Cognition CEO Scott Wu has criticized tech companies for focusing on token spending rather than AI output, after Meta and Amazon's internal leaderboards incentivizing token use backfired with employees gaming the system. Despite token prices dropping 90% since 2023, a BCG study found that 42% of workers save eight hours weekly with AI but 66% lack guidance on how to reinvest that time, indicating a fundamental misalignment between AI investment and productivity gains.
Summaries like this, in your inbox every morning.
Sign up free →What happened
Cognition CEO Scott Wu criticized companies for ranking employees by AI token spending rather than productivity, arguing the focus should shift to measuring actual output and return on investment. Meta and Amazon had created internal leaderboards to track token usage but scrapped them after employees used AI to complete useless tasks just to boost their rankings.
Why it matters
Companies like Uber have burned through entire annual AI budgets in months despite token prices dropping 90% since 2023, yet productivity gains remain elusive. A BCG survey of nearly 12,000 frontline employees found that while 42% reported AI saving them eight hours per week, 66% received little guidance on reinvesting that time, and half said they weren't spending it on strategic projects—suggesting misalignment between spending and business value.
What to watch
Wu argues that if engineers can ship three times more work, the GPU costs are justified—but only when companies measure AI by clear metrics like revenue growth, efficiency gains, or cost-saving rather than token consumption alone. Leadership clarity on AI strategy and accountability to business targets will be key to moving past the current paradox.
The 'tokenmaxxing' trend reflects a broader pattern of tech companies deploying AI tools without a clear strategy for measuring return on investment. While token prices have become cheaper—dropping 90% since 2023—companies have responded by increasing overall AI spending, emboldened by lower per-unit costs. This paradox mirrors what happened with internal leaderboards at Meta and Amazon: when incentives reward the wrong metric (tokens spent rather than value created), employees optimize for the metric rather than the business goal, resulting in wasted resources and employee frustration.
According to BCG's research, the productivity paradox extends across the workforce. Workers are saving significant time with AI tools, yet leadership has failed to communicate a clear vision for how that time should be reinvested. This gap between deployment and strategy is described by BCG's David Martin as a human-created problem of leadership communication, not a limitation of the technology itself. The implication the body supports is that companies now need to treat AI like any other workplace tool: define a business case, be selective about who has access, and hold teams accountable to measurable outcomes—revenue growth, efficiency gains, or cost-saving—rather than input metrics like token consumption.
No discussion yet for this article
Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.
Get Started FreeFree · takes 30 seconds · unsubscribe anytime
1 minute a day. The AI essentials.
200+ sources · Email / LINE / Slack
