How AI-Native Startups Actually Get Built

To me, a lot of conversations about AI startups sound strangely theoretical.

Founders want to talk about transforming industries. Investors love to describe massive platform shifts.

I’d rather dig into the messier parts where companies actually discover what changes their business.

That’s the thread that runs through this interview with True Ventures Managing Partner Puneet Agarwal and Mayank Mehta, CEO and co-founder of Gather.

Mayank’s a repeat founder; he built Pulse, which was acquired by Gartner. Prior to that, he was co-founder and head of product at Cooliris, which Yahoo acquired and folded into Yahoo Messenger.

Puneet invested in Pulse through True Ventures, so when Mayank was planning his next venture, he took an interest. “I wasn't thinking that much about the market,” he says. “I was actually just thrilled to be able to try to fund Mayunk again.”

Gather’s Original Thesis

Initially, Gather was called BeHeard Labs. Mayank and co-founder Anand Thaker were building a conversational AI platform for customer feedback with a simple thesis: traditional surveys are broken.

Speaking from experience, I only fill out surveys when I’m furious, ecstatic, or bribed with a coupon. Otherwise, I won’t take time to offer a company meaningful feedback. That’s a favor I reserve for friends.

Mayank believed AI could replace static surveys with real-time chats that used contextual follow-ups. The goal: elicit richer sentiment and deeper qualitative insights. The consumer side worked. In restaurants and retail environments, people scanned QR codes and started sharing all kinds of feedback they normally wouldn’t have bothered communicating.

Less friction, more feedback.

The companies receiving all of this new information weren’t logistically prepared to handle it, however.

“The assumption that they would find this invaluable from the get-go was something we still needed to go prove,” says Mayank. “So we ended up pivoting more into market research to go tell that story and break in.”

Not exactly a dramatic pivot. More like finding their actual wedge into the market.

As Puneet put it: “We’re funding at such an early stage, where it’s all about — nothing’s figured out.”

Building something that’s hard to explain?

That’s a good sign you’re working on something interesting — but it can also cost you fundraising, hiring, sales, and media opportunities.

I help early-stage founders sharpen the story around what they’re building: why it matters, who needs it, and why they’re the right team to build it.

If you’re preparing for a raise, launch, hiring, a panel, or an important customer conversation, I can help you pressure-test the narrative before the stakes get higher.

👉🏾 Learn more

The Part I Found Most Interesting

The strongest section of the episode comes later, when Mayank talks about rebuilding the company around AI-native workflows.

Their original outbound motion looked like fairly conventional SaaS:

cold emails
phone calls
manual personalization
traditional sequencing

It wasn’t working: A year of outbound efforts had produced roughly 10-15 meetings.

So over a weekend, the team built its own internal go-to-market engine:

ingesting contacts from Apollo
generating hyper-personalized outbound
creating multi-step sequences
auto-testing subject lines
self-correcting over time

“In the first two weeks of running our own growth engine, we had about 15 meetings and opportunities lined up,” says Mayank.

“AI isn't just SaaS plus,” says Puneet. “It is a whole new way to think about letting go of the ‘what’ or the ‘how’ and focusing on the ‘what’ and the delivery. And if you can build your company around that and really truly leverage that, I think there's a lot of potential there.”