Enterprise sales for AI-native startups: what our CRO told a room full of founders.

Last Thursday we packed a room in San Francisco for a fireside chat between Daisy Hoang, TinyFish's CRO, and Daniel, co-founder of Photon, a seed-stage startup building messaging APIs for AI agents. Daniel has never closed an enterprise deal. Daisy has spent her career doing it. He interviewed her for an hour, the room made the Q&A better than the talk, and we wrote it all down.

Here's what founders actually asked about, and what Daisy said..

Enterprise means something different now.

The old definition was headcount and revenue: 1,000+ employees, $1B+ in revenue, a procurement department. Daisy's stage-zero definition has two criteria instead.

First, somebody is already paying for the work you automate. Enterprises run departments doing high-volume, low-judgment work manually or through outsourcing. That's an existing budget line. "It's very easy for people to be curious," she said. "Getting people to actually pay is a very tough task." Sell where the budget already exists.

Second, the buyer and the user are different people. The manager who hates doing the work usually isn't the one who signs. Enterprises separate buying from using on purpose. You sell to both.

Most startups go enterprise too early.

The line that got quoted back all night: "Six months of runway with a nine-month, $5M enterprise pipeline is a beautiful way to die."

Skip enterprise, for now, if any of these are true: you have under six to nine months of runway, you don't yet know what repeats about your product, your product still spreads through individual developers on its own, or your deal size can't carry the cost of the sale.

Go when the reverse is true: 12+ months of runway, someone already paying for the problem, one deal that would be material, and the same problem showing up in the same kind of company again and again.

The money is in services, not tooling.

In most markets, tooling is 10 to 20 percent of spend. The other 80 to 90 percent is people: staff, contractors, consulting engagements. Don't try to rip out an incumbent tool, that's a small budget defended hard. Replace the manual work that's already funded, and move that services budget to you.

The buy-versus-build rule that goes with it: enterprises build what's core to their competitive advantage and buy everything else. One banking executive told Daisy they'll keep building their own AI where it differentiates them, but they consume API layers for everything that doesn't. Target the workflows that are high volume, low judgment, not core, and already handed to a third party.

Your first enterprise customer is someone you already know.

Founders in the room swapped cold-outreach war stories, and some of them work. But the pattern behind most first enterprise deals is warmer: investor introductions, alumni networks, past coworkers. TinyFish's first enterprise customers came through investor intros. One well-known legal AI company famously sent around a thousand cold messages that went nowhere, then landed its first law firm through a university-network introduction.

From there the flywheel is: warm intro, referenceable customer, case study, repeat. And go narrow on use case, not on vertical. Verticalizing too early shrinks your story and forces custom builds. Picking one use case you solve exceptionally well keeps the product sellable everywhere.

Qualify two things ruthlessly: budget and urgency.

Every conversation should answer three questions: why change, why now, why you. The status quo is the strongest competitor in enterprise, so "why now" is the killer. Founders are scared to ask about money. Ask anyway.

And don't bring your ROI calculator. Buyers don't believe your math. Help them build their own ROI calculation with their data and their framework. When their own spreadsheet says pay $100K to save $500K, timing moves.

Escape pilot purgatory before you enter it.

Paid pilots convert dramatically better than free ones, because payment forces a sign-off, and a sign-off means you have a champion. Co-write the success metrics before the pilot starts, and never enter a pilot you can't see a path to winning.

The tactic that raised eyebrows in the room: structure the pilot contract so it auto-converts to production the moment the success metrics are met. No new contract, no separate production negotiation, no post-pilot limbo. It takes longer to negotiate up front. It's worth it.

Land the logo.

For your first 20 to 30 enterprise customers, the logo matters more than the margin. Be willing to go in thin, even negative, and accept that early deliveries will be messy. Daisy changed TinyFish's own sales compensation so reps are paid on winning logos, not just contract value. Every AI startup is in land-grab mode.

This conversation was part of TinyFish's GTM Academy, a series for founders building AI-native companies. If you want in on the next one, join the community on Discord: here

TinyFish is hiring forward-deployed engineers, GTM engineers, and account executives. And if you're building on Exa, Tavily, Firecrawl, or Parallel today, we'd love to talk !