Investors and the Superhuman Mail founder on how finding product-market fit has changed in the AI era

Product-market fit in the AI era: the new rules

Finding product-market fit — the moment when a startup's product begins to sell consistently — has become even harder in the AI era. Despite record-fast growth at many companies, customers don't always stick around for long, and products are easier to copy than ever. That was the topic discussed last week at TechCrunch Disrupt by Iconiq partner Anu Bharadwaj, Sapphire Ventures partner Rajeev Dham, and Superhuman Mail founder Rahul Vohra.

Pilots are easy, production is hard

“It has never been easier to get a lot of pilots, because many companies are open to trying new things,” said Bharadwaj. “But going from pilot to production — that's what's really hard.” She noted that founders often present revenue metrics to investors without clarifying what kind of revenue they mean: “Is it real ARR or run rate? There are many ways to muddy the waters.” Recently, discussions on X have focused on how AI founders inflate ARR (which can mean either “annual run rate” or the more reliable “annual recurring revenue”): a common trick is to include pilot customers who have been using the product for a couple of months in the annual run rate, creating the impression that the revenue will last all year.

Red flags and green flags

Bharadwaj stressed that it is in founders' interest to be honest about revenue quality. Red flags are when revenue depends heavily on a few customers, or when customers can easily switch to another provider for a better price: churn and retention depend on this. Vohra advises tracking customer engagement and revenue retention (dollar retention), especially for consumer products, and figuring out what counts as “good, great, and excellent” metrics in your field. At Superhuman, he said, in addition to ARR they tracked an internal metric called “delighted ARR” — revenue only from highly engaged customers; this helped them understand how “sticky” customers were and whether they would stay for the long term. “Those users are still users, those companies are still customers,” he said.

When to pivot and when to optimize

Vohra suggested talking to customers about the product's value and about how they would feel if they could no longer use it — and honestly assessing their answers. “If you're at 15%, 14%, 13% or below — consider a full pivot: take the codebase, the team, and build something entirely new. But if you're at 20%, you can quite plausibly get from 20% to 40% through optimization,” he said. Dham added that it's important for a founder not to confuse having a vision with having a good business. According to Bharadwaj, at the seed stage a pivot is acceptable — at that stage she isn't even looking for product-market fit, but rather assessing “whether the founder has a plausible hypothesis and the agility to maneuver as circumstances change.” Dham noted that right now he especially values founders with multiple pivots: it shows they've had to make very hard decisions in a constantly shifting moment. In his view, product-market fit has to be constantly re-earned: “Even with a billion in revenue, these companies are re-earning their product-market fit. You have to keep innovating in the product.”

Why it matters

The discussion is relevant to founders and investors in AI startups: according to the panelists, fast revenue growth no longer guarantees durability, and inflating ARR through pilot customers distorts the market picture. The practical takeaway from what was said is to separate hype from reality when assessing revenue quality, track retention and engagement, and at low metrics (below ~15% by Vohra's method) consider a full pivot instead of optimization. The editorial team notes that these are recommendations from the discussion participants, not verified data about specific companies.