If I had asked people what they wanted, they would have said faster horses.
There’s no evidence Henry Ford ever said that. The line only gets tied to him decades after he died.
His real story is the opposite. Ford launched the Model T in 1908 and kept it largely unchanged for nearly 19 years. Meanwhile, General Motors rolled out a new model every year, in colors other than black. By early 1927, Chevrolet was outselling Ford. That May, Ford shut down its plants for six months to replace the Model T.
Ford didn’t lose because he ignored what people said, but because he ignored what they did.
That’s the core of the book The Mom Test, by Rob Fitzpatrick. People lie about the future to be nice. So you don’t ask “would you use this?”. You ask about the past: the last time it happened, what they did, what it cost them.
Talking to users is the product manager’s job
Plenty of engineers think so. Users are product’s problem. Ours is the ticket.
But a ticket is a user’s problem secondhand, already filtered. Building only from tickets is asking “would you use this?” through a middleman.
The usual way engineers try to grow into product has the same flaw. Ask for the title and wait. Post a demo and count the reactions. Both are opinions about the future, and Fitzpatrick has a name for that kind of feedback: compliments.
Compliments feel like progress. And commitments are progress.
Receipts, not requests
I’ve been at Automatiq since 2022, working remotely from Brazil on a team that builds automation for ticket resellers. I started on the frontend, and I’m still there. But more and more of my work has moved toward product: analytics, dashboards, web vitals, NPS surveys, and the people who actually use what we build.
So I stopped collecting opinions about my career and started collecting evidence:
We put real user monitoring on every page of our product, so we see where people actually wait instead of guessing.
We track NPS, so what users feel is a number I can point to, not a guess.
Before we migrated our legacy screens, I audited all of them, so the plan started from what exists.
We talk to users directly whenever we can. I don’t always love it, but it’s one of the parts of the job I value most.
None of it came with a new title. All of it is product engineering.
I call it receipts over requests: don’t ask for the job, show you’ve been doing it.
The catch is: receipts nobody sees are a diary. They have to land where decisions happen: a short doc, a five-minute demo, a Slack thread, a planning meeting, a roadmap review. And it's usually slow. AI can build the dashboard in an afternoon, but it can't make users show up any faster.
Ford didn't need anyone to tell him what people wanted. The answer was parked on his own street, with a Chevrolet badge on it. He just looked too late.
The title is always the last thing to arrive.
What AI did, what I did
Writing has always been my thing, and I don’t want to lose that. My rule is simple: I write and Claude reviews. This one bent the rule.
Claude found the Ford story, shaped the post into STORY, the writing framework I use now, and dropped it into my Notion content queue. It also got the most important line wrong: it wrote that Ford “never said” it, when the honest version is that there’s no evidence he did. Checking that, and deciding what I could actually sign, was mine.
From my week
I'm reading The Mom Test. This whole post started on one of its pages.
Keep committing,
Eder