“I disagree and commit and hope it becomes the most watched thing we’ve ever made.” Jeff Bezos sent that reply to an Amazon Studios team that wanted to greenlight a show he had doubts about. He told the story in a letter to Amazon shareholders, and the line is unusual because it does two things at once. He says plainly that he still disagrees, and in the same sentence he wishes the project well.
Most people who lose an argument at work don’t do either of those things, and instead they go quiet or say “fine” in a tone that means the opposite. Some of them tell a colleague later that it was never going to work. The disagreement doesn’t go away but just moves somewhere the person who made the call can’t see it.
Bezos also had an easy version of the problem, because he was the boss and committing cost him very little. This post is for the person on the other side of a decision like that, who argued, lost and now has to help the plan work. It explains what the idea actually means and what has to be true before anyone can commit, and then it walks through what committing looks like over the first three months. It ends with the cases where you shouldn’t commit at all.
What does “disagree and commit” actually mean, and where does it come from?
Disagree and commit means you argue your case as well as you can while a decision is still open. Once the decision is made, you act on it in good faith even if you still think it’s wrong.
The best-known version is Amazon’s. Its leadership principle, Have Backbone; Disagree and Commit, says leaders “are obligated to respectfully challenge decisions when they disagree.” It ends with a short sentence: “Once a decision is determined, they commit wholly.” Bezos wrote in his letter that the phrase “will save a lot of time.” He also said it isn’t only for junior people. Bosses should use it too, and the Amazon Studios reply was his own example.
The idea is older than Amazon, but its history is less certain than many articles suggest. Wikipedia’s entry on disagree and commit says Intel practiced it under Andy Grove. Plenty of posts say Grove coined the phrase, but nobody has shown that he did. Wikipedia also says Scott McNealy used the phrase at Sun Microsystems some time between 1983 and 1991. His line was blunter: “Agree and commit, disagree and commit, or get out of the way.”
What the phrase doesn’t mean matters more than where it came from. You don’t have to change your mind. Dave Kellogg, a software executive, admits on his blog that he first read “commit” as “change your mind.” He came to see it as carrying out the decision “as if it were the one you had voted for.” So committing means two things other people can see. You do the work the decision needs, and you don’t undermine it. What you privately believe about the decision is still yours.
One more boundary before going on. This post is about decisions. If what’s bothering you is how your boss behaves, that’s a different conversation, and the guide to giving feedback to your boss covers it.
What has to be true before you can commit?
Most advice treats commitment as something you either have or you don’t. In practice it’s much easier to give when three things are true. You were heard before the decision was made, and you were told why it went the other way. You also know what happens next, including when the decision will be looked at again.
That’s a way of thinking about it, built from the work described below. It isn’t a tested checklist. But each condition points at something you can ask for, and that matters when you’re the one who lost. Strictly speaking, the conditions are the decider’s job because they owe you a reason and a review point. But they don’t always offer them, so you may have to ask.
The first condition is being heard before the decision. A 1990 study by E. Allan Lind, Ruth Kanfer and P. Christopher Earley tested this in a lab. The 179 undergraduates in their study of voice and fairness were each given a goal for a task. Some could give their view before the goal was set. Others could only give it afterward, and some weren’t asked at all. The abstract reports that “Both pre- and postdecision voice led to higher fairness judgments than no voice.” Being heard before the decision counted for more than being heard after it. The study measured how fair students judged the process to be, not commitment at work, but it still matches what most people feel when a decision is announced before anyone asked them. If you weren’t heard in time, you can still ask for your concern to be written down alongside the decision. A short decision log does this well, because it puts your view on record for the review without reopening anything.
The second condition is being told why. W. Chan Kim and Renée Mauborgne wrote about this in their Harvard Business Review article on fair process, published in January 2003. Their idea is that people cooperate more willingly with a decision when they trust the process that produced it. Explanation is one of their three principles. Blue Ocean Strategy’s summary of fair process says a clear explanation helps people trust that a decision was impartial, “even if their own ideas have been rejected.” If nobody told you why, ask. Frame it as a request for information you need to do the work. You want to know what tipped the decision so you can make the plan succeed. That framing matters because “Why did you decide that?” can sound like the start of round two, while “What tipped it?” asks for something useful instead.
