AI Automation’s Existential Workplace Crisis

In a recent internal survey, Anthropic, the company behind the Claude AI model, asked 132 of its own staff what it’s like to work with the AI all day.

On the surface, the numbers are exactly what investors want to see. They use the AI for about 60% of their tasks. Productivity is up around 50%. Claude Code can string together roughly twenty actions before it needs a human.

Six months ago it was ten.

But buried in that survey is a line that tells a different story: “It kind of feels like I’m coming to work every day to put myself out of a job.” Ouch.

That’s not some random dude who’s just playing a guessing game about the future. It’s someone in the trenches every day. You know this feeling already.

You start your day by handing work to a textbox. You open your laptop, clear the overnight noise, and drop the first task into the little AI box in the corner of your screen. It reads the description, sketches a solution before your coffee cools, and adds the tasks and follow-ups so you don’t have to. You nudge a few lines, run what it suggested, and push out results that fit your company’s standards.

Last year, those minor adjustments would have chewed up your morning. Today, it’s twenty minutes. By lunch you’ve completed three tasks, your inbox is clear, and your manager mentions how efficient you’ve been getting lately.

Five years ago, a morning like that sounded magical. Now it just feels like more work landing on fewer plates.

You didn’t just inherit new tools. You inherited a second job nobody talks about: managing the AI.

The velocity is real. That pit in your stomach is, too. And if the people at the front lines building this stuff every day are already there, that means the rest of us are driving down the same highway. We’re just a few exits back.

The moments where you used to think through a problem are vanishing. You’re checking that the project you’re creating and prepping is ready for rollout, not understanding why it works in the first place. Inside Anthropic, staff say about a quarter of their work is stuff that wouldn’t have happened without their new AI partner in crime. Cleanup tasks, experimental dashboards, improvements that were never worth doing manually.

So you’re not just getting there faster, you’re doing work that exists solely because the AI made it cheap and efficient enough to bother with. If that work only exists because the model can do it, whose job is it, actually?

At Anthropic, they talked about collaboration getting thinner. Juniors don’t ask their seniors questions as often. Instead, they just ask the AI. One manager said younger people just don’t come to them much now. Maybe you’ve experienced something like this yourself. On the surface, it sounds great, right? It isn’t.

That sound of fewer people bothering you and less interruption in your day isn’t efficiency. It’s actually the apprenticeship model breaking in real time.

The old path was simple enough: start with grunt work, build fundamentals of the business, gradually handle more complex tasks. Eventually, you’re the person who knows the industry inside and out. Except half of those ladder rungs just disappeared with more being removed with each new model rollout, while everyone’s out there pretending the ladder is fine.

No one’s going to be there to tell you you’re losing your edge. The AI sure isn’t going to. There’s no dashboard for “still understands the system from the ground up” or “can jump in and take over when the autocomplete breaks.”

You have to be the one to decide which problems you’re willing to solve the hard way, when the AI is stuck running in circles. You also have to recognize when you’re drifting from “using a tool” to admitting to yourself, “I submitted this and I don’t really know how it looks.”

As painful as it may seem to admit, your second job is now managing your own obsolescence while you get faster at your main job.

There’s no one tracking how hollow your skills are getting while the productivity graph looks better every day. That quiet, unpaid work of watching your own talents decay and knowing when to course correct now falls on you, whether you asked for it or not.

If you’re a manager, you’re managing other people’s hollowing out, too. It’s no longer just your job to manage the workload. Now it’s figuring out which work still has to happen without AI so your team can learn anything at all on their own.

It’s making people explain the why behind the AI’s solution so they don’t freeze the first time it sounds confident and wrong. Or worse, so they don’t tell the LLM “sure, go ahead and run that.” The tools aren’t going away. If your company decides to shy away from them, there’s surely someone else that will continue using them full steam ahead.

Like it or not, you’re locked into our brave new world. It’s now part of your workflow. You can’t take the offramp anymore. You can only decide how you drive.

Every time you open an AI tool, you’re training a revised version of yourself. The question that matters most is whether you’re training the one that got really good at delegating everything to the machine, or if you’re the one who stayed sharp while everyone else let their skills quietly drain away.

Read the full article here: https://medium.com/human-offset/youre-50-faster-at-work-congrats-now-you-have-a-second-job-9553a812e493