Most delegation in small companies happens in one of two ways, and neither works.
The first is the handoff. You explain the task once, usually while walking somewhere, and hope. What comes back is either wrong or late, you fix it yourself, and you file the whole experience under "faster to do it myself."
The second is the permanent apprenticeship. You never quite let go. You review everything, you're copied on everything, and two years later the person still can't run the thing without you in the room. From the outside this looks like careful management. From the inside it's the same bottleneck with more meetings.
What sits between them is a method. This is the one we use, and it has three stages.
Before you start
Two questions, both quick.
Is this task worth handing over? Handing over something you do twice a year is rarely worth the effort of teaching it. The tasks that repay the investment are the recurring ones, and the ones where you are the only person who can do it.
What does "done" look like? Not the steps, the outcome. If you can't say what finished work looks like, you can't teach it and you certainly can't judge it later. This is where most handovers are lost before they start, and it's covered properly in why the work keeps coming back to you.
Stage one: shadowing
They watch you do the task. They take notes. They ask questions while you work.
The instruction that matters here is for you, not them: narrate your thinking, not your actions. What they can see is the steps, and the steps are the easy part. What they cannot see is why you skipped the second option, why you phrased the email that way, why you called instead of writing. That judgment is the actual content of the task, and it stays invisible unless you say it out loud, which feels unnatural the first few times.
Make them keep the notes. Two useful things happen. You find out what they think is important, which is often not what you think is important, and the notes become the process document you were never going to write.
One round is rarely enough for anything with variation in it. Have them shadow you two or three times if the task changes shape depending on circumstances, because the whole point is for them to see the circumstances change.
Stage two: collaborating
You do the task together, and you course correct as it happens.
The trap in this stage is that it quietly turns back into stage one. You have the keyboard, you're doing the work, they're watching and nodding, and you both feel productive. That isn't collaboration, it's a second shadow.
Give them the controls. They write the draft while you sit beside them. They speak on the call while you're on mute. You interrupt, correct, take over when it genuinely matters, then hand it back. Correction in the moment is the cheapest feedback that exists, because they get it while they still remember what they were thinking.
Expect this stage to feel slow and slightly irritating. That's the cost, and it's paid once.
Stage three: reverse shadowing
They do the task. You watch. You say nothing.
The discipline here is entirely yours. Watching someone do something worse than you would, in real time, without intervening, is the hardest part of the whole method. Intervene and you're back in stage two, and worse, you've told them in front of whoever else is present that they're not really running it.
Save it for after. Then be specific. "You lost the room when you went into the numbers before the context" is feedback. "That went well, maybe tighten it a bit" is not.
Repeat this stage as many times as it takes.
Knowing when to stop
The test is simple: you stop when you have no feedback left to give.
Watch them do it and come away with nothing to say. Then see it happen again. Two clean runs where they got the task done successfully without you stepping in, and where you genuinely couldn't improve on it, and they're done. That's the exit.
Two things count as stronger evidence than a clean run, and both are worth waiting for on anything important.
The first is a judgment call. They hit a fork you never discussed and chose the way you would have chosen, for reasons they can explain. That tells you they've absorbed the thinking rather than the checklist.
The second is a situation you never covered. Something goes sideways in a way that wasn't in any of your three demonstrations, and they handle it. You can teach steps. You can't teach that, which is why seeing it is such good news.
Notice what the test is not. It isn't a length of time, and it isn't a number of weeks. It's repetitions, so a weekly task moves through the stages in a couple of months and a quarterly one takes a year. Rushing the calendar is how people end up back at the handoff.
What to run this on
The method works on any task, but if you're building a bench rather than just clearing your desk, run it on work where someone has to get an outcome through other people. Five that work well in a small company:
Chairing the weekly meeting. Shadowing is them watching you run it with your reasoning narrated afterward. Collaborating is them running half the agenda. Reverse shadowing is them running it while you sit there like any other attendee, which is harder for you than for them.
Onboarding the next new hire, start to finish. This one is unusually good because it has a natural end point and a visible result: either the new person is productive or they aren't.
Running one client relationship end to end. Start with a smaller account. The staged version of this is how owner-operators eventually hand over the relationships they think they can never hand over.
Presenting the numbers. Not sending a report. Standing up, explaining what happened, and taking questions they can't fully anticipate.
Handling a difficult conversation. Shadow one with you leading, do one together where they open and you close, then one they run with you in the room. This is the rep that most reveals whether someone will be able to manage, and the one owners most often skip.
Where this comes from
None of this is new. Teachers have used "I do, we do, you do" for decades, and medical training has its own version in see one, do one, teach one. We didn't invent the shape.
What's rare is applying it to the things an owner does. Most business advice on delegation stops at "you should delegate more," which is true and useless. The three stages turn it into something you can start on a specific task, on a specific day, with a specific person.
Why this is bench-building, not just delegation
Handing over a task buys back your time, which is reason enough. But run this on the right work and it does something else: it tells you who can lead before you have to bet a role on it.
You will learn more about whether someone can manage from watching them chair four meetings than from any interview, appraisal, or gut feeling. And if the answer turns out to be no, you've found out at the cost of some slightly worse meetings rather than at the cost of a promotion, a team, and a year.
That's the whole argument for the leadership bench. This is the mechanism.
This is one spoke of the leadership bench, our model for building a leadership pipeline in a small company. See also why the work keeps coming back to you and what a first-time manager has to unlearn.