The 'use as much AI as possible' era is ending

For two years the incentive was usage for its own sake — internal leaderboards for who burned the most tokens. That's reversing. The question is no longer 'are you using AI' but 'what did that spend actually return.' If you build with AI, now is the moment to know your cost per output, not just your output. Take one real workflow, measure what it costs to run, and test whether a cheaper or open model holds the quality. The teams that win the next year aren't the ones spending the most — they're the ones who can tell you what each call buys.