Ethan stares at the blinking cursor of Amazon’s internal AI assistant. It’s 3:00 PM on a Tuesday in a glass-walled office in Seattle. He doesn’t actually need the AI to help him. He knows exactly how to write the email he’s currently drafting. He’s been doing this job for six years. But Ethan triggers the prompt anyway. He asks the machine to "summarize" a three-sentence paragraph he just wrote.
He waits. The blue loading bar crawls across the interface. The AI spits back a version that is slightly wordier and significantly less clear. Ethan hits "accept." For an alternative view, see: this related article.
He isn't being lazy. He is being a survivor.
Deep within the sprawling architecture of Amazon’s corporate culture, a new ghost has entered the machine. It isn't the AI itself, but rather the way the company measures human "success." When leadership decides that a new tool is the future, they don't just ask people to use it. They track it. They graph it. They turn "usage" into a high-stakes leaderboard. And as any student of human nature knows, when a measure becomes a target, it ceases to be a good measure. Similar analysis on this trend has been shared by Forbes.
The Algorithm of Anxiety
Consider a hypothetical corporate ecosystem—let's call it the "Efficiency Trap." In this environment, every keystroke is data. When Amazon rolled out its suite of generative AI tools for employees, the intent was noble on paper: reduce the "drudgery" of corporate life. The tools were supposed to draft memos, debug code, and find internal documents.
But a funny thing happens when you tell ten thousand Type-A overachievers that their performance reviews might be tied to how much they "leverage" new technology. They start using it for everything. Even the things that take longer with AI than without it.
Imagine a software engineer who can fix a bug in thirty seconds. Instead of just typing the fix, she now prompts the AI to explain the bug, then prompts it to suggest a fix, then spends two minutes correcting the AI’s hallucinated syntax. Why? Because somewhere in the cloud, a dashboard shows her "AI Engagement Score" is at the 90th percentile.
She is trading her actual time for the appearance of digital fluency.
This isn't just a quirk of the Amazon office. It is a fundamental shift in how we value work. We have moved from valuing the output—the finished, working code—to valuing the process—how much "innovation" was sprinkled on the task. It is a performance in the most literal sense. It is theater.
The Invisible Toll of the "Usage" Mandate
There is a psychological weight to doing work that you know is unnecessary.
Anthropologist David Graeber famously wrote about "bullshit jobs"—roles that serve no purpose other than to make someone else feel important or to fulfill a bureaucratic requirement. What we are seeing now is the birth of the "bullshit task." These are real jobs, performed by brilliant people, who are being forced to inject artificial complexity into their day to satisfy a metric.
The friction is real. Every time an employee stops to engage an AI for a task they could do manually in half the time, they break their "flow state." They lose the thread of their creative thought. They become an assistant to the assistant.
The irony is thick enough to choke on. A tool designed to save time is being used to waste it, all so a middle manager can report to a Senior VP that "95% of the department has integrated AI into their daily workflow."
Numbers don't lie, but they often hide the truth.
If the data says usage is up, the project is a "success." But the data doesn't show the frustration of the developer who had to fix the AI's mistakes. It doesn't show the exhaustion of the project manager who spent an hour prompting a bot to write a meeting summary that he could have dictated in five minutes. It doesn't show the quiet erosion of morale when smart people are treated like data-entry points for a machine learning model.
Why We Cheat the System
Humans are remarkably good at identifying the path of least resistance to a reward. If the reward is a "Meeting Expectations" rating on a performance review, and that rating is linked to AI usage, the human will find the fastest way to use AI.
We see this pattern throughout history. In the early days of the Soviet Union, factories were often given quotas based on the weight of the goods produced. When a nail factory was told to produce a certain weight of nails, they simply made a few massive, unusable, five-pound nails. They met the metric. They failed the mission.
Amazon's staff are simply the modern version of those nail-makers.
The "five-pound nails" of the modern office are the pointless summaries, the over-engineered emails, and the AI-generated slide decks for meetings that could have been a Slack message. We are generating "content" at an industrial scale, not because it helps us communicate, but because the act of generation is what is being tracked.
This creates a feedback loop of mediocrity. The AI learns from the "usage" data, thinking these unnecessary tasks are what humans truly value. The humans continue to perform these tasks to satisfy the managers. The managers use the high usage numbers to justify buying more AI licenses.
Everyone is "winning," yet the work is getting worse.
The Ghost in the Dashboard
There is a specific kind of loneliness that comes from working alongside a machine that is being used to monitor you.
When you speak to people who have spent their lives in these high-pressure environments, they describe a feeling of being "hollowed out." It is the sense that your expertise—the years you spent learning how to craft a perfect argument or how to spot a flaw in a system—is no longer the point. The point is to be a conduit for the company's latest technological obsession.
But the real problem lies elsewhere. It isn't just about wasted time. It’s about the death of intuition.
When we outsource our basic thinking to a tool just to inflate a score, we stop practicing the skills that made us valuable in the first place. If you never write your own summaries, you eventually lose the ability to synthesize information quickly. If you never debug your own code, you lose the "feel" for the logic of the system.
We are at risk of becoming a generation of "prompt engineers" who don't actually understand the engines we are prompting.
The Mirror of the Metric
The story of Amazon staff gaming the AI system is a mirror held up to the entire tech industry. It asks a painful question: Are we building tools to help people, or are we building tools to prove that we can build tools?
If the goal was truly productivity, the metric would be simple: "Did the work get done faster and better?" Instead, the metric is: "Did you use the AI?"
This distinction is the difference between a tool and a master. A hammer is a tool; you use it when you need to hit a nail. If your boss told you that you had to use a hammer for 40% of your day, you’d eventually start hammering screws, hammering the table, and hammering the walls just to keep your job.
That is the state of the modern corporate office. We are all just looking for something to hammer.
The danger is that this behavior becomes the new baseline. We stop noticing the absurdity. We accept that "work" is just the act of moving data from one AI window to another. We forget that the most valuable things a human can bring to a company are the things an AI cannot do: empathy, moral judgment, and the ability to say, "This task is unnecessary."
But in a culture where "usage" is king, saying a task is unnecessary is an act of rebellion.
Ethan finally hits send on his email. He looks at the "AI Assisted" tag at the bottom of the draft—a digital badge of honor that tells his bosses he is a "forward-thinking innovator." He feels nothing like an innovator. He feels like a man who just spent twenty minutes doing a ten-minute job.
He closes his laptop and walks toward the window. Outside, the Seattle rain is steady and gray, a rhythmic patter against the glass. It is a natural system, indifferent to metrics, unbothered by usage scores. It just falls because it must.
Inside, the blue lights of the servers hum, waiting for the next unnecessary prompt, hungry for the data that proves they are needed, while the humans behind the screens quietly wonder when they stopped being the ones in charge.
The screen stays dark for a moment before Ethan's phone pings. It’s an automated notification from the internal system, congratulating him on reaching his "AI Engagement Goal" for the week.
He doesn't smile. He just puts the phone in his pocket and walks away.
