Most advice on this subject is about the word. Practise saying no. Say it without apologising. The word is not what fails. A refusal comes apart when it arrives carrying no information, because the person hearing it then has to work out why, and the cheapest available explanation is that you did not feel like it.
Everything that follows is one idea applied five ways: replace the refusal with a constraint the other person can check, and hand the decision back to them.
- A bare no makes people invent a reason, and the invented one is never flattering.
- Say what you are already committed to, with dates, not that you are busy.
- Hand back a decision instead of making it for them.
- Never renegotiate scope and deadline in the same sentence.
- The hardest no is to work you would genuinely enjoy.
Why a refusal lands badly
When you decline something there are three readings available: you cannot, you will not, or you do not much care. Only the first is neutral. If you give no information, the listener picks between the other two, and which one they pick depends on their mood rather than on you.
Information changes the category. A constraint they can verify is a fact about the world. An unexplained refusal is a fact about you. People forgive the first and remember the second, which is why the same answer can cost nothing on Tuesday and cost you a relationship on Thursday.
This also explains why "I am swamped" performs so badly. It is unfalsifiable, everyone says it, and it happens to be true of the person asking as well. It sounds like information and functions as noise.
The shape that works
Three parts, in order, and the order matters.
Show you understood the ask. One sentence, specific enough to prove you were listening. Not "sure, I hear you" but a restatement of the actual thing, including the part they are most worried about.
State the constraint in units they can check. Named commitments, real dates. "I have the migration through Thursday and the board pack after it" is checkable. "My plate is full" is not.
Hand back a decision with two real options. Not a rhetorical choice where one arm is obviously unacceptable. Two things you would genuinely do, and let them pick. This is the move that converts a no into a piece of collaboration, and it is the part people skip.
The reason it works is not psychological trickery. It is that you have given them something to do. A refusal ends a conversation and leaves them stuck. A choice continues it and leaves them in control, which is what they wanted from the interaction in the first place.
Five conversations
Your manager
The goal is a trade, not a refusal, and the trade should be theirs to make.
"Happy to take the reporting piece. Right now I have the migration finishing Thursday and the supplier review due the following Monday. I can start the reporting on Friday and have it with you mid week after, or I can start it tomorrow if the supplier review moves to the end of the month. Which is more useful?"
Nothing there is a no, and the outcome is frequently that one of the two things quietly turns out not to have been urgent.
A client
Here the answer is usually yes, with a price attached, and the discipline is to keep the price and the timeline as separate sentences.
"That is a good addition and it is outside what we scoped. It is roughly two days, which puts delivery at the eighteenth rather than the eleventh, and it is X on top. If the eleventh is fixed, we can ship as scoped and do this straight after. Tell me which matters more and I will send the revised note."
The mistake to avoid is absorbing it "just this once" without saying so. An unpriced favour becomes the new baseline immediately, and the next request will be larger, which is not the client being difficult. It is you having told them this is free.
A colleague
No authority sits behind this one, which makes it easier to decline and easier to damage.
"I cannot take this week, I am on the migration until Thursday. Two things that might help faster than me: the template in the shared drive covers most of this, and Priya did the same thing for the Berlin account last month. If it is still stuck on Friday, come back to me."
Declining while still being useful costs you a minute and buys most of the goodwill that saying yes would have.
A recruiter
Be brief and be clear, because vagueness here generates six more emails.
"Not looking at the moment, thanks. If something comes up that is fully remote and senior, that is the shape worth sending. Otherwise I would rather not clutter your pipeline."
Saying what would interest you is what makes this a no that keeps the door open rather than a no that ends the relationship.
A friend or family member
The one where capacity is usually not the real reason, and pretending it is will be detected.
"I am not going to be the right person for this, and I would rather say that now than do it slowly and badly for three months. If you want, I will sit down for an hour and help you work out what to look for in someone who does this properly."
Honesty about fit is kinder than a scheduling excuse, because a scheduling excuse invites them to come back when you are less busy, and you will have to do this again.
The one we got wrong
Someone asked for a quick look at a spreadsheet that was behaving oddly. We said yes, because it was going to take twenty minutes and it did.
The mistake was not saying yes. It was that nothing in the yes had an edge to it. No end, no scope, no statement of what we were and were not taking on. Which meant that when the same file came back the following week, and the week after, there was no moment where anything changed. We had not become responsible for it in a single decision. We had become responsible for it by never once saying where the twenty minutes stopped.
What should have gone with the yes was one sentence: happy to look this once and tell you what is wrong, but I am not the right person to own it, so let us work out where it should live. That sentence costs nothing on day one and is nearly impossible to say on day forty.
The general lesson is that a yes without a boundary is not a smaller commitment than a yes with one. It is a larger commitment with the size hidden.
When to say yes to work you do not want
The framing that everything unwanted should be declined is wrong, and following it makes people worse at their jobs.
Say yes when the work is genuinely rare and would teach you something you cannot get another way. Say yes when refusing would put a real cost on someone who has carried you before, because that ledger is real even though nobody writes it down. Say yes when the request is small and the relationship is new, since early goodwill buys the ability to decline later.
Say no when the reason you are being asked is that you are reliably the fastest yes. That is not a compliment about your skills. It is a description of a routing problem, and every time you accept you reinforce it.
Saying no to yourself
The same shape works on your own ideas, and it is harder, because there is nobody to hand the decision back to.
The move is identical: name the trade out loud. Not "I would like to build this too" but "building this means the other thing does not ship this quarter". Written like that, most ideas answer themselves. The ones that survive were worth the swap, and now you know what you paid.
If you find yourself repeatedly saying yes to your own new ideas at the cost of finishing anything, the problem is usually upstream in how you pick work at all. Our piece on choosing which tasks are worth the effort is about the same decision in a different setting.
If you change one thing this week
Delete "I am swamped" from your vocabulary and replace it with a named commitment and a date. That single substitution does most of the work in this article, because it converts a statement about your feelings into a statement about the world, and the world is something the two of you can negotiate with together.
Frequently asked questions
What if my manager says everything is a priority?
Then the ranking is being made anyway, just by you and silently, which is the worst place for it to happen. Put the list in front of them with dates attached and ask which one slips. If the answer is still everything, you have learned something important about the job rather than about the request.
Should I explain my reasons in detail?
One line of constraint, not a paragraph of justification. Long explanations read as guilt and invite negotiation on each point. Name the commitment and the date, then stop talking, which is the part people find hardest.
How do I say no to someone senior to my own manager?
The same shape, with the routing added. Say yes to the priority and ask whose work it displaces, then bring your manager into the decision rather than resolving it alone. Quietly absorbing work from above is how people end up accountable to two queues with capacity for one.
What if I already said yes and now cannot deliver?
Raise it the moment you know, not the day before. An early warning with a revised date is a manageable problem. A missed deadline discovered on the deadline is a different and much larger one, and the difference is almost entirely in the timing.
Does this work in writing or does it need a conversation?
Writing is usually better for the constraint, because the dates are on record and nobody has to remember them. A conversation is better when the relationship is strained or the no is a real disappointment, since tone carries things text cannot. When in doubt, talk and then confirm in writing.
How do I stop being the person everyone asks?
You are usually asked because you are the fastest yes, not because you are the only option. Slowing the yes down is most of the fix. Answer requests once a day rather than instantly, and route repeat asks to the place they should have gone, which teaches the pattern without any confrontation.
How this was put together: the scripts here are patterns we use and have watched work, adapted so nothing identifies a specific person or client. The failure described above is our own. We have not quoted figures on how often these approaches succeed, because we have not measured it and any number would be decoration rather than evidence.