Token Use Gone Wrong: Why Burning Credits is a Bad Idea

Legora’s chief tech officer, who joined the legal AI startup in 2024, warned that many firms are wasting money by rewarding employees for using the most AI tokens instead of real productivity. He explained that some teams get a leaderboard and then try to climb it by burning huge amounts of tokens on tools like Claude, Codex or Cursor. “This is a really stupid way to do anything, ” he said on the “20VC” podcast. Instead, the CTO suggests that companies focus on outcomes. He proposes hack days or demo sessions where staff show off projects and the efficiency gains they achieved, rewarding people for output rather than token consumption. “Reward them for being effective and efficient and having more output, not for necessarily using AI, ” he added. Legora’s rapid growth makes careful token use critical, the CTO noted. He questioned whether a 20 % efficiency boost justifies spending thousands of tokens, pointing out the high opportunity cost.“If we spend a ton of tokens to learn something that might only give us 20 % better, it’s not worth it, ” he said. The tech sector is shifting from tokenmaxxing to token capping, and some companies are seeing the problem. Uber recently capped each employee’s monthly spend on AI tools at $1, 500 after overspending. Amazon shut down an internal dashboard that encouraged employees to climb a leaderboard by using AI, stating it was never meant to promote usage for its own sake. Industry leaders are also calling out unlimited token budgets. Cerebras Systems CEO Andrew Feldman called the idea “boneheaded from the get-go, ” urging teams to use cheaper, open‑source models and focus on practical gains. “Use a lower-cost model instead of a Ferrari to go grocery shopping, ” he joked, stressing the need for smart token management. The trend shows that companies are learning to balance AI enthusiasm with fiscal responsibility, aiming for real productivity improvements rather than token bragging.