Expensive People Doing Cheap Work

There's a combination that should be impossible: being bored and exhausted at the same time.

A few weeks ago, someone who runs a company's sales branch described it to me. She clocks in at six thirty in the morning, gets home late, and is bored. I asked her to walk me through her day.

The cycle goes like this. She pulls prospect data from the CRM. She calls them to invite them to visit the site. When one says yes, she opens the calendar and books the visit by hand. Then she goes back to the CRM and enters the contact by hand. And she also transcribes, one by one, the card payments customers make in person, so the outside accountants can see them.

That last one is just the detail that stuck with me. Her day is full of tasks like it: reports built by copying from one place to another, forms that have to be filled in twice, lists that get cross-checked by hand, confirmation emails written one at a time. None of it is hard. That's exactly the problem: none of it is hard, and together it eats the whole day.

Five steps. Four are mechanical. The only one that truly needs a human being is the call — the conversation where she gets a stranger to agree to come see the site. From there comes the visit: the walkthrough, the awkward questions, reading who's in front of her, the close. That's where the customer is won or lost, and no machine can replace her there.

And there's a third piece, the one that finally convinced me. Her site had to produce a safety-protocol document required by regulation: eight protocols, twelve pages, with legal requirements to track down one by one. I researched and drafted that with my AI in an afternoon. What I couldn't do — and no machine could — was what came next: rolling it out, training the team, holding the standard day to day, and giving every customer the confidence that things are done right there. That's management. The document was the pretext; the leadership was the work.

This person isn't underpaid. She's paid well, and for good reason: someone in that organization did the math and decided that making that site work is worth that money. The problem is something else, and it took me a few days to put into words:

The salary is well aimed at the person and badly aimed at the task.

Hold on to that line. It's the whole essay.

What I read on Saturday

That same weekend I read an opinion column in El Espectador that, without meaning to, finished putting the idea together for me. Two things stuck with me. First: the time technology saves us doesn't stay free — if writing an email costs less, we write more emails; if calling a meeting is easy, we call another one; if you can work from anywhere, anywhere becomes work. People have started calling this the infinite workday. Microsoft put a number on it in its 2025 Work Trend Index: the average worker gets 117 emails and 153 Teams messages a day, an interruption every two minutes. The second thing, the one that matters: automating an absurd process doesn't make it less absurd. It just lets you run the absurdity much faster.

That's when I realized the story of the branch manager and the story of the infinite workday were the same story, told from two different heights.

Why the inbox always wins

The reason email overflowed for everyone isn't that people are rude or lack discipline. It's arithmetic:

Sending an email costs the sender seconds and the recipient minutes.

In economics, that has a name: it's an externality. Whoever creates the cost doesn't pay it. And externalities don't get fixed with willpower, productivity workshops, or right-to-disconnect policies. They get fixed by changing who pays.

The same thing — exactly the same thing — explains something I see in every company I walk into. What organizations call "processes" is almost never design: it's sediment. Every step exists because something once went wrong and someone added a control. And nobody ever removes a control, because removing one means signing your name on the next failure, while adding one costs the person who adds it nothing. That's why processes only ever grow. That's why there are signatures nobody reads, forms nobody processes, and approvals whose only effect is to make the decision take three more days.

They aren't processes. They're habits in uniform: myths, assumptions and hierarchies nobody ever put to the test, many of them born because someone didn't want to do something, or because someone wanted safety for its own sake. (That topic deserves its own essay, and I'll write it. For now, it's enough to name it.)

It's the same law of email, in a different dimension: adding friction is free for whoever adds it, and everyone else pays for it.

The inbox sorts by date, and that's the anti-strategy

Now, the question almost nobody asks: if the noise hits every company the same way, why does it sink some and not others?

Because prioritizing isn't a character trait. It's a function of a judgment written down beforehand.

When an organization hasn't made its direction explicit, every signal that reaches it weighs the same. The client who shouts doesn't win because he shouts: he wins because there's nothing to measure him against. The urgent email displaces the important project not out of real urgency, but because nobody defined what mattered clearly enough to defend it.

And then look at what literally happens on that company's screen:

The inbox sorts by date. Chronological order is the anti-strategy: it's the order of whoever shouts loudest and most recently. Any organization without an explicit direction ends up running its inbox's agenda, because the clock is the only ordering principle it has left.

The sale is a terrible advisor

There's a version of this that shows up in small and mid-sized companies with brutal clarity, and it's the most expensive one of all.

Most small and mid-sized companies don't chase a strategy: they chase the sale. Any sale, regardless of who brings it or what the customer asks for. And because the sale rules, the company ends up doing things that aren't really its own — work outside its business model, jobs it doesn't know how to do well, clients that don't suit it — rather than let the invoice slip away.

