- Indeed doesn't want to ever have an AI token usage leaderboard.
- Indeed's chief information officer, Anthony Moisant, told Business Insider that it can create the wrong incentives.
- "We track token use, but we use it in the background," he said.
Indeed is closely watching how its employees are using AI, but it's staying "far, far away" from any sort of "Tokenmaxxing" leaderboard.
"We track token use, but we use it in the background," Indeed's chief information officer, Anthony Moisant, told Business Insider in an interview. "We definitely are not going to use a leaderboard."
Moisant said there's nothing "inherently wrong" with the types of reported leaderboards that Meta and other Big Tech companies are using to encourage employees to use as many tokens as possible. He just wants to track metrics that are closer to outcomes.
"I think anytime we have a metric or a measure like this that is part of an incentive system, we create that perverse incentive and people start to do things," Moisant said. "Even if it's not malicious, they just inherently are following the incentive. They're following that carrot that they're chasing, and it often leads to bad results."
Across tech, there's a push to make sure employees are using AI enough in their jobs. Nvidia CEO Jensen Huang recently said that he expected an engineer making $500,000 to use the equivalent of $250,000 in AI tokens.
Tokens are how large language models break down words into numerical inputs and outputs, essentially the building blocks of AI chatbots like OpenAI's ChatGPT.
Instead of focusing on token use, Moisant said Indeed puts more stock in how quickly it is shipping products and how customers are responding to those changes.
"Activity is not valuable to our business, and it's not valuable to the job market," he said. "Outcomes are — better matching, faster matching. That's what we're going for."
Moisant said that in organizations that crave data like his, it's easy to put too much focus on what's easy to measure. It's a mistake, he said, that Indeed made before when it tracked the percentage of code written, a metric that was popular when generative AI coding tools like Anthropic's Claude Code or Cursor were starting to gain popularity.
"We used that metric for about three, four months, and we realized that, yeah, it was a fine proxy metric, but it wasn't about the outcome," he said. "It actually wasn't changing. It wasn't reshaping productivity really."
How Indeed is thinking about its AI costs
"Tokenmaxxing" also coincides with the reality that many companies are spending much more on AI than they did six months, let alone a year ago.
Moisant said Indeed's AI bills are projected to be 4x higher than in 2025, most of which is attributable to R&D usage. He said the board of directors recently discussed how to balance staying within their projected budget while not missing out on potential advances enabled by AI tools.
"We had this rich debate around, is that really what we want? Because the moment we start orienting ourselves towards budget, we also slow down the productivity that we're going after, the outcomes that we're going after," he said.
Ultimately, Moisant said it's about being vigilant about spending even in high ROI areas that basically have "the green light" to "spend as much as we need." One of those areas, he said, is for "Time-to-value related things," the time it takes for Indeed to go from an idea to shipping a product to a customer.
"I think the transparency, the understanding of how we're spending money and who's spending it in the background, we are looking at token use, but what we're looking at it from like, 'Hey, this is actually very interesting. Team one is sort of spiking. Why are they spiking? Are we seeing value?'" he said.
