Are Companies Pushing AI Too Hard, Too Fast?
The pressure to integrate AI into every corner of business isn't slowing down — it's accelerating. But a growing body of evidence suggests that the rush to deploy AI at scale is creating as many problems as it solves, particularly for the workers expected to make it all work.
Two recent reports paint a concerning picture of where some companies are heading, and offer a useful warning for everyone else.
When AI Creates More Work Than It Saves
According to reporting by The Guardian, Amazon has been pushing its employees to use AI tools across virtually every aspect of their jobs — whether or not those tools are actually helpful. Over half a dozen current and former Amazon employees described an approach that was, in their words, "haphazard," with management tracking AI usage while simultaneously deploying tools that aren't ready for the demands being placed on them.
One software developer, identified as Dina, described a job that used to be about writing code and has since become largely about cleaning up after AI. Amazon's internal "Kiro" tool, she explained, frequently produces flawed outputs, forcing her and her colleagues to spend extra time auditing work that AI was supposed to accelerate. Despite this, the message from management remained consistent: go faster, use AI more, speed is the priority.
The gap between what leadership expects AI to deliver and what frontline workers actually experience isn't new — but this example shows how damaging it can be when that gap goes unaddressed.
Dina put it plainly: "You don't look at the problem and go, 'How do I use this hammer I have?' You look at it and go, 'Is this a problem for a hammer or something else?'"
Days after speaking to The Guardian, Dina was laid off.
Token Counts as Performance Metrics
Separately, a New York Times report reveals that some tech companies have started factoring AI usage statistics directly into employee performance reviews — rewarding workers who use AI heavily and flagging those who don't.
The usage figures involved are staggering. One OpenAI engineer burned through 210 billion AI tokens in a single week — enough to fill Wikipedia 33 times over. At Anthropic, one user of the company's internal coding tool consumed over $150,000 worth of AI tokens in a single month. Some employees are reportedly competing informally on usage metrics, while managers hand out larger "token budgets" as a form of reward.
The underlying logic is that more AI use equals more productivity. But this assumption deserves scrutiny. Heavy token consumption doesn't necessarily reflect better output — it may just as easily reflect workers automating tasks indiscriminately, or generating volume without meaningful results.
It also carries a darker undercurrent. Some Amazon workers told The Guardian that by using AI at an accelerated rate, they worry they're essentially training the systems that will eventually replace them. That's not a trivial concern, and attaching it to performance reviews makes it considerably harder to ignore.
What This Means for Your Organization
If you're thinking about how to roll out AI in your own workplace, these stories offer two clear lessons.
The first is practical: forcing AI into workflows where it doesn't fit doesn't boost productivity — it undermines it. When employees spend time correcting AI errors, second-guessing AI outputs, or navigating tools that weren't built for their specific tasks, the net effect can be negative. AI works best as a targeted solution to real problems, not a blanket requirement imposed from the top down.
The second is human: tying performance reviews to AI usage — and signaling to workers that their value is measured by how much they use the technology — sends a message that erodes trust and stokes legitimate anxiety about job security. Workers who feel pressured rather than empowered are less likely to use AI thoughtfully and more likely to use it performatively.
The companies that get this right won't be the ones that move the fastest. They'll be the ones that ask the harder question first: where does AI actually help, and where is it just noise?