You take pride in one thing above all: your code is cleaner than anything an AI produces. Build the same feature and the draft the AI spits out always has something to fix, while your own comes out tidy on the first pass. So today, once again, you spend two hours hand-wiring the payment integration yourself. Handing it to the AI would only mean more time spent reviewing it—and more than that, it feels wasteful to give away the very thing you're best at. Two hours later the code is clean and connected, but the first customer interview you'd planned for the day slips to next week, again. You spent one more hour on what you do well, and one less hour on what decides whether the business lives.
This scene holds the trap solo founders fall into most often: clinging to the work you're good at and handing off only the work you're bad at. It feels intuitive, but draw the line this way and you end up spending your most valuable hours on your cheapest tasks. The test for what to delegate needs rebuilding.
Delegating What You're Bad At Is Only Half Right
The first answer that comes to mind when you think about delegation is 'hand off what I'm bad at.' The founder who came up through marketing offloads development and design; the one who came up as a developer offloads copywriting and bookkeeping. As a way of covering your weak spots, it isn't wrong. It's just only half right. This test leads straight to the conclusion that of course you should do the things you're good at yourself—and that conclusion is what produced the two hours in the scene above.
The other half is filled by a 200-year-old trade theory. It's exactly what David Ricardo (D. Ricardo) set out in 1817 with the principle of comparative advantage. The point isn't absolute skill but opportunity cost. Even someone who's better at both tasks can't do both at once. Hold on to one and you have to give up the other. So the question isn't whether you're better at it, but what you give up while you're holding it. The principle was first used to explain trade between nations; swap the trading partner from a country to an AI and it applies unchanged.
The Better You Are at It, the More It Costs to Hold On
Return to the developer whose code is better than the AI's. In absolute skill, the human is ahead. Hand the coding to the AI and quality drops a little, and closing that gap with review takes time too. By the test of 'delegate what you're bad at,' coding is obviously a do-it-yourself job. But comparative advantage looks at the next column over. The hour he spends coding is also an hour he could have spent on product planning or a customer interview. If the value created by one hour of planning exceeds the difference in code quality, then handing the coding to the AI and keeping only the final review for himself grows the output of the whole business.
Here the test flips. The question to ask isn't 'Is the AI better than me?' but 'What am I giving up while I hold on to this?' The better you are at something, the more you want to keep it—but an hour spent on what you do well is expensive precisely because the work you forgo in that hour is valuable too. The true price of the two hours spent writing clean code isn't the outsourcing fee or the tool subscription; it's the first customer you never met in that time. The more a task is one you cling to because you're good at it and it feels wasteful to give up, the more often the value you forgo by keeping it turns out to be large.
Labor economics has already drawn the grain of what comparative advantage calls the 'work you'll miss least.' David Autor (D. Autor) and colleagues found that technology replaces routine tasks—work whose steps can be written down as rules—while it actually complements non-routine cognitive tasks such as judgment, creativity, and reading context. The work that's good to hand to an AI is the work whose procedure can be documented; the work to keep on your side is the work that takes judgment. And routine tasks you can write down as steps tend to be the kind where the result is much the same no matter who does them—in other words, the work you'll miss least. The line comparative advantage draws by opportunity cost and the line labor economics draws by the nature of the task land in the same place.
More evidence comes from a field experiment. In a study by Erik Brynjolfsson (E. Brynjolfsson) and colleagues observing 5,179 call-center agents, introducing generative AI produced an average 14% gain in productivity, and the effect was largest—34%—among newer and lower-skilled workers. Read from the solo founder's side, the implication is clear: the payoff from delegating to an AI is largest in the areas where you're weakest. If you came up through marketing, you stand to gain more by putting AI on development, design, and bookkeeping—the areas where you're the rookie—than on the marketing you already know. The 'work you'll miss least' from comparative advantage and the 'work you're bad at' from the field experiment point the same way. That a 200-year-old trade theory and recent call-center data, starting from such different places, arrive at the same conclusion is something you can't see by looking at only one of them.
