How to Have a Hard Conversation With Your Cofounder
You have probably already decided to have this conversation. You decided weeks ago, and then a launch happened, and then it seemed like the wrong week, and now the thing you need to say has accumulated interest. What follows is a method rather than an argument — how to prepare, what to say in the first two minutes, how to keep the middle from collapsing into a dispute about intentions, and what to do with each of the four reactions you are likely to get. The aim is not a comfortable conversation. It is one that changes something.
What’s inside
Why this is harder than any conversation you have as a manager
Four things to settle before you start
The opening two minutes
The middle: separate the facts from the meaning
Four reactions, and how to answer each
Close with something written
If it goes badly
Where Blomma fits
Why this is harder than any conversation you have as a manager
Founders who deliver difficult feedback competently to their teams often cannot do it with a cofounder, and there are structural reasons rather than personal ones.
There is no authority gradient. Managerial feedback works partly because the role gives it standing. With a cofounder you have no standing except the relationship itself, which means the conversation’s only support is the thing you are putting under strain.
There is no process. Performance conversations have a review cycle, a template, a documented expectation. Cofounder conversations have no container, so raising one is itself an event — the request for a meeting communicates something before you have said anything.
The history is asymmetric. You have years of shared context, which cuts both ways: they will hear the current issue against everything that came before, including things you have long since let go and they have not.
And there is no exit. You cannot manage a cofounder out of your day-to-day the way you can restructure around an employee. Whatever the outcome, you both still have to run the company on Monday.
That combination explains the delay. It does not make the delay cheaper — unsaid things between founders compound, and the team can usually see the avoidance well before either of you names it.
Four things to settle before you start
Preparation is most of the outcome, and it is where the postponement usually hides. Founders who feel unready are usually missing one of these four.
What you actually want.Not what you want to say — what you want to be different afterwards. Write it in one sentence. If you cannot, you are not ready, and the conversation will become a general airing that resolves nothing. “I want us to agree who decides on pricing” is workable. “I want him to understand how this has been for me” is a feeling, not an outcome.
Three specific instances.Concrete, recent, and describable without interpretation. Instances can be discussed; characterisations can only be disputed. Three is the right number — one reads as an isolated incident, and five reads as a case you have been building, which is exactly what your cofounder will accuse you of.
Your own part in it.What you contributed: feedback you never gave, a boundary you never defined, resentment you let accumulate rather than raise. Not as a concession. Because they will raise it, they will be partly right, and having already accepted it lets you stay in the conversation instead of defending.
Which conversation this is.Are you asking for a change, renegotiating the arrangement, or opening an exit? Your cofounder will try to work this out in the first minute, and if you do not tell them they will assume the worst version and respond to that instead of to you.
The opening two minutes
The opening determines whether the next hour is a conversation or a defence. It has four parts and takes about ninety seconds.
Name what this is.“I want to talk about how we’re dividing decisions. This isn’t me questioning whether we should be doing this together.” One sentence of frame, one sentence of reassurance if the reassurance is true. Do not offer it if it is not.
Say the thing.Directly, early, in plain language. “I don’t think we’ve ever defined who decides on pricing, and it’s cost us three arguments in two months.” The instinct is to build up to it. Building up is worse: your cofounder knows something is coming, and every sentence of preamble raises their anticipation of how bad it is.
Give one instance.Specific, recent, factual. Not all three yet — one, so they can engage with something concrete rather than a summary.
Hand it over with a real question.“How do you see it?” Then stop talking, and let the silence run longer than is comfortable. This is the part founders skip, and skipping it converts the conversation into a presentation. You cannot learn anything about the situation while you are still explaining your view of it.
What to avoid in the opening: any sentence beginning “you always” or “you never,” any account of their motives, and the phrase “I’ve been thinking about this for a while” — which is true and lands as evidence that you have been quietly accumulating a file.
The middle: separate the facts from the meaning
Most cofounder conversations fail in the middle, and they fail in one predictable way: the two of you collapse what happened together with what it meant, and then argue about intent. Intent is unfalsifiable, so the argument has no end.
