Uber’s eye-watering AI bill shows enterprises are ‘still measuring AI success through consumption rather than outcomes’ – and it's warping our perception of ROI and productivity

Uber has been forced to introduce a monthly cap on AI use after accruing a monumental bill – and experts suggest it’s an incident that’s muddying the waters on ROI.

In April, Uber CTO Praveen Neppalli Naga revealed the company used its entire annual AI budget in just four months, per reports from The Information.

Indeed, the huge bill came after the company encouraged increased use of the technology, incentivizing staff by introducing a leader board to track consumption.

Recent reports from Bloomberg claim the company has now introduced a $1,500 monthly cap per employee, based on the tools they use - which includes Claude Code and Cursor.

Employee usage will also be tracked through an internal dashboard, although caps can be increased given employees are given express permission by the company.

Tokenmaxxing is the new industry buzzword

The move by Uber comes amid rising concerns about the cost of AI use at companies across a range of industries, with “tokenmaxxing” having recently emerged as the latest industry buzzword.

Tokenmaxxing is a trend whereby enterprises measure AI usage based on the number of tokens consumed by employees. Essentially, it’s a way to measure productivity by showcasing the use of the technology – the higher the token use rate, the more productive you are.

In March, Nvidia CEO Jensen Huang suggested that token use could be used as a means of measuring productivity among software engineers. Companies such as Meta, meanwhile, have reportedly introduced internal leader boards akin to Uber’s aimed at showing off AI use.

The trend has already sparked controversy in the industry, largely due to the fact that increased usage doesn’t necessarily equate to higher levels of productivity.

Martin Reynolds, field CTO at Harness, said the introduction of a spend cap at Uber is “treating the system rather than the cause”, and companies are basing success with the technology on usage, rather than with tangible examples of productivity improvements.

“The real issue is that many organizations are still measuring AI success through consumption rather than outcomes,” he said.

Reynolds suggested that tracking AI use “made sense” during the early days of adoption and was crucial as a means of measuring uptake levels as well as how it influenced workflows.

Now, though, usage metrics have the potential to “distort behavior” and incentivize workers to use the technology simply for the sake of it.

“Employees are rewarded for generating more prompts, tokens, and model interactions – regardless of whether those activities create meaningful business value,” he said.

Distorted ROI metrics

The long-term outcome here, according to Reynolds, is that tracking AI use in this manner could create a convoluted picture for business leaders.

Enterprises think they’re getting bang for their buck and employees are embracing the technology, but it’s needless waste - and that will distort how they measure return on investment (ROI).

“As a result, business leaders can’t confidently say whether AI usage is driving results or simply inflating the bill,” he added.

ROI with AI has been a long-running challenge for enterprises, both in terms of measuring it and whether they’re actually reporting clear-cut gains.

In a survey conducted by PwC in January, executives reported growing frustration at the lack of financial returns. Similar research from IDC shows that many organisations are disregarding poor return with AI largely due to a fear of missing out and falling behind competitors.

Measuring success

According to Reynolds, a $1,500 spending cap for Uber employees “doesn’t do much to tackle the root of the issue”. Organizations need to prioritize cost visibility and build a culture where usage can be attributed to tangible business outcomes.

This, he said, will play a key role in “highlighting whether it genuinely contributed to customer experience or revenue growth”.

“That’s how you make AI economically intelligent,” he said. “Until organizations can measure value, not just spend, they'll keep swinging between uncapped excess and blunt rationing – and leaving the real ROI on the table.”

FOLLOW US ON SOCIAL MEDIA

Follow ITPro on Google News and add us as a preferred source to keep tabs on all our latest news, analysis, views, and reviews.

You can also follow ITPro on LinkedIn, X, Facebook, and BlueSky.