Accenture Once Penalized Non-AI Users. Now It Rations.

Six months ago, Accenture was threatening to slow-track promotions for employees who weren't using AI tools enough. Amazon built internal leaderboards so staff could compete on who burned the most AI tokens. The message was loud: maximize your AI usage, or fall behind.

That era is over. Accenture, Amazon, and Meta are now all restricting employee AI usage and cutting AI budgets, according to a new TechCrunch report. Amazon deleted its AI usage leaderboards. Companies are banning AI for routine tasks like converting PDFs to presentations. CFOs, COOs, and CIOs are all asking the same question at once: are we getting any return on what we're spending?

From 'Tokenmaxxing' to Token Rationing

The industry even has a term for what just ended: 'tokenmaxxing' — the period when burning more AI compute was seen as a sign of progress, and companies rewarded the habit. What's replaced it is the opposite: scrutiny. 'We're hitting this inflection point where AI is becoming material to the cost structure,' Justice Kwak, Accenture's agentic AI strategy lead, told TechCrunch. 'Leadership are still asking whether they're getting value from what we're spending on in the context of AI.'

The pivot mirrors what happened with cloud spending after the first wave of cloud adoption. Companies rushed in, bills arrived, procurement teams pushed back, and the conversation shifted from 'move everything to the cloud' to 'what are we actually getting for this?' AI is following the same arc, compressed into a shorter timeline.

What This Actually Means for People Who Build

If you're an individual building tools with AI — rather than a corporate team — this news lands differently. You're not subject to someone else's AI budget, and you're already doing the thing large companies are struggling to prove: using AI for specific tasks with a clear output you can see and measure.

The companies being cut are the ones who adopted AI as a brand move. Teams that automated a real workflow — reduced their support queue, cut research time in half, built a product they couldn't have built otherwise — will have no trouble justifying the cost. The reckoning is arriving for the ones who never defined what 'good' looked like.