Open your calendar and look at Wednesday. Now look at how much of Wednesday is work you already handed to somebody else on Monday.
That is the part of delegation nobody warns you about. The handoff itself usually goes fine. You explain the job, the person understands the job, they go and do the job. Two days later a piece of it is back on your desk, and not because they failed. Something came up that the instructions did not cover, and the only person allowed to say what happens next is you.
Delegation in a small business rarely fails at the moment you give the work away. It fails on the return trip. William Oncken Jr. and Donald L. Wass opened their Harvard Business Review article on this problem with a question that is still the entire diagnosis: “Why is it that managers are typically running out of time while their subordinates are typically running out of work?” (hbr.org). Something keeps traveling back up the line. It is not the task.
What actually comes back is the next decision
Watch the shape of the interruptions for one week and the pattern is hard to miss. The refund request that is four days outside the policy. The customer who wants the Tuesday slot that you normally hold for commercial jobs. The supplier who is out of the part you specified and has offered a substitute. The invoice that is forty dollars short of the quote.
In every one of those, the work got done. What bounced back was a decision, and it bounced because nobody had said in advance who was allowed to make it.
This is why the standard advice does not help. “Give clearer instructions” improves the handling of the expected case, and the expected case was never the problem. Your people can already do the routine ninety percent. The ten percent that does not fit the instructions is the ten percent that lands on you, and it lands on you every single time, which is why the work feels delegated on Monday and undelegated by Wednesday.
The scale of business this matters in is smaller than most management writing assumes. Of the 36,207,130 small businesses in the United States, 6,395,635 have paid employees, and among employer firms the Small Business Administration’s Office of Advocacy reports that family-owned firms employed 15 employees per firm while non-family-owned firms employed 11 employees per firm (advocacy.sba.gov). A business that size has no management layer to absorb an exception. There is the person doing the work, and there is you.
The check-in is the return channel, and you installed it yourself
Here is the part that took me longest to see. Most owners, having been burned once, respond by adding oversight: a daily stand-up, a Friday review, a “just run it past me first.” It feels like the responsible move.
But a check-in with no agreed decision boundary is not oversight. It is a scheduled appointment for the next decision to come back to you. You have not reduced the number of decisions you make. You have batched them and put them in your calendar, and because they now arrive in a meeting rather than as an interruption, it feels like progress.
The Society for Human Resource Management puts the missing half in one sentence: “Give your team members the authority to make decisions related to their tasks” (shrm.org). Almost every delegation conversation covers the task and skips the authority. The task is the easy half to describe, which is probably why it gets all the attention.
So the four moves below are not about explaining work better. They are about deciding, in advance and in writing, who owns the next move when reality does not match the plan.
Move 1: Write the rule, not the steps
Pick one recurring job that keeps coming back to you. Instead of documenting how to do it, write the rule that settles the common case without you.
“Handle the refund requests” is a task, and it will be back on Wednesday. “Refund it without asking me if the order is under sixty days old and under one hundred fifty dollars, and the item is unopened” is a rule, and it will not. The difference is that the second version answers the question the person was going to ask you.
You will not get the rule right from memory, because you have never had to say it out loud. The fastest way to find it is to work backward from the last five times that job interrupted you and ask what you decided each time. Your own past decisions are the draft. If the job is one you have never written down at all, recording yourself doing it once is a faster start than trying to describe it from scratch, which is the case we made in just record yourself.
One rule. Not a manual. A manual is a project you will not finish, and it will not help anyway, because the thing that comes back is never the step, it is the judgment.
Move 2: Put two numbers on the boundary
A rule covers the common case. The boundary covers everything else, and it needs to be numeric, because “use your judgment” is not a boundary and everyone knows it.
Two numbers do most of the work. A dollar ceiling: below this amount, decide it and tell me afterward. A time ceiling: if fixing it costs less than this many hours, fix it. Below both, act. Above either, bring it to me. For a three-person shop, fifty dollars and thirty minutes is a real starting boundary, and it will feel alarmingly low for about a week and then obviously too low.
Set the number by asking what it costs you to be consulted. If a decision is worth eighty dollars and interrupting you costs twenty minutes of your time plus the delay to the customer, consulting you is the expensive option. Owners routinely set these ceilings below the cost of their own involvement, which guarantees the interruption pays for itself in the wrong direction.
Set it higher where the decision is reversible, and keep it tight where it is not. Refunds, scheduling and substitutions can be undone. Firing a supplier, discounting a contract or promising a date cannot be undone cheaply, and those belong to you at any dollar value. Sorting decisions by whether you can walk them back is the same test we used in a decision-making framework for small business owners.
Move 3: Change what the check-in is for
Keep the check-in. Change what happens in it.
A check-in should report decisions already made, not collect decisions waiting to be made. The sentence you want to hear is “here is what came up and here is what I did about it,” not “here is what came up, what do you want to do?” That single change is the difference between a review and a queue.
This also gives you a clean diagnostic. When somebody arrives with an open question they could have closed under the rule and the boundary, that is not a judgment failure on their part. It is a gap in your rule, or a boundary set too low, or a person who has learned from experience that deciding without you gets them criticized. All three are yours to fix, and the third one is the most common and the least comfortable.
Be honest about that last case. If the first few decisions somebody makes under the new boundary get second-guessed, the boundary is dead and every decision comes back permanently. You get one or two chances to react badly before the whole arrangement quietly reverts.
Move 4: Keep an exception log and let it write the next rule
Every time something falls outside the rule, one line goes in a shared document: what happened, what was decided, who decided it. Not a form. One line.
Then read the log once a week and apply a single test: has this exception now appeared twice? If it has, it is not an exception, it is an unwritten rule, and you write it down. This is the mechanism that actually converts supervision into delegation over months rather than in one heroic afternoon, because the rules get built out of real events instead of imagined ones.
