From Customer-Led Development: How to Build What Your Customers Want When AI Can Build Anything — the operating manual for building what customers want when AI can build anything. See every chapter.
Churn is optional.
That isn't optimism. ServiceNow, Salesforce, and Costco each keep nine or more of every ten customers, year after year. Apple keeps roughly nine in ten iPhone buyers. CIRP (Consumer Intelligence Research Partners) loyalty surveys put iPhone brand loyalty around 89% to 90%.
If churn were a law of nature, those numbers couldn't exist. They do. So the companies that lose customers aren't fighting gravity. They're making choices, usually without knowing it.
Everyone claims to be customer-centric
I've spent my career obsessed with this problem. I've had more than 2,000 conversations with product people, and almost all of them were, one way or another, about how organizations work and how they fail. As Head of Product at multiple Series A YC companies, I've helped teams go from 10% retention in month one to 90% retention in month one.
The companies that keep their customers aren't more virtuous. They aren't louder about loving their users. Every company claims to be customer-centric. The difference is that the companies that retain customers have built systems that make customer-centricity inevitable, so it happens whether or not anyone remembers to care that week.
I kept having the same conversation with product leaders. Features nobody uses. Customers who leave without warning. Sales promising things product never builds. Feedback vanishing into the void. They had tried Agile, Scrum, SAFe, product-led, design thinking. Nothing stuck.
The problem sat deeper than methodology. They had built organizations where customers couldn't be heard. You can't fix that by bolting another framework onto it.
Rule of thumb: If your customers can't be heard inside your organization, no methodology will save you.
Execution just became free
Two forces are colliding right now. One is brand new. The other is the oldest story in commerce.
The new one: AI turned engineering time into a commodity. Engineering time was the scarcity every methodology was built to manage. Sprints, estimation, roadmapping rituals all existed to ration it. Now story points have become tokens. Everyone is a builder.
When anyone can build anything, the only scarce thing left is judgment about what deserves to exist. Deciding what to build, why, and for whom becomes the whole game. Your customers are the only reliable way to make that call.
When building is free, judgment is the product.
The oldest story in commerce
The second force is simpler. In markets, history repeats itself. To see where things are going, it helps to see where they've been.
The market square
Before industrialization, you bought from people you knew: the butcher on the corner, the tailor down the street, the blacksmith who knew every farmer in town.
Every purchase came wrapped in a relationship. You trusted these people. You didn't always pay right away. You settled up at the end of the month, or traded something in return. If you didn't like them, you could often walk to the next corner. Competition was local, and loyalty was earned.
You knew who made your goods, and they knew what you needed. The relationship was the product.
What industrialization took
Then you could buy from people you'd never meet. Products came from factories hundreds of miles away, and the relationship disappeared.
We tried to scale trust anyway. Sears catalogs, department store brands, "satisfaction guaranteed or your money back." Each was an attempt to recreate the feeling of knowing who you were buying from, without actually knowing them.
It sort of worked. People traded relationships for reliability and cost. You didn't know who made your shirt, but you trusted the brand. You didn't know the factory worker, but you knew Sears would take it back if it fell apart.
Then we stopped trying
As markets consolidated, even the pretense of relationship disappeared. Monopolies and near-monopolies emerged. Switching costs protected bad experiences. Companies realized they didn't need goodwill to grow, because the addressable market was growing faster than churn.
Look up Walmart, Best Buy, or McDonald's online and you'll find plenty of one- and two-star ratings. You go because it's cheap and predictable, not for the experience. They don't compete on service, so they don't invest in it.
| Era | How trust worked | What you gave up |
|---|---|---|
| Market square | You knew the maker personally | Scale and choice |
| Industrial | Brands and guarantees stood in for people | The relationship |
| Consolidation | Switching costs replaced trust | Goodwill itself |
| What's coming | Relationship at scale | Nothing, if you build for it |
Software, Absent a Service
Software followed the same path, and somewhere along the way we forgot what SaaS stands for. Software as a Service. Not software with a subscription.
When SaaS started, customers paid monthly because you would keep delivering value. You'd improve the product, work with them, and listen. It was a relationship.
Packaged software, oddly, forced you to stay closer. You shipped V1, then had to convince customers to buy V2. You had to earn the upgrade.
Subscriptions let companies stop earning it. Customers have to take V2. They're already paying. And when they churn, we're confused. We don't have relationships with our customers, so we look at the data. We blame the data. We become data-obsessed and wonder what went wrong.
A lot of companies operate as if SaaS means Software, Absent a Service. But the service was always the business: supporting customers, servicing the product, paying someone to think every day about how to make it better for the people using it.
