This week on Keep Going, I sat down with Richard White, founder and CEO of Fathom—the AI note-taker built for the flood of Zoom, Meet, and Teams meetings we all sit through. Fathom was early. It launched back in 2021 before the LLM boom, betting that AI would get better and note-taking would become table stakes for knowledge work. They were right—but it didn’t feel like that at first.

When they debuted on the Zoom app marketplace, they picked up nearly 100,000 signups in a month. Sounds like a dream, right? But that wave of attention turned into a near-death moment for the team. Out of those 100,000 users, just 100 were active. Total. For a while, it looked like they’d built something nobody wanted.

So what happened? And how did they fix it?

Lesson #1: Not all signups are equal.

It turned out most of those early users weren’t the right users. Many didn’t have a single meeting scheduled when they joined. COVID-era Zoom meant folks were using it to call their grandparents or attend virtual happy hours. Most of those who clicked “install” had no real use case. The team realized that their onboarding funnel didn’t filter for quality—and that was the problem.

Lesson #2: Filter hard, even if it shrinks your numbers.

They added friction. The signup flow got longer. They started blocking signups from personal emails, pushing low-quality users to waitlists, and asking for proof of usage—like syncing a work calendar or joining a test call. That worked. The active user count went up, not because of more traffic, but because the traffic was better.

Richard put it clearly: “I'd rather someone have no opinion than walk away thinking the product didn’t work for them.”

Lesson #3: Bad onboarding is worse than no onboarding.

The team also realized their onboarding process wasn’t helping people succeed. Instead of stripping everything down to make it easier, they actually added steps—but ordered them carefully. They drew inspiration from video games: start easy, build confidence, then ask for more. This gave users a sense of momentum and kept them engaged through the setup process. As a result, their activation rates climbed from 30% to 80%.

Lesson #4: Use the wrong users to get better.

Ironically, those low-value early users became a gift. They weren’t ideal customers, but their volume gave the team the data they needed to refine their flows. They learned how to stress-test the funnel and got better before the real flood of quality users arrived.

Lesson #5: AI is finally catching up to the vision.

Now, Fathom is doing what it always wanted to do. Its AI can summarize meetings better than most humans. Soon, it won’t just take notes—it’ll understand what’s happening across all your meetings. It’ll tell you who got upset, what topics keep coming up, and which conversations need follow-up. In other words, the AI is starting to understand context. And that’s the next frontier.

Richard’s vision is simple: “Let AI do the pre- and post-meeting work, and let you skip the meetings you don’t need to be in.” That means fewer meetings, better decisions, and searchable, usable knowledge across your company.

Keep Going Takeaway:
The lesson from Fathom is about the long game. It's not just about getting people to try your product. It's about getting the right people to try, and giving them the best possible shot at success. That takes friction, patience, and willingness to ignore the vanity metrics.

You can check out Fathom at fathom.video. No special links. Just five carefully designed steps.