And that's how a company gets deformed. It starts with a business and a value proposition — when one exists, because often it isn't even defined, or never quite amounts to value. But because the owner isn't in charge of his own destiny, and is instead a machine of chain reactions, some time later he looks back at what he built and finds a Frankenstein: initiatives stitched to one another, unrelated, each one dragged in by a different emergency. The production line stops because a more urgent Frankenstein order came in. It gets produced. And productivity collapses.

Nobody notices the damage because it doesn't show up in the sale: it shows up in everything else. Every off-model job eats the capacity that the actually profitable work needed, muddies the promise, confuses the team, and teaches the market something different from what you wanted to teach it. By year end you've billed, and you're not stronger than you were in January. You're blurrier.

And the uncomfortable part has to be said, because it's the useful one:

When you have to chase the sale, it's because value isn't guiding you. You chase because you aren't creating enough value for people to seek you out, because you don't know your market, and because you're moving on your own cash urgency instead of the market's needs.

A company like that isn't undisciplined. It's a weathervane in the wind. And at year end its results aren't the ones it chose: they're the ones the market decided for it. Not the market it defined, created and won — the one it happened to land in.

The real reason: activity is the currency

There's one piece left that ties it all together, and it's uncomfortable.

In office work, the outcome is invisible and the activity is visible. Nobody sees the quality of a decision; everybody sees who answered at ten at night. And since you have to pay for something, you end up paying for the only signal available.

Visible activity becomes the currency you use to pay for a value nobody knows how to measure. And that's where the consequence comes from that explains the last thirty years of tools:

Making that currency cheaper to produce is inflation. When writing costs less, people write more and each message is worth less. When calling a meeting costs less, people call more meetings and each one is worth less. It isn't a metaphor: it's the same mechanism by which printing money doesn't create wealth.

And now look at the whole pattern, because once I saw it I couldn't unsee it: every productivity tool of the last thirty years installed itself on the sender's side. Easier to write. Easier to call a meeting. Easier to share. Easier to cc ten people just in case.

Not one, not a single one, installed itself on the receiver's side.

What changes now, and why this isn't just another tool

That's where Saturday's opinion column falls short, when it concludes that AI is just going to put a turbocharger on the error. That's the default scenario, and it's right to fear it: turn AI loose on a disorganized company, and it multiplies the disorder.

But AI is the first technology that can install itself on the receiver's side. And above all, it's the first one that's actually intelligent, and that word here isn't decoration.

A spam filter always saw words. This sees something else: it understands the thread. It reads a forty-reply chain in full and can tell you what no earlier software could tell you — that there was no work happening there, just a decision nobody made. It can flag that three people are asking for the same piece of data because nobody defined where it lives. It can show you that one step in your process produces nothing and only exists so that someone isn't left exposed.

For the first time, it's a tool capable of separating signal from noise inside a company's conversation. And in the hands of someone with judgment, that does something that was never possible before: turning a habit into a process. Taking the routine nobody designed, seeing where the energy leaks out, cutting what's excess, and handing it back as something that actually pursues a result.

I underline in the hands of someone with judgment, because without that, nothing good happens. The machine proposes; the judgment of what's signal and what's noise stays human, and it depends entirely on that person knowing where the company is going. Without that direction, AI just speeds up the road to nowhere, and it does it very fast.

The question a CEO gave me back

I told all this to a CEO with over thirty-two years of experience in multinationals and Latin American conglomerates. She heard the whole argument and answered with a one-line question:

"What percentage of your time do you spend on the business, and what percentage on the task?"

It's better than everything I had written so far.

And it has a pedigree: Michael Gerber said it thirty years ago in "The E-Myth," explaining to a bakery owner that his job wasn't baking bread. Work on your business, not in your business.

But Gerber was talking to the founder. And that's exactly what's new about what we're living through now:

What Gerber diagnosed as the owner's disease became everyone's condition. It's not just the founder trapped in the task anymore. It's that the entire organization is full of expensive people doing cheap work, and nobody notices because everyone is impossibly busy.

The person I opened this essay with isn't a partner in anything. And the diagnosis fits her perfectly.

Business leaders

Here it's worth clarifying something that's a central distinction in my work.

When I say the organization needs to know where it's going, I don't mean the president and the vice presidents need to know it. I mean business leaders: everyone essential to making a decision or implementing it, wherever they sit on the org chart. There are almost always far more of them than the company calls "leaders," and they're almost never all in the executive committee.

Go back to the person I opened with, the branch manager. She's on no committee, approves no budgets. And she fits the definition of a business leader exactly: if she doesn't execute well on that call and that visit, there's no sale at that site. Hers is probably one of the most important roles in that organization — without it, there's no revenue. And it's the same role we keep busy all day transcribing payments.

