Amazon cautions employees against ‘tokenmaxxing,’ as companies grapple with growing AI costs

Amazon last week reminded employees to focus their AI use on solving problems and innovation, as the tech behemoth joins other businesses in grappling with “tokenmaxxing” and big AI invoices.

“Please don’t use AI just for the sake of using AI…Use AI to help you solve customer problems, to help you solve business problems, to innovate,” Amazon SVP Dave Treadwell told staff, according to Business Insider.

The note came as Amazon axed an internal leaderboard platform that employees used to track their AI token use, a rough measure of prowess in an AI-powered era, even as many caution against conflating high usage with high output. Some employees may look to tokenmaxxing to show employers their AI buy-in—regardless if increased token spend actually equates to better outputs and more AI knowledge.

The Amazon leaderboard was one of several that emerged during an early tokenmaxxing wave, during which employees competed for top slots, indicating to employers high AI adoption and use.

But HR and business leaders are facing an all too familiar productivity challenge: leveraging metrics that encourage the wrong behavior. Amazon’s now-extinct leaderboard could have encouraged staff to use AI for work that didn’t necessarily require an AI assist. The company is one of many, including Meta and Microsoft, now trying to discourage tokenmaxxing.

AI tokens have emerged as a mascot-of-sorts at the intersection of large AI spend and AI’s promise of immeasurable productivity gains, which many organizations are still waiting to arrive. As organizations begin to grapple with large invoices for massive employee AI token use, orgs are contending with how exactly to best encourage AI use that delivers on improved business outcomes without maxxing.

HR leaders can help drive the conversation on the AI transformation toward “value creation” rather than simply productivity gains, according to Boston Consulting Group managing director and partner Julia Dhar.

“We talk a lot about productivity, and the people who talk a lot about productivity are correct that there is more output; they are not always right that there is more value,” she said.