Everything is repricing

Ada This is Bets and Ships, out loud. I'm Ada.

Sam And I'm Sam. Quick note before we start. We're both AI voices, and an AI wrote this script, working from Chris's essay, Everything is repricing. So think of this as the conversation you might have after reading it. The essay itself, with every source, is on the site.

Ada Okay. So, Sam, I want to start with a strange week.

Sam Go on.

Ada February this year. In a single day, something like two hundred and eighty-five billion dollars comes off the value of software, data and financial companies.

Sam Because of what? A scandal? A bad earnings call?

Ada A product launch. And one simple line of thinking that went through the market all at once. If an AI agent can do the work of five people, and your software is sold per person...

Sam ...then you're selling to a number that's about to shrink.

Ada Exactly. And the thing Chris points out is that none of this was news. You could have made that argument two years earlier.

Sam So what changed?

Ada Enough people believed it on the same afternoon.

Sam Huh. That's a slightly unnerving way for prices to move.

Ada It is. But the real point of the essay is that this one loud event is an example of something quieter, happening in a lot of places at once.

Sam Where?

Ada Four places. Product managers. Designers. How software is priced. And how software companies are valued. His claim is that there's one mechanism under all four.

Sam Which is?

Ada Making things got cheap.

Sam That sounds almost too simple.

Ada It does, so let's be careful about what "things" means. Not good decisions. Not insight. The stuff. A spec document. A screen design. A working prototype. A summary of fifty customer interviews. A first draft of some code.

Sam The deliverables.

Ada The deliverables. Each of those used to take a skilled person days, sometimes weeks. Now it takes minutes, and the result is often pretty decent.

Sam Okay, but why does that shake up careers? People still need those things.

Ada Here's the part I found clever. For years, we judged professionals by what they delivered. Not because the deliverable was the real value. Ask anyone senior and they'll tell you the real value is judgment. But judgment is hard to see, and the deliverable was a good enough stand-in for it.

Sam Like a portfolio.

Ada Like a portfolio. If your designs were great, you were probably making good calls. That link held for a long time.

Sam And now anyone can put together a great-looking portfolio.

Ada Right. So the stand-in stops telling you anything. The portfolio still looks good. It just doesn't separate one person from another anymore.

Sam So employers go looking for something else.

Ada They go looking for the judgment itself. And the people who did everything right under the old rules feel like the ground moved under them. Chris describes it as a measurement problem that feels like betrayal.

Sam Oof. Yeah. That lands.

---

Sam Let's start with product managers, since that's Chris's world. What's going on there?

Ada The headline numbers look like they contradict each other. Job postings for product managers are up on last year. But the market overall is still well below where it was at the peak in twenty twenty-two.

Sam So, good news or bad news?

Ada Both, which is why nobody agrees. The more telling thing is which jobs. There's a well-known example where Atlassian cut about sixteen hundred jobs, about a tenth of the company, and said the savings would pay for AI and enterprise sales.

Sam So if you were one of the sixteen hundred, it's a layoff.

Ada And if you only read the announcement, it's a hiring story. Both are true.

Sam What about pay?

Ada That's the striking part. In late twenty twenty-four, Glassdoor numbers put U.S. product managers with an AI focus at roughly double what traditional ones made. Around two hundred and forty-five thousand dollars, against around a hundred and twenty-four. The gap has narrowed since. This year's averages are closer to a hundred and ninety-eight, against a hundred and fifty-one.

Sam That's still a big gap.

Ada It is. When the same title pays that much more depending on one word in front of it, you're starting to look at two different jobs that happen to share a name.

Sam So what's actually different about the job?

Ada Think about where a product manager's week used to go. Writing specs, because a document was the only way to show an idea without tying up engineers. Waiting a week for research to be pulled together. And arguing in meetings, because running an experiment cost more than arguing about it.

Sam And now?

Ada You can build a clickable version of the idea in an afternoon. The research summary takes an hour. Experiments are cheap. So the meeting becomes the expensive option.

Sam Ha. The meeting is the expensive option. I like that.

Ada So a big chunk of the coordinating work doesn't get automated, exactly. It just stops being needed. What's left is the deciding, and that part got harder.

Sam Harder how?

Ada More options to choose between. Products that give a different answer every time you ask. And new kinds of risk that nobody owned five years ago.

---

Sam Designers next. Same story?

Ada Same shape, different words. Teams are getting smaller and more specialized. The work that's growing is the work that needs judgment. The pure production work isn't.

Sam And I'd guess the portfolio problem is even sharper for designers.

Ada It is. The design world has a neat way of putting it. AI didn't replace designers. It made the output cheap. And hiring managers have quietly changed their question. It's less, is this person talented? And more, can I trust this person when things are unclear?

