Joining an AI startup pre-product market fit
When Henry Modisett joined, there was no product-market fit, no user base, no roadmap. Just a hunch, a prototype, and a small team of engineers. For most designers, that would be a daunting risk. For Henry, it was exactly the kind of environment he was looking for.
After several years in management, he craved the immediacy and satisfaction of making things again.
“There’s nothing more gratifying than building something and seeing someone use it, even if it’s just one person,” he said.
That impulse pulled him back into startup life — not with a specific idea, but with a desire to get close to the metal again.
What convinced him to join wasn’t a polished pitch. It was the people.
He had worked with two of the co-founders at Quora and knew them to be exceptional engineers. One had joined Quora at 16 and was already a world-class competitive programmer.
“If someone like that wants to build something with you, you say yes.”
At the time, the team was experimenting with a “text-to-SQL” prototype — a tool that let users query a database using natural language. It was still a tech demo, but Henry saw something bigger: a universal interface for information.
He joined as the first and only designer, at a time when there were no product managers or front-end engineers. Just engineers, AI researchers, and Henry — writing production code and shaping the experience in real time.
Progress came through relentless, high-velocity iteration
Henry wrote and pushed code daily, sometimes multiple times a day. There were no mocks, no exploration phases, no layers of approval. The product was the prototype.
The turning point came when the team connected the backend to Bing’s search API. Suddenly, users could ask anything and get a real-time answer.
“It wasn’t about building a Slackbot or research tool anymore,” Henry said.
“It was obvious, this needed to be an iPhone app. This was a company.”
What made this breakthrough possible wasn’t just technical talent, it was the team’s freedom to move fast without waiting for permission. No stakeholder reviews. No alignment meetings. Henry had full autonomy to design, build, and ship.
Dogfooding was key
He credits their ability to evolve quickly to constantly using their own product.
“You can’t mock up generative software. You don’t know what it’s going to do until you ask it. So you have to build it, try it, break it, and build again.”
Perplexity found traction shortly before ChatGPT launched — not by chasing trends, but by spending six months in deep, intuitive exploration. They trusted their instincts, moved fast, and built in the dark.
That kind of environment isn’t for everyone.
But for a designer like Henry, it was exactly the challenge he was looking for.
You might also like