The third condition is knowing what happens next. Kim and Mauborgne’s third principle is that people should know what’s expected of them once a decision is made. The person who lost needs one more thing, which is to know when the decision will be checked. Ask for a date or a signal, because “When will we look at how this is going?” is enough, and so is “What would make us rethink it?” Asking gives your concern somewhere to go, so you don’t have to carry it around with you.
What does committing look like on day one, in week two and at month three?
Commitment shows in what other people can see you do, because your private view of the decision is yours and nobody can check it anyway. What they can see is how you talk about the plan and how hard you work on it. They can also see what you do when something goes wrong. Those behaviors change over the first few months because the pressure on you changes. On day one, people are watching how you take the loss. The first evidence is coming in by week two, and there’s usually a result to look at by month three.
Day one: what you do and say in the first hours
There are usually three audiences on day one, and each one needs something different from you.
Start with the person who made the call. They need a clear signal that you’re going to act on the decision, and they need it soon. A short message the same day works better than a long silence, because silence after a lost argument reads as sulking, even when you’re only busy. If there’s a risk you want on record, mention it once as something to watch rather than a reason to reconsider. The guide to pushing back on your manager has a short recap message for exactly this moment. What doesn’t belong in that message is the argument itself. If you find yourself writing “I still think,” save it for the review.
Your peers are the second audience. Some of them were on your side, and others will just ask how it went. The temptation is to recruit them. “Can you believe they went with that?” feels like harmless venting. But it builds a small group that’s quietly waiting for the plan to fail, and you’ll have started it. The opposite mistake is to perform agreement you don’t feel. If you tell everyone it’s a great call, you’ll have to take that back later, and people remember. The honest version is simpler. You argued for something else, the decision went the other way and you’re working on it. That’s true, which means it gives nobody anything to gather around.
The hardest moment often comes with someone more senior, such as your manager’s manager. They ask you directly whether you agreed with the decision. It can feel like there are only two answers. You can say yes, which isn’t true. Or you can say no and seem to be undermining your boss. But there’s a third answer. You can say that you argued for a different approach and that you’re now behind the plan and doing your part. Both halves are true, so you aren’t lying. You also aren’t asking anyone to overturn the decision, which means you aren’t working against your boss. A senior leader who asks that question is usually checking two things. They want to know whether the team heard the other side, and whether everyone is now pulling in the same direction. Your answer tells them yes to both.
If you manage people, day one has a fourth audience, because your team will ask about the decision and notice whether you sound convinced. The guide to change management at work covers how to pass on a decision you didn’t agree with.
Week two: the first sign that your concern was right
This example is an illustrative composite rather than a real person. Megan is a product manager at a software company. Her team had to decide how to launch a new billing page. She argued for releasing it to a small group of customers first. The page hadn’t been tested with customers who pay by invoice, and those are the company’s largest accounts. Graham was her director, and he chose to release it to everyone at once. “We promised it to sales for this quarter,” he said, “and a slow rollout breaks that promise.” He listened to her and explained his reason. He also set a review for the end of the month.
So all three conditions were met, and Megan committed. She wrote the launch notes and answered the first support questions herself. She didn’t mention her worry to anyone outside the team.
Three support tickets came in from invoice customers during the second week. Each one said the new page showed the wrong due date. Three tickets isn’t a crisis for a company with thousands of customers, but it was exactly the risk she had raised.
She had two easy options. The first was to say nothing, because the tickets were small and raising them might look like gloating. But staying quiet would let a small problem grow while the one person who’d seen it coming just watched. That isn’t commitment either, because a committed person wants the plan to work. The second option was to forward the tickets to Graham with a note like “Looks like the invoice issue.” Every word of that would be true, but the timing and the wording turn information into a point scored. Mark Schwartz put the rule plainly on the AWS enterprise strategy blog: “You can’t be passive-aggressive or later say ‘I told you so.’”
What Megan did instead was report the tickets as news about the plan, along with what she was doing about it. She sent Graham the three tickets and pointed out that they all came from invoice customers. She said she’d asked engineering how the due date was worked out and would know by Thursday. She also asked whether he wanted the invoice accounts watched more closely before the review. She didn’t mention that she’d predicted it.