Sam And a portfolio can't answer that.

Ada Because a portfolio shows what you picked. It hides everything you threw away, and why. And the throwing away is the part that's still hard.

Sam That's a good way to say it. You also hear people say product and design are merging.

Ada Chris pushes back on that a little. His view is that both jobs were a big pile of execution with a smaller core of judgment. Take the execution away from both, and what's left looks similar. Not because they merged. Because the same layer came off each of them.

Sam So where does judgment actually come from?

Ada From finding out whether you were right. Watching people struggle with the thing you shipped. Checking the number after launch. And here's the uncomfortable bit. If you never check, years of experience don't turn into judgment.

Sam Thirty launches, no follow-up...

Ada ...is thirty launches of experience, and not much sense to show for it. He admits that was partly true of him, too.

---

Sam Okay. Money. The software pricing part.

Ada This is my favorite bit. Remember the per-person pricing?

Sam The thing that spooked the market.

Ada Right. Software has mostly been sold by the seat, because the number of people using a tool was a decent stand-in for how much you got out of it. More people, more work, more value.

Sam Until agents do the work, and nobody's logged in.

Ada Exactly. And then it flips. The better the product gets at doing the work, the fewer seats a customer needs, and the less the vendor earns. It punishes both sides for the product working.

Sam So what are companies doing about it?

Ada There's a survey of a few hundred software CEOs from this spring. Nearly all of them, ninety-seven percent, plan to move away from seat pricing within two years. And ninety-four percent say seat pricing fits their product today.

Sam Wait. Both?

Ada Both. And Chris argues that isn't confusion. It's what it looks like when a measure still works, but you can already see where it stops working.

Sam Like standing on a floor you can see cracks in.

Ada That's the picture, yes.

Sam So will they all switch to charging for results? Per resolved support ticket, that kind of thing?

Ada That's the story people tell. In practice, most companies are landing on a mix. A steady base fee, with usage charged on top. Credits have become really common too. Pure pay-for-results is still rare.

Sam Why? It sounds fairer.

Ada Because you have to define the result precisely enough to put it on an invoice. And the customer has to trust how you counted it. Those turn out to be product problems, and often nobody owns them.

Sam And meanwhile, the buyers...

Ada ...are getting stricter. They check who actually uses the licenses. And vendors have been pushing renewal prices up by folding AI features into bundles. Buyers have a name for that. The AI tax.

Sam So vendors aren't sure what to charge for, buyers are counting every seat, and prices are going up anyway.

Ada Which is why a lot of software deals this year have felt so tense.

---

Sam Back to that crash, then. Was the market right?

Ada Chris thinks it was right about the direction, and early on the timing. Nothing about those companies' revenue changed that week. What changed was how confident people felt about the revenue still to come.

Sam And there's a counterweight, isn't there? A lot of AI projects not paying off.

Ada There's a widely quoted MIT figure that most enterprise AI pilots, something like ninety-five percent, don't show a measurable return. People argue about how that was measured, so hold it loosely. But put it next to the fact that most product teams have already shipped AI features...

Sam ...and you get: everyone's doing it, few can prove it's working, and the markets have priced it in anyway.

Ada Right. The optimists have to explain the failures. The skeptics have to explain why the repricing keeps happening.

Sam And what's Chris's read on that gap?

Ada That it's mostly a product problem, not a model problem. What is this for? What does a good answer look like? How would we know it worked?

Sam Which are very much product manager questions.

Ada He noticed that too.

---

Sam So if I'm listening to this, and I work in product or design, what do I actually do?

Ada The advice is almost disappointingly simple. Before you make a call, write down what you expect to happen. Then go back later and check.

Sam That's it?

Ada That's most of it. The old proof, the polished deliverable, doesn't prove much anymore. What does is a record. The bets you made, what you predicted, what you turned down, and how quickly you noticed when you were wrong.

Sam And hardly anyone keeps that record.

Ada Which is exactly why it's worth having. It's cheap to start, and very hard to fake.

Sam Anything else?

Ada Two things. First, the squeeze lands in the middle. The solid, competent people, and the solid, competent products, whose value was the deliverable. Second, whoever gets good at defining what good means, clearly enough that someone else can check it, is in a strong position. Pricing needs that. AI features need that. Hiring needs that.

Sam And the timing?

Ada Expect it to overshoot in both directions. Some of the panic will reverse, and then a slower, more permanent version arrives. So don't trade on the panic, and don't trade on the shrug. Trust the underlying shift, and be patient about when.

Sam Making things got cheap. So everything that was priced on the things is finding a new price.

Ada That's the episode.

Sam The full essay, with the charts and every source, is Everything is repricing, on Bets and Ships. Thanks for listening.

Ada See you next time.