The weekly read is also where this system either survives or dies, and a recurring calendar reminder does more for it than discipline does. That is the same finding that came out of the goal-setting research in why weekly reports beat plans, where the group that performed best was the group that had been sent reminders rather than left to remember.
Keep the log where the work already happens. A second system nobody opens is worse than no system, because it produces the feeling of having built something.
Where to keep the rules and the log
The rule, the boundary and the exception log need one shared home that everyone can see. Three real options, with prices read from each vendor’s own pricing page today.
Trello is the simplest thing that works, and for most small teams the free tier is the finished tool rather than a trial: Trello lists its Free plan at “$0USD” and “Free for up to 10 collaborators per Workspace”, with Standard at “$5USD Per user/month if billed annually ($6 billed monthly)” and Premium at “$10USD Per user/month if billed annually ($12.50 billed monthly)” (trello.com). A board with a card per recurring job, the rule written on the card, and a list for exceptions covers everything above.
ClickUp makes sense if you want the log to be searchable and reportable later. It lists a Free Forever plan, Unlimited at “$7 Per user/month, billed yearly” or “$10 Per user/month, billed monthly”, and Business at “$12 Per user/month, billed yearly” or “$19 Per user/month, billed monthly” (clickup.com).
Asana is the option to look at when several people are handing work to each other rather than everything routing through you. Its Personal plan is listed at “$0” and “Free forever”, with Starter at “$10.99” per user per month billed annually or “$13.49” billed monthly, and Advanced at “$24.99” billed annually or “$30.49” monthly (asana.com).
If none of that appeals, a shared document with two headings does the job. The tool is not the intervention here. The written rule is.
The same rule applies when you delegate to software
This is worth saying plainly, because a lot of small businesses are now handing work to an assistant rather than a person. Between September 2024 and August 2025, 7.6 percent of businesses used artificial intelligence, and the second most likely group to have used it was businesses with fewer than five employees at 8.2 percent, behind businesses with more than 250 employees at 11.4 percent (advocacy.sba.gov). The smallest firms are not the laggards here.
The return-trip problem transfers exactly. If a tool drafts your quotes and you then read every line to decide whether the numbers are right, the next move came straight back to you and the only thing you delegated was the typing. That may still be worth it. But test it honestly: after the handoff, who makes the next decision? If the answer is still you, you have bought a faster draft, not a delegated job.
The useful application is one step back from the work. An assistant is genuinely good at turning your last five decisions into a first draft of the rule in Move 1, and at spotting that three lines in your exception log are the same exception wearing different words. Both are tedious pattern-matching jobs on text you already own. If you would rather start from a structured prompt than a blank document, the Delegation Skills Trainer at BusinessPrompter.com is written for this build, filed under Personal Development and summarized as learning to delegate effectively to multiply your impact. It sits behind that site’s Pro membership.
What none of it should be used for is deciding whether the person is ready. That judgment needs context no tool has, and getting it wrong in the cautious direction is how a capable employee learns to stop deciding.
What this does not fix
A written rule does not make a person ready for work they have never done. If somebody has never handled an unhappy customer, a refund boundary hands them a hard conversation with a number attached to it, which is not the same as training. Sit in on the first two.
It also does not help with the genuinely one-off decision, and you should not try to make it. The point of writing rules for the recurring ninety percent is to buy back the attention the exceptional ten percent deserves. Sorting what actually deserves that attention is a separate problem, and one we took apart in a priority rule for SMB owners.
And it is worth being clear about what a boundary is for. Giving somebody a spending limit and a documented rule is how a person becomes able to run a part of the business without you, which is the thing that makes them promotable and makes the business worth something when you are not in it. Used the other way, as a way to extract more output from the same people while removing the interruptions that told you they were drowning, the exception log stops being a training record and becomes a surveillance log. The log is meant to be read for patterns in the work, not for patterns in the person.
Frequently Asked Questions
What should I delegate first?
The job that has interrupted you most often in the past two weeks, not the job that takes the most time. Frequency is what makes a rule worth writing, because a rule pays back every time the situation recurs. A four-hour task that happens twice a year is worth less to systematize than a five-minute decision that reaches you daily.
How do I set the dollar limit without exposing the business?
Start from what being consulted actually costs, then check the downside. If a wrong decision at that ceiling would be annoying rather than damaging, the ceiling is probably still too low. Keep it tight on anything you cannot reverse, such as discounts, promised dates and supplier changes, and let it run higher on things you can undo, such as refunds and scheduling.
My team keeps asking anyway. What am I doing wrong?
Usually one of three things. The rule does not actually cover the case they keep hitting, so check the exception log before blaming the habit. The boundary is set below the decisions they face, so it never applies. Or an early decision made under the boundary got second-guessed, and they have concluded that asking is safer than deciding. The third is the most common and it is fixed by backing a decision you would have made differently, at least the first few times.
Does this work if I have no employees at all?
The rule and the boundary still apply, with the contractor or the tool as the other party, and they matter more rather than less, because a contractor cannot absorb ambiguity by reading the room. Write the rule and the ceiling into the brief. The exception log is what tells you, a few months in, whether you are buying a service or supervising one.
Start with the one that interrupted you yesterday
Do not build a system. Take the single job that pulled you off something else yesterday, write one sentence that decides its common case, attach a dollar figure and a time figure, and tell one person that below those numbers they do not need to ask.
Then watch what comes back, because that is your real documentation. Everything that returns in the next two weeks is either a gap in the rule or a signal that the person does not yet believe the boundary is real, and those two problems look identical from your desk while needing opposite responses.
