Uber, Starbucks AI investments expose enterprise ROI gap

Uber burned through its entire 2026 AI budget in four months. Every dollar went to Anthropic's Claude Code, and the spend was not the result of misuse or runaway experimentation. Developers adopted the tool exactly as the company intended. The problem, as Quartz reported in late May, is that Uber's president and COO Andrew Macdonald could not draw a clear line between that token consumption and any measurable improvement in customer experience.

That admission, from a C-suite executive at one of the world's largest technology platforms, crystallized a tension that has been building across enterprise AI deployments. At almost exactly the same time, Starbucks quietly discontinued an AI-powered computer-vision inventory management system it had deployed less than nine months earlier. Restaurant Dive reported that employees described the system as unreliable, citing persistent miscounts and mislabeled products. The tool had been positioned as a key plank in CEO Brian Niccol's strategy to fix in-store product availability. It did not survive contact with daily operations.

The ROI gap is now a board-level conversation

Macdonald's candor carries weight precisely because of his seniority. Speaking to Quartz, he framed the core issue plainly: enterprises will need to treat token consumption as a line item weighed against headcount and other cost centers, not as a separate innovation budget shielded from scrutiny. The link between higher usage and better business outcomes, he said, is simply not there yet.

That tension is not unique to Uber. Techerati reported on a Gartner forecast projecting that by 2028, the cumulative cost of AI coding assistants will become a significant budget concern for enterprise technology leaders broadly. As generative AI coding tools scale across development teams, the token and licensing costs will require the same financial governance applied to any other enterprise software category.

For CIOs and IT procurement teams evaluating multi-year software agreements today, the Gartner projection has a direct implication. A tool that looks affordable during a 10-developer pilot can look entirely different when it runs continuously across hundreds of developers, each generating high token volumes across long working days.

Starbucks' nine-month lesson in parallel testing

The Starbucks case carries a distinct operational warning, separate from the cost question. The inventory system, announced in September 2025 as part of a broader push to improve in-store accuracy, was integrated into live operations without a sufficient parallel-run phase. According to Restaurant Dive, employees raised accuracy concerns repeatedly, and the company eventually pulled promotional materials tied to the system before discontinuing it entirely.

Computer-vision inventory tools are not inherently unreliable. But the Starbucks outcome illustrates what happens when a high-visibility operational system skips the extended parallel-run period that would normally surface edge cases before they affect store-level decisions. For operations leaders in retail, food service, and any sector where inventory accuracy directly drives customer experience, this is a concrete data point.

A pattern forming across enterprise AI

Forbes contributor Gene Marks, writing about both cases, observed that large enterprises are discovering AI deployments at scale introduce cost structures and reliability variables that were not visible during the evaluation phase. The observation holds up against the specifics. Uber's four-month budget burnout was caused by legitimate, approved adoption, not rogue spending. Starbucks' inventory system was a strategic priority, not a skunkworks project.

OpenAI, meanwhile, is pressing forward on the infrastructure side of this equation. Techerati reported that the company has unveiled a custom inference chip called Jalapeño, developed with Broadcom, which it expects to deploy with Microsoft and other infrastructure partners through 2026. The underlying cost of AI inference is something major vendors are actively working to reduce. Whether those economics flow to enterprise customers through lower pricing, or stay as vendor margin, will be a critical variable in 2027 and 2028 renewal cycles.

What this means for your team

• Model token consumption at full deployment scale before signing AI coding assistant or operational AI contracts. Pilot-scale economics are not predictive of production costs.
• Require a parallel-run phase of at least 90 days for any AI system touching inventory, fulfillment, or customer-facing accuracy before live cutover.
• Build explicit ROI thresholds and a defined review timeline into AI vendor agreements. If the link between usage and business outcome cannot be measured, the spend cannot be defended.
• Watch inference cost trends closely. Custom silicon investments by major AI vendors may shift pricing in 2027-2028; factor that uncertainty into multi-year budget models.