AI Startups Must Re-Find PMF Every 3 Months

Elena Verna, Head of Growth at Lovable, has spent 20 years leading growth at Miro, Dropbox, and Amplitude. She estimates that only 30-40% of what she learned over those two decades still works today. That figure deserves more attention than it usually gets.

Anyone running an AI startup has probably felt the ground shift, but her interview puts a concrete number on it and traces why.

Why Traditional PMF No Longer Holds

In the old model, finding product-market fit meant you could scale for years on that foundation. AI moves fast enough that customer expectations reset every three months. Companies hitting $200M ARR are still having to prove themselves from scratch each cycle, and there is no sign that pace will slow.

The most counterintuitive takeaway from her interview: funnel optimization contributes roughly 5% of growth. The remaining 95% comes from shipping new features. When the whole category is moving rapidly, obsessing over micro-optimizations in the funnel is a way to lose ground while feeling productive. The teams that do not understand this tend to find out the hard way, usually after spending months on conversion rate work that doesn’t move the needle.

What Lovable Does Instead

Engineers announce their own updates directly. The founder posts progress on Twitter every single day. The product feels alive, customers stay curious, and competitors are left reacting. The constant public shipping is itself a form of market positioning.

Lovable also absorbs the cost of free distribution aggressively. AI products are expensive to run, but they fund hackathons, sponsor events, and hand out credits, treating that spend as marketing rather than cost. Word-of-mouth compounds over time in a way that paid acquisition does not. Their Discord has hundreds of thousands of members helping each other, which amplifies word-of-mouth, improves retention, and creates a sense of belonging that is hard for a competitor to replicate quickly.

On the distribution side, short videos showing what the product can actually do outperform paid ads by roughly 10x in Lovable’s experience. I’ve found the same pattern holds: showing beats explaining every time.

What to Build Toward

There will be inflection points, moments when model performance leaps or pricing collapses dramatically. Those moments reshape markets faster than any growth campaign can. The teams that are positioned to move when the window opens, with capital secured in advance and distribution channels already warm, will capture disproportionate value. The teams still trying to optimize their funnel when the inflection hits will find themselves a step behind with no obvious way to catch up.

• Elena Verna Interview