations, the localization of training materials for various languages and markets, and the creation of personalized interactive dashboards for daily management.

However, the strategic ambiguity of the rollout is where the real risk—and the real test—lies. Sukumar told TechCrunch that Uber has not yet settled on a rigid set of mandates for how employees should utilize the tool, stating that the company's goal is to provide powerful tools while allowing teams the space to experiment.

To bridge the gap between open-ended experimentation and structured efficiency, Uber implemented a targeted tactical strike: the deployment of 'AI pods.' As reported by TechCrunch, Uber paired 30 AI engineers with various teams across the organization for two-week sprints designed to automate key portions of specific jobs. This mechanism is critical because it combines technical AI expertise with the domain-specific knowledge of the employees performing the work, theoretically identifying the highest-leverage automation opportunities that a general-purpose rollout might miss.

Despite the scale of the deployment, the financial metrics remain opaque. Uber has not disclosed the total spend associated with the rollout of ChatGPT Work. Instead, the company is focusing on behavioral and outcome-based metrics. Sukumar told TechCrunch that Uber will evaluate the technology based on how employees use it, the specific workflows they develop, and the overall effectiveness of their work. These findings will then inform future AI deployment strategies across the business.

For the broader SaaS market, this is a pivotal moment. OpenAI is aggressively scaling its enterprise arm to compete with rivals like Anthropic. While OpenAI has successfully transitioned from a consumer-facing brand to a business favorite, the industry is watching to see if these massive seat-license agreements translate into leaner operations.

If Uber can prove that 30,000 employees using agentic AI leads to a measurable reduction in manual overhead or an increase in operational velocity, it provides a blueprint for the rest of the Fortune 500. If the tool remains a fragmented productivity aid used sporadically across departments, it suggests that enterprise AI may still be in the 'land grab' phase—where the primary value is captured by the vendor rather than the customer's bottom line.