🎧 Prefer to listen?

Your boss just sent an email. All employees must now use AI tools — every day, tracked, leaderboard’d, and metric’d. Welcome to 2026, where using AI isn’t optional anymore, and the people who work at Amazon have already invented a word for what happens next: tokenmaxxing.

I’ve been watching this unfold for a while now, and honestly? It was inevitable. Companies spent billions on AI infrastructure. Shareholders want to see ROI. So the mandate comes down from the top: everyone uses AI, and we’re going to track it. The problem is that “tracked AI usage” and “actual productivity” are not the same thing — and the Amazon story proves it.

What tokenmaxxing actually means

The Financial Times reported that Amazon set a target: 80% of developers should use AI tools every single week. To enforce this, the company introduced individual usage targets and an internal leaderboard tracking “token consumption” — the raw units of data processed by their in-house AI agent.

What happened next was entirely predictable. Employees started gaming the system. They call it “tokenmaxxing” — running personal tasks through the company AI just to inflate their numbers. One employee on Team Blind described launching 10 sub-agents to analyze Slack conversations whenever their project manager said something dumb. “Great use of GPUs,” they added.

This isn’t one rogue employee. Multiple workers across Amazon described the same pattern: the pressure to use AI creates “perverse incentives” where the metric becomes the goal, not the actual work. It’s Goodhart’s Law in real time — when a measure becomes a target, it ceases to be a good measure.

Why forced AI adoption backfires

I’ve talked to people at other companies too, and the pattern is the same everywhere. When you mandate tool adoption from the top down, you get compliance theater. People use the tool just enough to stay off the radar, not because it actually makes their work better.

The real problem isn’t resistance to change — it’s that AI tools don’t help with every task. Writing a first draft? AI writing tools can genuinely speed that up. Summarizing meeting notes? Sure. But a lot of developer work, strategic thinking, and creative problem-solving doesn’t get faster just because you throw an LLM at it.

If you’re an individual trying to figure out which AI tools actually matter — before your company forces one on you — I wrote about the tools I actually use every day. That list comes from real usage, not corporate mandates.

The metrics nobody’s asking about

Here’s what’s wild about the tokenmaxxing story. Amazon isn’t just tracking this internally — they’re using these numbers to justify hundreds of billions in AI infrastructure spending. If a meaningful share of token consumption is performative, what does that mean for the financial projections driving the entire AI boom?

SnapChat employees on Team Blind are openly coaching each other on how to inflate their numbers. “If companies use brain-dead metrics to judge people,” one worker wrote, “then you need to learn how to f**k them over right back.”

That’s not a healthy adoption curve. That’s a system eating itself.

What actually works instead

Forced adoption is the wrong approach. Here’s what I’ve seen actually get people to use AI tools effectively:

Start with the pain, not the tool. If someone’s drowning in repetitive tasks, show them how automation in 15 minutes can help. If they hate writing status reports, demonstrate what a good prompt can do. You can’t push people toward a solution before they’ve felt the problem.

Let people discover naturally. My automation pipeline didn’t come from a mandate — it came from me being annoyed at wasting time. The best AI adoptions I’ve seen are bottom-up, driven by people who found a genuine use case.

Measure output, not input. Token consumption tells you nothing. Track tasks completed, time saved, quality improvements. If someone ships better work faster because they used AI, great. If they’re burning tokens to hit a leaderboard, that’s just waste.

Build the culture, not the mandate. I didn’t plan to learn AI tools — I just stumbled into it because the tools solved a real problem. That’s the kind of adoption you want. Organic, because the value was obvious.

For non-technical teams — this matters even more

If you’re not a developer, forced AI adoption can be even worse. Non-technical employees get handed tools they don’t understand, with zero training, and are told to hit metrics. That’s a recipe for frustration, not productivity.

I wrote a beginner’s guide to no-code for exactly this reason. The tools should work for you, not the other way around. If your company is pushing AI tools and you’re not sure where to start, building your first AI workflow is a better investment of your time than gaming any leaderboard.

And if you’re worried about privacy when your company mandates AI tool usage — you’re right to be concerned. I covered the privacy problem nobody talks about in detail.

The bottom line

Amazon’s tokenmaxxing problem is a preview of what every company will face as they push AI mandates. The companies that figure out organic adoption — where people use AI because it genuinely helps, not because a dashboard says they should — will win. The ones that keep mandating usage metrics will keep paying employees to game them.

If you want to start using AI tools because they actually help — not because a leaderboard says you should — head to Start Here and I’ll walk you through it.