This division of labor changes the kind of work more than it reduces the total amount. As Daron Acemoglu (D. Acemoglu) and Pascual Restrepo (P. Restrepo) have laid out, automation shrinks the human share of existing tasks while creating new ones. Hand off drafting, and into the empty space come new jobs: setting the standards for review, refining the instructions, and curating—choosing which of several outputs to keep. The work that shrinks is the work whose steps can be written down; the work newly created is the work that takes judgment. Where you hand off the work you'll miss least, the more valuable work of judgment moves in.
Three Ways to Redraw the Line with Comparative Advantage
First, next to each task write one line: 'what I can't do while I do this.' Lay out the past two weeks of tasks and, beside each one, note the work that got pushed aside while you did it yourself. Next to two hours of code you'll write 'customer interview'; next to squaring the books, 'planning the next product.' Any item where the work you gave up is worth more than the task itself is a candidate to hand off—even if it's work you're good at. It's an exercise in reproducing the comparative-advantage calculation right there on paper.
Second, flag the tasks you're keeping because you're good at them. The tasks you handed off because you're bad at them are already gone. The problem is the tasks you didn't hand off because you're good at them. On your list, mark the items tagged 'I do this myself because I'm good at it,' and pick out the ones whose steps can be documented. If a task is something you're good at and can be written down at the same time, being good at it is no longer a reason to keep it. That task is your first candidate for delegation.
Third, put AI on your weakest areas first. If you're stuck on where to start delegating, look first at the areas where you're the rookie. As the call-center experiment confirmed, that's where the payoff is largest. One caveat: work you've never done yourself is hard to review even after you hand it off, so do it yourself once—just enough to develop an eye for telling a good result from a merely plausible one—before handing it over. The first delegation in a weak area starts not with eliminating the work, but with getting yourself to a place where you can review it.
From Productivity to Profitability
Redrawing the line with comparative advantage doesn't end at simply saving time. The point is to pour the hours you free up by handing off the work you'll miss least into the work where judgment compounds and customers multiply. Two hours that raise code quality by 0.1 and two hours that uncover a first customer's real problem are not the same two hours. The first ends with a single output; the second sets the direction of every output that follows. This trade—less of the work you're good at, more of the work that's valuable—is the first button to fasten in connecting faster hands to revenue. The moment you change the test for delegation from 'work I'm bad at' to 'work I'll miss least,' the division of labor crosses over from a tool of productivity to a tool of profitability.
This series unpacks a single manuscript, one piece at a time. It starts from Running a Company by Yourself (『혼자서 회사를 경영한다는 것』), a book that re-binds the standard theories of accounting, economics, management, and investing—across disciplinary lines—into the problems of the one-person business. The next piece takes up how the table of tasks you've handed off this way comes to serve as the org chart of a company with no employees, and why that boundary is a moving line that has to be redrawn every quarter. If you've ever felt your business slipping precisely because you were holding on to the work you do well, try this: on a single sheet of paper, start by writing, next to each task, the work you gave up for it.
Concept Appendix
- The principle of comparative advantage — Proposed by David Ricardo (D. Ricardo, 1817). The idea that, regardless of absolute skill, if each side specializes where its opportunity cost is lower and then trades, both sides come out ahead. Swap the trading partner from a country to an AI, and it applies directly to how you split tasks between yourself and the AI. - The routine / non-routine task distinction — Set out in labor economics by David Autor (D. Autor), Frank Levy (F. Levy), and Richard Murnane (R. Murnane). Routine tasks, whose steps can be written down as rules, are replaced by technology, while non-routine cognitive tasks that require judgment and context are complemented by it. It becomes the grain along which you decide what to hand off and what to keep. - The task-based model of automation — Proposed by Daron Acemoglu (D. Acemoglu) and Pascual Restrepo (P. Restrepo). Automation carries a displacement effect that reduces the human share of existing tasks while also creating new ones, explaining that delegation changes the kind of work more than the total amount of it.
About the series — Insights based on the manuscript of Running a Company by Yourself: A Management Framework for the Solo Founder in the Age of AI (『혼자서 회사를 경영한다는 것 — AI 시대 1인 창업자를 위한 경영 프레임워크』). Each piece takes up a single management decision.
