Automate the Drudgery, Keep the People

There are, roughly, two pitches for bringing AI into a company, and they lead to completely different places.

The first pitch is subtraction. It counts heads. It asks how many people you can remove, how much payroll you can cut, how quickly you can shrink the team and show a better number to whoever’s watching this quarter. It’s the pitch a lot of the market is running right now, and it’s the one I refuse to make. I tell clients, plainly, that I’m not there to see how many people we can lay off.

The second pitch is multiplication. It asks how good your people can become when the machine takes the parts of their job that were never the point (the drudgery, the copy-paste, the soul-flattening repetition) and hands them back the time and energy to do the work only a human can do. That’s the only version I’m interested in building, and not just for moral reasons. I think it’s also the one that actually wins.

“I tell clients, plainly, that I’m not there to see how many people we can lay off.”

The junior-staff dilemma

Let me be honest about the hard case, because it’s easy to preach multiplication until you’re standing in front of the counterexample.

I once helped automate a research process: a genuine fit for the technology, a task it handled beautifully, cutting the time involved by something like eighty percent. A real success by every technical measure. And then the client, delighted, promptly laid off the junior staff who used to do that work. The efficiency I built became a pink slip for the people at the bottom.

I’ve never fully shaken that. It forced a question I now carry into every engagement: who does that job in ten years? If you automate away the entry-level rung (the work juniors do to learn, to earn their footing, to become the seniors of tomorrow), you’re not just cutting a cost. You’re sawing off the bottom of the ladder that produced every expert you currently rely on. It’s short-sighted even in pure business terms, and it’s exactly the kind of “win” that looks great on a spreadsheet and terrible in five years. The technology didn’t make that choice. The client did. But I was in the room, and it changed how I talk about what I build.

You can’t just fire the other ninety-nine

There’s a little parable I use to puncture the “automate the whole team” fantasy. Imagine a hundred farmers working a field by hand. Someone invents a tool that makes one farmer as productive as a hundred. The naive conclusion is: great, fire the other ninety-nine. But that’s almost never how it actually plays out, and it misunderstands what the tool did.

Because now that one farmer with the tool can do far more than the old field required, so the smart move usually isn’t to shrink to one farmer doing the old amount of work. It’s to let a hundred tool-equipped farmers do a hundred times the work: farm more land, do it better, take on things that were never possible before. The tool doesn’t have to mean fewer people doing the same job. It can mean the same people doing vastly more, or better, or entirely new things. Which of those you get isn’t decided by the technology. It’s decided by whether the person deploying it is thinking about subtraction or multiplication. Same tool. Opposite outcomes. Entirely a choice.

Digital twins, done honestly

One of the more interesting shapes this can take is what I think of as a digital twin. You take a real, named employee and give them an AI counterpart (clearly labeled as the AI version, matched to how that person works) that covers the hours the human can’t. Five people plus five twins can cover the ground of ten, with the humans doing the real, judgment-heavy, relationship work and the twins handling the after-hours routine.

I like this model, but only with the guardrails that keep it honest. The twin is always disclosed as a machine, never a bot pretending to be the person. And the twin belongs to the arrangement, not the ether: if the employee leaves, their twin gets shut off, because it was built around a real person’s role and shouldn’t outlive its accountability. Done that way, the twin extends a valued human instead of replacing them. Done the other way (a fleet of undisclosed bots wearing employees’ names), it’s exactly the subtraction game with a friendlier interface, and I won’t build that one.

What the freed time is for

There’s a gap in the “multiply, don’t subtract” philosophy that I have to close honestly, because a cost-cutter will drive a truck through it: when you take the drudgery off someone’s plate, you’ve created free hours, and free hours look exactly like slack to a person holding a spreadsheet. If you don’t deliberately redirect those hours toward something valuable, you haven’t multiplied anyone. You’ve just built the case for cutting them, and handed it to the exact person you were arguing with. Freeing time is only half the move. The other half is designing what the time is for, and that half doesn’t happen by itself.

So you design it, on purpose, role by role. The support person who used to spend the day on password resets and status lookups now spends it on the hard calls: the upset customer about to leave, the complicated situation that needs real judgment and real empathy, the relationship worth saving that a bot would have torched. The analyst freed from assembling the same report every week now spends that time on the interpretation, the “what does this actually mean and what should we do,” which is the part that was always the point. The junior freed from rote grinding spends it learning the craft, building toward becoming the senior you’ll need in five years, the exact rung the subtraction crowd saws off. In every case the freed hour gets reinvested into the distinctly human work, the stuff the machine can’t do, the stuff that was getting crowded out by busywork.

That reinvestment is what “upskilling people into the creative brains behind the tools” actually means in practice: not a slogan, a scheduling decision. It’s the difference between a tool that makes your people more valuable and a tool that makes them redundant, and the tool is identical in both cases. What differs is whether the person deploying it did the second half of the work: deciding, deliberately, that the point of the efficiency was to buy your people the room to be more human, not to have fewer humans in the room. Get that decision right and the whole thing compounds. Skip it, and you’ve just automated your way to a smaller, sadder company and called it progress.

Better service beats cheaper service

Here’s the strategic argument for the humane version, for anyone unmoved by the ethical one. Most companies reach for AI to make service cheaper. The bigger prize is using it to make service better.

The best businesses I’ve ever studied competed on relationship and care, sometimes charging a real premium over identical products purely on the strength of how they treated people, and customers gladly paid it, and came back, and told their friends. That kind of high-touch, deeply-human service is expensive to deliver at scale, which is exactly why it’s rare and exactly why it commands a premium. Now: what if you used AI not to remove the humans from that equation but to free them for it, to strip away the administrative sludge so your people can spend their time being extraordinary to customers instead of buried in busywork? You don’t get that by racing your competitors to the cheapest possible automated experience. You get it by using the machines to make your humans more available, more informed, more present. Cheaper service is a race to the bottom that AI will happily win for everyone until it’s worth nothing. Better service compounds: into loyalty, into pricing power, into a moat.

So this is the whole of my philosophy about AI and work, and it fits on a bumper sticker: automate the drudgery, keep the people. Use the machines to delete the parts of jobs that were never worthy of a human in the first place, and reinvest every hour you free into making your people the creative, capable, deeply-human brains behind the tools. The technology is genuinely neutral about which future you build. It’ll subtract or it’ll multiply, whichever you point it at. I only know how to point it at multiply, and I’ve stopped apologizing for that being the whole business.