AI use 25% — Human directed the thesis, structure, all claims validated against sourced data. AI drafted prose and organized research.
Two days ago I was helping one of our founders think through a hard organizational change, and I reached for an old reflex: a McKinsey-style implementation-and-change deck I’ve kept for years. Stakeholder maps, the adoption curve, the burning platform, the eight steps from urgency to anchoring. Solid material. Close to state of the art when I first learned it. I started walking the founder through it, and somewhere around the third slide I realized I was handing them a map to a country that no longer exists. Nothing in the deck is wrong. It just belongs to a world that is quietly ending.
I co-founded OSS Ventures on the bet that AI would reshape industry the way Lean reshaped it over twenty years. I thought I understood what that meant. The last twelve months have taught me I was more right on the direction than I thougt, and more wrong on the details. We’ve built new companies from scratch and sat next to our portfolio founders while they tried to navigate a mess none of us had a map for. What surprised me wasn’t the technology. It was watching which kind of person wins, and discovering it wasn’t the kind I’d have bet on three years ago.
Everything below is from the frontlines, which is another way of saying it is a confession. We got things wrong. We backed instincts that didn’t survive contact with reality, and we were slow to see some of what I’m about to describe. So take all of it with a healthy dose of salt. These are patterns from one studio, looking at one slice of the world: B2B, industrial, unglamorous. But they’ve repeated often enough that I’d put money on them. In fact, I do, for a living.
Buying became an act of faith, so don’t be an asshole
The first thing we noticed is that buying AI has become an act of faith.
The SaaS era trained buyers to evaluate a thing. You looked at a feature list, sat through a demo, called two references, and made a decision about a product that would be more or less the same a year later. AI broke that contract. What you buy today is half-obsolete or transformed in six months. The truth of yesterday is not the truth of tomorrow. So the buyer isn’t really choosing a product anymore. They’re betting on a trajectory, which means betting on the people steering it.
That changes who wins. Our founders who are present, reliable, and straight, the ones who pick up the phone when something breaks at 9pm and tell you the hard truth before you ask, are overperforming, sometimes spectacularly. The transactional ones are struggling. I mean the high-ego, high-impact operators who could close a deal and move on, the archetype the old game so often rewarded. It isn’t that being an asshole suddenly became a crime. It’s that you cannot sustain an act of faith with someone you don’t trust, and faith is now part of the purchase.
Doing it well means changing the company, not installing software, so be prepared to do it
The second thing is that doing AI well means changing the company, not installing software.
A good SaaS product could be bolted onto a process and left there. AI, when it actually works, doesn’t sit on top of a process. It dissolves it and rebuilds it. That asks far more of a founder than a clean integration. It asks for empathy, because you are changing how real people spend their days, and most of them are frightened of exactly that. It asks for thoroughness, because the value and the danger both live in the details. And it asks for the patience to reshape an organization rather than ship a tool and disappear.
The slick operator who could win the last decade on charm and a beautiful deck is having a hard time. Detail orientation, deeply unfashionable for years, is quietly back in fashion. The gap I keep seeing is between the founder who can dazzle in a demo and the one whose deployment survives a week on the actual shop floor. Those are not the same person, and the second one is worth ten of the first right now.
The constraint moved from building to imagining, and that’s more rare than you think
The third thing is that the binding constraint moved from building to imagining.
For fifteen years the scarce resource was building. You had an idea, and the real question was whether you could ship it, scale it, afford the engineers. AI collapsed the cost of building almost to zero. Now the question is no longer whether you can build the thing. It’s whether you can even imagine what is suddenly possible. That puts a premium on courage, on invention, and on taste, which is the rarest of the three.
Here is the cruel part. Some of our most capable founders, people who were elite in the SaaS era, are paralyzed. Not because they lost any ability, but because the space of what they could now do is so large it overwhelms them. They were world-class at operating inside constraints that no longer bind, and freedom turns out to be harder than constraint. Meanwhile a handful of our founders have never been more alive. They are moving at a speed and a level of ambition that would have looked delusional two years ago, and it is working. Watching that second group is, I’ll admit, one of the great pleasures of this job.
The team you need changed underneath you and you need to lead through the fog
The fourth thing is that the team you need has changed underneath you.
AI is hollowing out some roles completely while creating others that are almost impossible to hire for. The repeatable team-building playbook, hire ahead of the curve and slot people into a known org chart, doesn’t hold when you don’t know what the org should look like in six months. The scarce thing now isn’t a skill you can put on a job description. It’s leadership through uncertainty: the ability to act fast and with conviction while staying humane, to make genuinely hard calls without curdling into cruelty, to hold a group together while the map keeps redrawing itself. The rarest competence in this era turns out not to be technical at all. It’s character under pressure.
What it adds up to : the new game
Put the four together and they rhyme.
The SaaS era paid for transactional efficiency, surface-level breadth, flawless execution inside known bounds, and a scalable, repeatable way of building teams. The AI era pays for the opposite: faith, depth, imagination, and a kind of human judgment you cannot mass-produce. The thread running through all of it is almost embarrassing to write down, because it sounds like something printed on an office wall. Character is back.
I mean something specific and unsentimental by that, though. Not virtue for its own sake. The point is that the actual conditions of this moment, the speed, the uncertainty, the depth of the change AI forces through an organization, happen to turn trust and depth and courage and judgment into competitive advantages rather than nice-to-haves. For once, the market is paying you to be good. That is not a moral claim. It’s an observation about what is working.
The McKinsey deck is still on my laptop. I won’t delete it. It’s an honest record of how we used to believe transformation worked, back when it was a process you could manage with enough rigor and a clear enough chart. What I’ve watched over the last twelve months is messier and more human than any framework I own: founders failing, founders flying, all of us improvising in real time and pretending slightly less than we used to.
It is the worst of times to be the founder built for the world that’s ending, and the best of times to be the one built for the world arriving. I have never been more grateful for the seat.