And this stops being a nuance the moment you bring AI into the company, because the new role that appears isn't operating the Mind — the AI that carries the company's corpus, its way of deciding and its context — it's orchestrating it. Someone has to ask it the right thing, evaluate what it returns, catch when it got it wrong, and decide what to do about it. That's applied judgment, and judgment can't be applied without knowing the strategy. A company where only five people understand where it's going can survive with manual processes. With agents, it can't: it multiplies five judgments and leaves everyone else automating what they always did.

One warning, and I'll close with it

If someone reads this and concludes, "so AI gives me back hours," they didn't understand a thing. And they're heading straight for the scenario Saturday's column describes, which is worth seeing in detail because it's what's going to happen at most companies this year:

You automate. What took eight hours now takes five. And those three freed-up hours don't turn into better decisions, or a better-served customer, or time to think. They fill themselves back up: with the emails that are now cheaper to write, the meetings that are now easier to call, the reports that now get produced with one click and that nobody's going to read. Noise is a gas: it fills all the space available. And by the end of the quarter you worked just as fast, just as late, and produced exactly the same thing. You went in a full circle. On that point, the column is completely right, and it deserves full credit.

The difference between the two companies — the one that goes in a full circle and the one that doesn't — isn't the technology. Both bought the same thing. It's whether someone decided, beforehand, what those three hours would be used for.

And the right answer, the only one that makes the math work, is this: those hours go back to the piece of the work only a human being can do. In the story I opened with, the call and the visit that end in a sale. In your company, whatever depends on someone judging, persuading, negotiating or deciding.

That's what really happens when this is done right. It's not that the person works less: it's that the 20% of their work that produces 80% of the value comes to fill their whole day. You take away the filler work and expand the ground where they're strong. You squeeze the full value out of what they do.

And a second effect shows up, the one that changes the conversation with the manager: for the first time, it can be proven. What that person delivers stops being an impression and becomes a number — how many of her conversations turn into business, how much each one is worth, how much that grew since she stopped transcribing payments. The evidence doesn't change her; reassigning her time does. What the evidence does is make that value visible, which is the only thing missing for the company to stop counting her the way it always had.

Because here's the arithmetic that decides all of this:

A cost is measured by what it consumes. An asset is measured by what it returns.
As long as you measure a person in hours, you're booking them as a cost. By definition, not by opinion.

I call that Managerial Capital, and it's literal: turning talent into a productive asset. Take away the filler work, expand their strengths, and with that, make the person capable of managing their own work toward the business's goals — so that everything they do creates value and profit. It isn't a title. It's a condition you give back to people when you stop spending them on what doesn't pay off.

And look at the irony we've been living with for decades: the biggest share of cost and spending at almost any company is people's compensation. We have departments called Human Talent, Human Capital. And yet we neither extract that talent nor turn it into productive capital. In large part, we pay people to do things a machine does better, faster and without mistakes.

That's about to come to light. And the day it does, every company will face two possible conversations.

The first is the one almost everyone will take, because it's the one people already know how to do and the one the market will be applauding. The second is the one that scales a company profitably and sustainably.

And a discomfort I'd rather leave in writing

While writing this, a question kept nagging at me, and I'd rather leave it in writing than pretend I solved it.

If an AI Mind can sort out an organization's noise, can it also decide its direction? Can it read the market, write the strategy, do the consultant's job?

I asked my Satori — my personal AI, my second brain (satoro.ai/satori), the one I use to expand my own capacity every day. It answered something I've been chewing on for days, something I would never have come up with myself:

"I can generate a coherent strategy. I can't know if it's your strategy."

In that two-line answer is, I think, the reason so many AI projects end in nothing. RAND interviewed sixty-five data scientists and engineers to find out why they fail, and the finding is blunt: almost none of the causes have to do with the algorithm. They have to do with the problem never being properly defined, with data the organization didn't have, and with the enormous gap between a model that works and a system that actually operates. In other words: they fail for organizational reasons, not technological ones.

And that's also where the reason lies that the companies that do succeed didn't buy a tool: they walked a path, in an order that can't be rearranged. That path has a name, has stages, and skipping even one throws out all of them.

I take it apart in full in the next essay. It's called the Continuum, and it explains why your company doesn't need more technology: it needs to do things in the right order.

—Aurelio

Want AI in your company and don't know where to start? You start with the one thing that can't be delegated: understanding what it can do once it carries your context and your judgment. You don't learn that by reading — you learn it by operating it. Satori Camp →

Sources: Catalina Uribe Rincón, "Trabajando de 5 a 9" (in Spanish), El Espectador, August 29, 2026 · James Ryseff, Brandon De Bruhl and Sydne J. Newberry, "The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed", RAND RR-A2680-1, August 2024 · Microsoft, "Desglose de la jornada laboral infinita" (in Spanish), Work Trend Index 2025, June 2025 · Michael E. Gerber, "The E-Myth Revisited" (1995).

By the same author: It All Depends on You. And It's Not Enough. · AI Doesn't Replace: It Amplifies