There’s a simple test for whether a message like that is information or a told-you-so. Ask whether you’d send the same message if someone else had raised the risk. If you would, it’s information. If it only makes sense because you were right, it’s a point scored, and the other person will hear it that way.
Graham’s reply was mixed. He thanked her for catching it and asked engineering to fix the date. But he didn’t slow the launch, and he said three tickets didn’t change the plan. He also asked a little sharply whether this was the issue she’d raised in the meeting. She said it was the same area and left it there. The fix shipped the following week. Whether invoice customers needed more than a fix was still an open question when the review came around.
Month three: when the risk happened, when it didn’t and when you can’t tell
By month three there’s usually some kind of result. What a committed person does depends on which result it is.
| What happened | What you do | Why it matters |
|---|---|---|
| The risk you raised happened | Bring what you saw to the review as facts. Don’t treat the outcome as a win for you. | The review needs facts. A told-you-so makes the decider defend the decision instead of fixing it. |
| The risk didn’t happen | Say briefly that you were wrong about it. | It costs little and makes your next disagreement easier to believe. |
| You can’t tell yet | Ask for the review that was promised. | Without one, the question stays open and the doubt spreads. |
If the risk didn’t happen, then you were wrong and that’s fine. One sentence to the person who made the call is enough. It shows them you’ll admit when you were wrong, so they’ll take your next disagreement more seriously.
If the risk did happen, the hard part is not treating it as your trophy. You were right about the risk, but the outcome belongs to the whole team, and you were part of the team that delivered it. So put what you saw into the review and let the facts do the work.
The third row is the hardest to handle, because a mixed result like Megan’s invoice tickets can be read either way. That’s where the review date you asked for on day one pays off. A review lets everyone look at the same evidence at once, and the next section covers when it’s fair to raise something new there.
How do you tell commitment from compliance or quiet resistance?
Commitment, compliance and quiet resistance can all look alike in the first week. They separate over time, and a few everyday behaviors show the difference.
| Behavior | Commitment | Compliance | Quiet resistance |
|---|---|---|---|
| When something goes wrong | Report it with what you’re doing about it | Wait to be asked | Let it grow, or point at it in public |
| What you say to peers | You argued the other way and you’re working on it | Very little | Hints that it was never going to work |
| Effort when nobody’s checking | Full effort on the parts that matter | Enough to avoid blame | Slow, or spent on something else |
| The risk you flagged | Watch it and share what you see | Forget about it | Keep a private file of evidence |
| When someone asks your view | Honest about the argument and the commitment | ”It’s fine” | Depends on who’s asking |
This table is a way to check yourself. It isn’t a validated test. The middle column is the one most people drift into without noticing, because from a distance compliance looks a lot like commitment. Bezos draws the same line in his letter. He says his Amazon Studios example wasn’t him thinking “well, these guys are wrong and missing the point, but this isn’t worth me chasing.” That thought is compliance, and he counts it as something different from commitment.
So can you be on board without really being on board? You can be fully behind the action while still thinking privately that it was the wrong call. What you can’t do is act as if the plan doesn’t apply to you, or as if its failure would be good news.
Can you ever bring it up again?
Yes, and there are two legitimate ways. The first is new information. Kellogg’s test for reopening a settled decision is a single question. He puts it like this: “if you want to re-open discussion on something we’ve already decided, do you have any new information that wasn’t available at the time we made the decision?” If the answer is yes, the decision may be worth looking at again, either now or later. If it’s no, the topic waits. A better argument for your original view doesn’t count as new information. Megan’s tickets were new, but her concern about invoice customers wasn’t, because Graham had already weighed it.
The second way is the scheduled review, and this is where the third condition pays off. If you know the decision will be checked on the 30th, you don’t have to fight for a hearing. You turn up with what you’ve seen. If there’s no date, every attempt to raise the subject looks like reopening the argument. You’re left choosing between silence and nagging, and neither helps the plan. If the review happens in a document or by email, the guide to disagreeing in writing covers how to make your case land there.
