Most repair attempts fail at the first step rather than the last. “I’m sorry, I’ll do better” is a promise about the future offered in place of an account of the past, and the other person is still holding the past.
There’s no script for this, and anyone offering you one is overselling it. What there is, is a way of holding the work. Three questions, in order: what did the breach mean to them, what can be put right now, and what evidence will they need next?
Should you apologise first?
No. Work out what actually broke first, or you’ll end up apologising for the wrong thing.
The same action doesn’t damage trust with everybody. Erosion is contextual in exactly the way trust is contextual. It depends on the person, the situation, what they were expecting and what they value. Two people can watch you do an identical thing and reach opposite verdicts.
It also matters what they think the behaviour says about you. Someone thinking you got it wrong is a different problem from someone thinking you lied to them. They don’t respond to the same words. Researchers have tested this directly, comparing how people react when the question is about someone’s competence and when it’s about their integrity. What helped in one case didn’t in the other. That study used allegations against strangers, so it won’t hand you a script. It does settle one thing, though. The same apology doesn’t do the same job on every kind of break.
Vague ownership gives you away. “I know things haven’t been great between us” tells someone you still can’t see it. This is usually where feedback earns its keep. Our view is that most erosion is unawareness rather than malice, which is our position rather than a research finding, and the person who can tell you what landed is the one who had to live with it.
What did the breach mean to them?
Start with the behaviour you can point to, then find out what changed in their judgement of you.
Did they stop believing what you say? Stop expecting you to follow through? Start doubting you could handle it? Did they feel unheard, or as though their interests didn’t count? You’re not trying to file the breach in a box. You’re working out which piece of confidence took the damage, because that’s the one you have to rebuild.
How big the break is depends on how far your behaviour sat from what that person values. Be twenty minutes late to someone who treats punctuality as ordinary friction and you’ve caused an inconvenience. Be twenty minutes late to someone for whom being on time is respect, and you’ve told them something about how much they matter. Same action. Different injury.
What can be put right now?
Own what you did, apologise where the responsibility is actually yours, and correct whatever can be corrected.
An apology that holds water shows two things: the behaviour, and what it created for someone else. “I’m sorry I missed the deadline” owns the behaviour. “I’m sorry I missed the deadline. You went into that meeting without the work you were expecting, and you carried the consequence of it” shows you can see it from where they were standing. That’s the part people are listening for.
Don’t invent the impact to make the apology sound better. If you know what it cost them, name it. If you only know how it landed for you, own that much. If you don’t know, ask, and then let them tell you rather than finishing the sentence for them.
Then fix what’s fixable. Repair should fit the harm. An explanation doesn’t correct a loss, and an apology doesn’t return what’s owed. If something was lost, return it. If a process was unfair, change it. If information should have been shared, share it. If the same thing could easily happen again, put something in the way of it.
Sometimes you can’t say everything. Confidentiality, someone else’s privacy, an investigation in progress. The standard was never tell everyone everything. It’s to be as clear as you legitimately can about what’s known, what isn’t, what you own and what you can’t discuss.
What evidence will they need next?
Most of repair happens here, and it’s the slow part. A promise helps. It isn’t the thing.
Make the next commitment specific enough that they could check it, then keep it. “I’ll see you Friday”, then turn up on Friday. If the problem was missed delivery, deliver when you said you would. If it was being surprised, warn them. If someone was left out, change the point at which they come into it.
Sometimes the fix isn’t only in your behaviour. If the same thing could easily happen again, change something so that it’s harder. Agree a check-in halfway through rather than both waiting for the deadline. Write down what finished actually means, so you’re not guessing at it separately. It doesn’t have to be a big change. The point is that they aren’t left relying on your word that you’ve learned. They can see something is different.
Then let the evidence stack up without asking for credit for it. Don’t audit their progress or ask whether it’s working yet.
Repair is not the promise. Repair is what becomes believable afterwards.
How long does it take to rebuild trust?
Nobody can tell you, and be sceptical of anyone who does.
We’ve looked. No study puts a number on how many kept commitments, or how many weeks, a real working relationship needs. The figures that circulate are invented, and an invented number does real damage here. Tell someone who’s been unreliable for three years that the rebuild takes three years and they won’t start at all.
What history does do is give the new behaviour somewhere to land. Experiments on trust recovery found people came back further after a breach where there was more positive experience beforehand. That’s not a clock and it isn’t a guarantee. Those were anonymous partners in a short game, not colleagues of ten years. Time passing isn’t repair, and a long good record doesn’t make a serious or repeated breach disappear.
Can trust ever be fully rebuilt?
Sometimes. Not always.
The people who study this make a distinction worth borrowing. There’s partial repair, where trust comes back up but not all the way, and there’s complete repair, where it returns to what it was. A review of forty years of the research treats those as two different endings rather than two points on the same road. Most of what’s been studied is the partial kind.
An apology helps. It also isn’t the finish line. In one experiment, people who got an apology after a breach put more back into the relationship than people who got nothing, and still less than people whose trust had never been broken in the first place. That was a short exchange between students rather than colleagues with a history, so read it as a hint about which way things tend to go, not as a prediction about you.
The finding that sticks is the one about lying. In a repeated game, a run of trustworthy behaviour brought trust back after someone had behaved badly. Where that same behaviour had involved deception, it didn’t fully come back. Not after a promise, not after an apology, not after a long stretch of doing better. Small study, young participants, a laboratory rather than an office. But it matches something most of us already know. Being let down and being lied to aren’t the same injury, and they don’t heal at the same speed.
A repaired relationship can end up narrower or more careful than it was. Sometimes it doesn’t come back at all. That’s a harder answer than a repair script, and a more useful one.
When this isn’t the right approach
Everything above assumes ordinary damage between two people who both want to carry on working together. It isn’t legal advice and it isn’t HR advice. If there’s a formal process running, a safeguarding concern, or any question about someone’s safety, that comes first and nothing here overrides it.
Power matters too. Repair is harder to do honestly when one of you can affect the other’s job. Someone who depends on you will often say things are fine long before they are, and none of the research below sorts that out for you. If the recovery looks quick and you’re the senior one, treat that as a question rather than an answer.
What repair is not
It isn’t authority. Reaching for the policy, the process or your position to close a trust gap papers over it, and it shows someone acting in line with the rules rather than with their values. “We were within our rights” belongs nowhere in a repair.
It isn’t a single conversation. The conversation opens the door. The pattern walks through it.
And it isn’t guaranteed. Sometimes people stop lending. Which is the best argument there is for practising the behaviour before you need it, because a long pattern works the other way too. If you’ve been consistent for years, an occasional slip is easier for people to absorb.
The three questions are Trust Leader’s own synthesis, not a tested method. They’re drawn from the repair research listed below, which supports something narrower than any sequence: breaches aren’t all the same, an apology is a beginning rather than a conclusion, and repair is often partial. An earlier version of this page set out five steps and suggested the rebuild had to run longer than the old pattern did. Both have been removed. They made the work look more predictable than the evidence allows.
None of this research tests the Trust Leader approach, and none of it shows that any particular practice will rebuild trust. Reviewed and updated 17 August 2026 by Scott Hunter.