Why would anyone with a ChatGPT subscription pay for a product that generates forms?
Jay Choi, CEO of Typeform, joined The Product Podcast to walk through how he’s answering that question:
The tension Typeform faces isn’t unique to them, so here are 5 lessons for AI Builders from Jay:
Defense before offense
“For our traditional forms business, we think about it as, how can we be defensive? And so we really put it on ourselves to create that same conversational interface into our core forms business.”
The first move Typeform made in response to the AI threat was to harden the core. The defensive thesis was simple: if someone could replicate your product’s basic functionality in a chat window, you needed to bring that same conversational interface inside your own product first. Typeform built an AI form builder that pulls in a URL, draws on millions of data points, and generates a contextualised form from a plain-language prompt.
💡 Before committing resources to AI-powered expansion, audit whether your existing product is defensible against substitution. Offense built on a leaking foundation doesn’t compound.
The happy churner problem
“We call them happy churners. People are not mad at the platform. They just use it, and then they are done with it, and then three months later, they use it again, and then they turn it off.”
Traditionally, churn users are seen as dissatisfied, but Jay introduced the concept of “happy churners” – users who weren’t unhappy with the product, they just didn’t need it anymore. They’d complete a project, cancel, come back three months later, and repeat the cycle. The insight was that churn here wasn’t a product quality problem but a symptom of the use case depth problem.
Typeform wasn’t desperate for better onboarding, a cheaper plan, or better customer service. They needed to find more applications beyond their traditional product.
💡 If your retention data shows low-friction cancellations, look at use case depth before you look at engagement mechanics. Users who leave without complaint are telling you the product solved a one-time problem. Do more research to see if it has the potential to solve a recurring one.
Triangulating on use case bets
“Research and growth tended to overlap, and so we did a bit of triangulation there, where we could get a lot of product depth with similar investments across our platform.”
Typeform narrowed from three possible vertical bets – growth, research, and talent – to two, with feature overlap being the deciding factor.
Growth and Research flows shared infrastructure requirements. Growth and Talent did not – talent required anonymity architecture, while growth required enrichment and automation. Building both simultaneously would have meant divergent platform investment with compounding complexity.
💡 When evaluating which vertical bets to prioritise, map the underlying feature requirements before modelling the revenue opportunity. Overlapping infrastructure is a multiplier; diverging infrastructure is a tax.
Breadth as an accidental moat
“The farther we go along the product breadth, it just seems to touch more surface area that feels more challenging and not worth it to replicate. Breadth matters now more than it did before.”
Jay’s reframing of product breadth is counterintuitive.
Conventional product strategy pushes toward depth – go narrow, go deep, own one thing completely.
Jay disagrees. In a world where simple, discrete use cases can be vibe-coded out of existence, the breadth of workflow coverage becomes the moat. The more surface area a product touches – form, enrichment, lead scoring, automated follow-up, CRM integration – the less worth it the alternative becomes.
💡 Re-evaluate whether your “go deep” strategy still holds in an AI-native competitive environment. End-to-end workflow ownership may now be more defensible than single-feature excellence.
Model agnosticism as product strategy
“We want to be model agnostic so we can, one, always choose the best model, and then two, some models are better for some use cases within our own product, and we don’t have to all-or-nothing our entire platform on any particular use case.”
Typeform made an early commitment to being model agnostic – and it wasn’t primarily a technical decision, but an important strategic move.
The reason is simple: the pace of model releases makes any fixed bet a liability. Different models already serve different use cases within the Typeform product. And without an AI observability platform, you can’t even answer the question of whether switching models actually improved anything.
💡 Treat your AI model strategy as a product strategy decision. Build for optionality, instrument for evaluation, and resist the pressure to consolidate on a single model before the market has matured.
Key Takeaway
Jay described a kind of strategic honesty that's harder to practice than it sounds. Typeform started by acknowledging they wouldn't win every use case – and built from there.
That shaped every decision downstream: which verticals to bet on, how to position AI without confusing customers, how to price closer to value without overcomplicating the model.
For product leaders navigating the same pressures, that starting point might be the most transferable thing from this conversation.
