Don't F*** Up the Culture: Psychological Safety Is the Founder's Job

Replacing one person in a UK tech team costs about £31,800, and the new hire takes 29 weeks to get up to full speed. That’s from an Oxford Economics study in 2014, so it’s likely more today. If you lose three good engineers in a year, that’s close to £100,000 before anyone has built anything.

Psychological safety is the cheapest way I know to keep that number down. It’s also the thing I talk about most when I go into a company as a fractional CTO, because it decides whether people tell you about problems early, while they’re still cheap to fix. And it sits with the founder, not with HR.

Thiel’s advice came after the money

In 2012, after Airbnb closed its Series C, Brian Chesky asked Peter Thiel for his single most important piece of advice. Chesky describes him as someone who “just gave us $150M”. His answer was “Don’t f*** up the culture.” Chesky wrote it up in a letter to the whole team in October 2013, and published it a few months later.

What I find interesting is the timing. Thiel didn’t say it at seed. He said it once the money was in and the hiring was about to start, because he thought most companies mess up their culture once they get to a certain size. Chesky’s view was that “if you break the culture, you break the machine that creates your products.”

Psychological safety is a big part of that. When people stop feeling safe to speak up, the problems are still there. You just stop hearing about them until they’re expensive.

Why the raise is the risky bit

I’ve seen this around funding rounds more than once. The team is under more pressure than it’s ever been, with investor updates, a roadmap with dates on it and a hiring plan to deliver. And the reward usually waits. Pay reviews and promotions get parked “until the round closes”, so people carry more stress with nothing to show for it yet.

If they’re patient and the round lands, the rewards arrive, and that’s the story people tell afterwards. But if the round slips, shrinks or doesn’t happen, the problem is bigger than a missed pay rise. What the founder said and what actually happened no longer match, and the team notices. That’s an integrity problem, and in my experience it damages psychological safety faster than anything else.

Then there’s the hiring that follows the money. A 2024 study by Bransby, Kerrissey and Edmondson followed more than 10,000 healthcare workers and found that new joiners started out feeling safer than the people already there, then lost it over time. In teams with low safety, it dropped sharply and didn’t come back. So the people you hire after a raise are the ones most affected by whatever culture is already there.

The five habits

These are the five habits I work through with founders and their leadership teams. Each one makes the next one easier, so I treat them as a cycle rather than a checklist.

1. Flatten the room

If your opinion ends the debate the moment you say it, people stop giving you theirs. So speak last. Ask the most junior person in the room what they think before you say what you think, and when someone disagrees with you in front of the team, thank them in front of the team.

2. Agree the norms

Write down how the team works: how decisions get made, how people disagree, how quickly you reply to each other, and what happens to pay and promotions during a raise. That last one is the one founders tend to leave out, and it’s the one people worry about most.

3. Act fast and fairly

Most teams have had a brilliant jerk at some point. If you tolerate one because they ship, everyone else learns how safe they really are. Deal with it quickly, and by the same rules you’d apply to anyone else, because the other four habits only work if people believe you will.

4. Listen and give space

Surgical teams that follow the WHO’s safety checklist start each operation with everyone saying their name and role, because the WHO found it makes people more likely to speak up later, especially junior staff. You can do the same thing in a meeting: ask who hasn’t spoken yet, then wait for the answer. It only works if you actually want to hear it.

5. Treat work as a bet

Call a new feature, a pricing change or a hire an experiment, and say up front what result would prove you wrong. Then a miss is something the team learns from, and nobody has to carry the blame for it. Amy Edmondson’s advice when something fails is to be curious before you start blaming.

You can see this with AI tools already. In Microsoft and LinkedIn’s survey of 31,000 people, 52% of those using AI at work were reluctant to admit it for their most important tasks. If your team is trying new tools and hiding it from you, you can’t learn from what’s working.

Doing what you said you’d do

The slide I adapted these habits from had them going round in a circle with nothing in the middle. I’ve added what I think all five depend on: what you say matches what you do.

Each habit is really a promise. “I want to hear disagreement.” “We’ll deal with bad behaviour.” “It’s fine to try something and get it wrong.” The team believes each one as far as they’ve seen you keep it, and breaking one makes the rest harder to believe. Chesky said culture is “a thousand things, a thousand times”, and I think that’s right. It’s mostly what you do on an ordinary week when the numbers are bad.

Around a raise, that comes down to three things. Don’t promise rewards the round might not deliver. If the timing slips, tell the team before they work it out for themselves. And if it doesn’t land, say so plainly and say what happens next. People will forgive a round that didn’t close, but they won’t forgive finding out about it from someone else.

Safety needs standards

Psychological safety on its own isn’t enough. In a 2011 study, Pearsall and Ellis found that teams leaning towards “the ends justify the means” cheated, but almost only when the team felt psychologically safe. Edmondson makes the same point from the other side: high safety with low standards gives you a comfortable team, but not one that’s learning.

So habit 3 really matters. With safety and clear standards together, you get a team that tells you the bad news on a Monday, rather than in an exit interview.

How to tell if it’s working

Engagement surveys are slow and easy to game, so I look at what people do instead:

• Who disagrees with you in meetings. If nobody ever does, that’s a warning sign.
• How long bad news takes to reach you. Time it from when the problem started to when someone told you. In a safe team that gap gets shorter.
• Who owns up first after an incident. In a safe team it’s the person who made the mistake, and they say so in the team channel.
• Who leaves. Good people leaving and mentioning “culture” is the most expensive signal of the lot, at £31,800 each.

If you want the engineering-team version of this, with the warning signs to look for in code review and post-mortems, it’s in my earlier post on psychological safety. And if you’re heading into a raise and would like a second pair of eyes on how the team is holding up, book a call.

Note 1. The five habits are adapted from a HumanOS slide shown at the Connexion Career Collective’s future-of-work summit in London. At the same event Bianca Errigo shared how her company measures culture, and found psychological safety to be the first thing a team needs for sustained high performance. The founder moves and the centre of the cycle are mine.