Keep them apart deliberately. First establish what actually happened — the sequence, the decisions, who knew what. This part is usually less contested than either of you expects, and agreeing on it builds the only footing the rest of the conversation has.
Then move to meaning, and own yours as yours. “When that happened I read it as you not trusting my judgment” is a statement about your interpretation, which invites correction. “You don’t trust my judgment” is a claim about them, which invites denial. The information content is identical and the outcomes are completely different.
Watch for two moves that end progress. One is escalation to the general — the conversation leaving the specific instance for a broader claim about how things always go. Bring it back: “Can we stay with the pricing decision for a minute?” The other is trading grievances, where each item you raise is met with one of theirs. That pattern feels like engagement and produces nothing. Name it directly and propose taking theirs properly, in sequence, once you have finished with yours.
Four reactions, and how to answer each
Four responses cover nearly everything you will get.
Agreement, quickly.“You’re right, I’ve been aware of it.” Genuinely common, and the risk is relief — you accept the agreement and leave without changing anything. Fast agreement without a structural change is how the same conversation happens twice. Go straight to what specifically changes.
Defensiveness.Justifying, contesting the instance, explaining the context. Usually the first response of someone who is surprised, and often followed by real engagement if you do not escalate. Acknowledge the substance of what they said, then return to the specific: “That makes sense. And I still want to solve who decides.” Repeat as needed without heat.
Counter-attack.They raise a grievance about you, immediately. Sometimes deflection, sometimes something real they have also been postponing — you often cannot tell in the moment. Take it seriously and sequence it: “I want to hear that properly. Can we finish this and then take yours, or would you rather do yours first?” Offering to go second is disarming and costs you nothing.
Withdrawal.Minimal engagement, agreement without substance, a clear wish for it to end. The hardest, because it looks like resolution. Do not chase it in the room. Name it gently and reschedule: “I don’t think we’re getting to it today. Can we pick this up Thursday?” A second conversation with a prepared counterpart beats an hour of pressing an unwilling one.
Close with something written
A cofounder conversation that ends with mutual goodwill and nothing recorded has not concluded. It has paused, and both of you will remember the agreement differently within a fortnight.
Before you leave the room, agree three things out loud: what specifically changes, who does what, and when you will check whether it is working. Then write two paragraphs into a shared doc that same day. Not a contract — a record. Verbal agreements between founders reliably evaporate and then return as a dispute about what was agreed.
And schedule the follow-up before you separate. Serious cofounder conversations are a sequence rather than an event, and naming the next one removes the need for either of you to re-initiate — which is the step that does not happen.
If it goes badly
Sometimes it does. The useful thing to know is that a bad first conversation is not the outcome.
Give it a few days. People process at different speeds, and the response that arrives on day three is frequently better than the one in the room — partly because your cofounder needed to hear it without an audience of one.
Then go back once, briefly and without relitigating: “I think I opened that badly. What I was trying to say is X.” Founders skip this because a failed conversation feels like a closed door. It usually is not, and taking responsibility for the delivery without withdrawing the substance is the single most effective repair available.
If a second properly-prepared attempt also does not converge, that is real information — and the point to consider outside help, whether a mediator, a board member with standing, or separate coaching so each of you can work out what you actually want before negotiating for it.
Where Blomma fits
The reason this conversation goes badly is almost never that founders lack courage. It is that there is nowhere to practise. You cannot rehearse it with your cofounder. Your team is not a safe audience. Your investors have a stake. The friends who know you both are already partisan.
Blomma is an always-on AI career coach with no position in the outcome. Use it to get clear on the one sentence describing what you want to be different, which is where most of the postponement actually lives. Use it to test whether your three instances are factual or interpretive. Use it to rehearse the opening two minutes out loud until they are ninety seconds rather than a fifteen-minute build-up — the highest-leverage preparation available. And use it to plan responses to all four reactions, so you are not composing them under pressure. When the situation warrants a human who has sat in a founding pair, you can bring one in..
The conversation you are postponing is almost always smaller than the version of it you will be forced into six months from now. Prepare it properly, open it directly, and write down what you agree. That is most of the method.