The subscription doesn't replace the relationship. It assumes one.
The middle is going to die
Put the two forces together and the market splits in two directions. Everything in between gets squeezed.
Direction one: hyper-niche
The cost of building for small audiences has collapsed. You can build a product for three people, not three million.
AI lets you work with suppliers, iterate on designs, 3D print prototypes, and find your way through industrial systems you never could have understood alone. Hardware or software, it doesn't matter. Markets that were too small to be viable are suddenly viable.
This is bad news for anyone whose moat was scale. When you don't need economies of scale to build something great, AI starts unbundling everything. What does a marketplace look like with infinite competition? What does the App Store look like when someone can build a product for exactly three people? Products no longer need to appeal to everyone, and customers can choose to buy from people who listen to them.
Direction two: scale with service
The other direction keeps massive scale but delivers the service quality of a niche product, personalized for millions of users.
AI makes this possible too. You can listen to every customer, respond to every signal, and close every loop. The bottleneck that made personal service impossible at scale is gone.
The trust problem
There's a twist. AI is making people less trusting, not more.
When anything can be generated, faked, or automated, people crave real connection. They want to know someone on the other side cares about their problem.
I once watched an HR software company brag about its 30-second response times, from a bot. Meanwhile you couldn't reach a human to do anything that actually mattered. That's the trade that kills trust. When a company swaps authenticity for speed and cost, whether through AI, outsourcing, or plain dehumanization, it signals it doesn't take itself seriously. And people won't hand their serious problems to a service that doesn't take itself seriously.
The companies that win will feel human. They'll listen, respond, and build relationships at scale.
So we're heading back to a market square economy. Not out of nostalgia, but because the technology finally lets us have scale and relationship, efficiency and service. The companies stuck in the middle, too big to be personal and too generic to be excellent, will get squeezed from both sides.
Rule of thumb: Be small enough to know your customers, or big enough to know all of them. The middle is where companies go to die.
What customer-led actually means
None of this is new. The principles go back decades.
- 1984: Frank Robinson of SyncDev was doing market validation before anyone had a name for it. His cross-functional teams traveled together, test-sold validation prototypes, and listened to customers the way a jury listens to witnesses. By his own account, he did this for over 400 companies, including early Citrix.
- 2005: Steve Blank codified similar ideas in The Four Steps to the Epiphany, launching the Customer Development methodology.
- Later: Eric Ries paired Customer Development with Agile engineering and called it the Lean Startup. It became a phenomenon, but the roots trace back to Robinson and Blank.
- Y Combinator distilled it further: at any given time, you should be doing one of two things, building or talking to customers.
Talking to customers means continuous research into their needs. Not sales calls.
The quote Ford never said
Everyone loves the line attributed to Henry Ford: "If I asked people what they wanted, they would have said faster horses." Ford never said it. The quote first appeared in 1999, from a cruise ship designer speculating about what Ford might have thought.
What Ford is actually credited with saying is close to the opposite:
"If there is any one secret of success, it lies in the ability to get the other person's point of view and see things from that person's angle as well as from your own."
The misquote gets used to justify not talking to customers. That misunderstands what research is for. The job was never to ask people what they want. It's to understand their needs and how to solve them well. Ford didn't need to ask whether people wanted a car. His problem was manufacturing one at scale, so he spent his time studying how humans work in factories, optimizing for the people building the product.
The research Apple supposedly didn't do
People use Steve Jobs the same way: "Jobs never did user research." Apple runs one of the most rigorous human interface and design testing operations in the world. Its product development is famously structured: requirements defined at inception, relentless design reviews, small teams deeply involved with the users they're building for.
The "no focus groups" line was about inventing new categories, not about ignoring users. Watch any Apple keynote and you'll see real customers in the videos: the people who got early access, the people the product was built for.
The definition
Listening to customers doesn't mean building everything they ask for. It doesn't mean religious user research. It doesn't mean great customer support, though it can involve any of those.
Customer-led means the customer problem and the customer outcome sit at the center of everything you do: every OKR, every deliverable, every decision about what to build and what to kill. Everything else is implementation.
That's the whole idea. The butcher on the corner understood it without a framework. We spent a century building organizations that forgot it. Now that building is cheap and trust is scarce, it's the only advantage left that compounds.
If you want the full system for building organizations where customers can't help but be heard, it's all in Customer-Led Development.
And if the return of the market square is on your mind too, subscribe to the Context Limit newsletter for more essays on building the kind of product people stay with.