What none of the sources answers is how many times you can lose before it stops being fair, so this part is only our view. If the same person’s ideas lose every time, the problem probably isn’t their commitment. It may be the way decisions get made, or the relationship with the person making them, and the guide to managing up is a better place to start on that. But we can’t tell you where the line is. Three losses in a row might be bad luck, while ten might be a pattern or a team that’s simply right more often than you are.
When should you not commit?
Everything so far assumes a decision you can live with, made through a process that mostly worked. Sometimes neither is true. Victor Hoang’s guide to disagree and commit on the Rework blog lists several cases where the principle breaks down. It’s a practitioner’s view rather than research, and the first three cases below follow it.
-
When going along would cross an ethical, legal or safety line. Disagree and commit is for judgment calls, where reasonable people can see it differently. It isn’t for decisions you believe are unlawful or unsafe. In that case the right route is your company’s formal channels. That might be compliance, legal or HR, and in some situations a regulator. Rules differ by country and by company, so check your own policies and local rules. This post isn’t legal advice.
-
When the decision can’t be undone and you know something the decider doesn’t. Some decisions are easy to reverse and some aren’t, and the guide to reversible decisions explains how to tell them apart. If it’s a one-way decision and you hold a fact nobody weighed, committing quietly buries it. Get the fact to the person deciding before it’s too late. If they still decide the same way with it in front of them, you’re back to an ordinary judgment call.
-
When the three conditions were never met, and you were refused when you asked. Go back to the three conditions near the top of this post. If you asked for a reason and didn’t get one, or asked when the decision would be reviewed and were told never, you haven’t really been heard. Hoang puts the underlying problem simply: “If voicing an objection costs you something, you won’t do it.” You may still do the work, and often you’ll have to. But be honest with yourself that it’s compliance. That’s a sign about the team’s psychological safety rather than a failure of your character.
-
When the phrase is used to end a conversation that never happened. “Let’s just disagree and commit” can be said before anyone has made their case. In that case the phrase is just an instruction to stop talking. Kellogg describes the reverse failure, where someone stays quiet in the meeting and then tries “a pocket veto” in private afterward. Both come from the same gap because the disagreement never happened in the open. If someone uses the phrase to cut you off, ask for a few minutes to make your case first. Say you’ll commit once you’ve had them, and then do.
None of these tells you when it’s time to leave. One overruled decision is normal working life. A long run of them, with the conditions refused each time, is something else. But no rule says whether that takes five decisions or fifteen, so only you can weigh that against everything else the job gives you.
How do you answer “tell me about a time you disagreed and committed”?
Interviewers ask this because it tests three things at once. They want to hear that you made a real case, with reasons, while the decision was still open. They want to hear that you accepted the outcome and acted on it. They also want to know what happened to your concern afterward, because that shows whether you kept paying attention.
A good answer follows the order things happened. Describe the situation and what you argued for, with your reasons. Say what was decided and what you did next, and finish with what happened and what you took from it.
Pick a real example. If you invent one, a good interviewer will find the gaps with two follow-up questions. Be wary of the story where you were proven right and your boss admitted it. It sounds like a strong answer, but it tends to come across as a told-you-so with a happy ending, so a mixed outcome is more believable. Maybe you were partly right, or the plan worked in a way you didn’t expect. Maybe you never found out. Any of those shows you can commit without needing to win.
Get the day-one sentence right before you need it
Most of what goes wrong after a lost argument happens in a few sentences. There’s the message to the decider on day one and the answer to a senior person who asks whether you agreed. There’s also the note about the first sign of trouble. Those are hard to get right on the spot, because you’re still annoyed and the stakes feel high.
They’re also easy to practice. You can say “I argued the other way, and I’m behind the plan” out loud with Merlin until it sounds like you mean it. Merlin can also play the vice president who asks whether you really agreed. Merlin runs natively in Slack and Microsoft Teams, so you can practice where you already work. What you tell Merlin stays between you and Merlin. Your conversations are private. If you’d like to see how you handle calls like these more broadly, the decision-making assessment is a good place to start.
Rehearse the “I’m behind it” conversation with Merlin before the next decision goes against you.