The real price of effortless work
There’s a new genre of document loose in the workplace. You’ve seen it. Polished. Structured. Grammatical. Complete. And something’s off: it took minutes to generate, seconds to forward, and the person whose name sits on it spent no effort on any of it. The report that analyzes nothing. The strategy deck that averages every strategy deck ever made. The three-page email that took four seconds to produce and wants twenty minutes of your life.
When one lands on your desk, you have two options. Both are bad.
Option one: rework it. Read it against your years in the field, fill the gaps, fix the confident nonsense. Congratulations, the four seconds the sender saved just became forty minutes of your time. And notice who paid. The effort didn’t vanish when the machine wrote the draft. It moved downstream, from a keystroke to senior judgment, from the cheapest point in the chain to the most expensive one. Multiply that across a company and the celebrated productivity gain is an accounting trick. The cost wasn’t removed. It was relocated to the people whose time was already the scarcest thing in the building.
Option two: pass it along. Feels free. It’s the pricier choice. Forwarding is certification. Your name on the chain lends the document a credibility its content never earned, because every organization runs on one assumption: whoever sends something has stood behind it. A forwarded document meets one of three ends. Someone downstream does the rework anyway, later, with less context, at a higher price. Or nobody does, and a decision gets made on top of it, and now your company is load-bearing on text no mind has ever inspected. Or nothing happens at all, and it just taxes everyone’s attention on the way to the archive.
Why does this flood exist? Because of a promise. You’ve heard it. It’s the loudest story in business right now: the machines are about to do everything a human mind can do. Cure disease. Unlock abundance. Solve what we couldn’t. Interestingly, the same voices also warn that the technology might end us. The two messages sound contradictory. They work as a team. Salvation or catastrophe, the conclusion is identical: this is the biggest thing ever, you can’t afford to look away, and the people building it deserve your money, your talent, and your deference.
So are we close or far? Honestly? I have no idea. Neither do the people who sound sure, and their certainty is usually the tell: check what they’re selling.
The case for close: the tools are real, and their progress keeps embarrassing the skeptics. Machines win gold at the hardest math competitions on earth. They produce research that professionals take seriously. Things declared decades away keep arriving on Tuesday afternoons.
The case for far: the destination has no agreed definition, so nobody can say what arriving would even look like. The dates slide. Every disappointing release becomes proof that the real thing is coming next time. The promise sits permanently close enough to demand your money now and far enough to dodge every audit.
I’ve been sold the future for twenty-five years. The internet, then mobile, then digital everything. Here’s the pattern I trust by now: the technology in each wave was real, the story wrapped around it was bigger than the technology, and the timeline always served the people telling the story.
Close or far is a fascinating question. It’s also the wrong one to run a company on.
There’s a better question, and it comes from the people who’ve lived with capable machines longer than anyone. In 1976, a computer helped prove a math problem that had resisted humans for a century. The proof was correct. Nobody seriously disputed it. And a big part of the profession was furious anyway. It took a while to articulate why, but the reason holds up: a proof whose reasoning is sealed inside a box makes nobody wiser, and making people wiser is what the work was for. Fifty years later, that fury looks like foresight. Math journals are drowning in machine-generated papers: plausible, confident, sometimes correct, and far too many for human reviewers to check. The machines have no reputation to lose, so wrong-but-confident output costs them nothing, while the field’s entire checking apparatus was built for scarce work from people with names at stake. Economists who model where this goes, if nothing corrects it, gave the destination a name worth sitting with: knowledge collapse.
Your company is a small version of that community, and the mechanics translate one to one. What a business knows is a shared practice. It’s built when people do real work. It’s certified because a name has always meant a mind. It’s transferred through common effort. The document flood attacks all three at once: production without learning, so knowledge stops being replenished; plausible volume that swamps the checking, so your internal review breaks the way the journals are breaking; output with no stake behind it, which voids the good-faith assumption your whole operation was priced on. One question cuts through all of it: does this make anyone wiser? Understanding gets built in the doing, stays in the head that did the work, and transfers to the next problem. Answers resolve one question, once, then depreciate. A company can produce more than ever while knowing less every quarter. From the inside, for a while, that looks exactly like success.
So how should you position yourself, whether you run a company or simply intend to still be worth something in five years? Five moves.
One: Use the machines, guilt-free, for everything where the result is all anyone needs. Drafts, summaries, computation, the hundred tasks where faster is the whole point. No guilt.
Two: match your checking to your producing. If the team can now generate ten reports a week but you can still only properly read three, you have two honest choices: build the capacity to check ten, or make three. Making ten and checking three isn’t productivity. It’s seven unexamined risks in circulation with your name on them.
Three: tell your people plainly: your name on a document means you can defend what’s in it, and forwarding counts. Passing something along says I’ve read this and I stand behind it. If you can’t say that, it doesn’t leave your outbox.
Four: use the third option. Send it back, with one question attached: what should I take from this? The person who produced it has the context, the sources, and the machine draft on screen. Fixing it costs them ten minutes. Fixing it yourself costs an hour and teaches them nothing. And there’s a quiet bonus: after the second return, the four-second documents stop arriving.
Five: adopt on your own clock. (I keep telling myself this one. My clock has other ideas. But I digress.) The pitch will insist you don’t have one. Adopt or die, your competitors are already ahead, there’s no time to think. Notice who benefits from your hurry: a rushed buyer skips the pilot, skips the questions, and buys the story bundled with the tool. So run it the boring way. Pilot small, write down what would prove it isn’t working before you start, expand what earns its place, kill what doesn’t. No company ever lost its market in the six months it takes to test something properly. And the louder the voice telling you there’s no time for that, the more that voice earns from your hurry.
Are we close to machines that solve everything? Still no idea. But one question survives both outcomes: is anyone wiser because this exists? Ask it of every tool, every process change, every document that survives to your desk. What I can’t tell you is what happens to the companies, and the professionals, that stop asking. I suspect we’re all about to find out.
Further reading, for the two essays that set this off, both worth your evening: “How AGI Became the Most Consequential Conspiracy Theory of Our Time” in MIT Technology Review, and “Knowledge Collapse” in Boston Review.
Leave a comment