Most AI startups selling to enterprise don’t die because the tech is bad.

They die because nobody buys it. CB Insights looked at 431 VC-backed startups that shut down since 2023. Poor product market fit was a cause for 43% of them. And 20 of those were Series B or later. They raised, they had traction, and still.

Over the last 3 years at my studio I worked on exactly this with Digimarc, Almetra AI, MarvelX and Siktio. Some reached PMF, some got a lot closer. And to be honest, it’s always the same playbook.

Almetra just raised a $19M Series A, with Bosch and Siemens Energy as customers. Watching that happen after working on their stack was a good week.

So here’s the full thing. Roughly 24 weeks.

Quick disclaimer: this is for startups selling to enterprise. Long sales cycles, buying committees, security reviews. It’s its own game.

Weeks 1-4: Sales calls disguised as research

Book 20 calls with people who match your ICP. The ask is simple: “we built this tool, you are exactly who we built it for, can we get 30 min to understand your pain points?”

You ask about all their pains. Then at the end: “wanna see what we built?”

Why this works:

• It doesn’t feel like a sales call, so they actually talk
• You show you care about their problem first
• Worst case you get great input. Best case you’ve got a buyer.

And the thing is, you don’t get much time with enterprise buyers anyway. Gartner found B2B buyers spend only 17% of their buying time meeting suppliers, and that’s split across every vendor they look at. Getting 30 minutes where they talk and you listen is huge.

Steve Blank said it best: “There are no facts inside your building, so get the hell outside.”

Weeks 5-12: Test features before you build them

Your engineers should build the core platform. But don’t let them build features you haven’t validated yet.

Pendo looked at usage across 615 software products and found 80% of features are rarely or never used. That’s a lot of engineering salary going nowhere.

So test first. I build realistic prototypes with Replit and Claude Code, so founders and sales reps can show a personalised demo to every new lead. They walk into the call with something that looks like it was built for that one company.

User testing and validation on steroids. And if nobody cares about a feature, you lost a week. Which is fine.

Weeks 13-24: Look like a safe bet

Showing a cool feature is nice. But the enterprise buyer is asking a different question: will this blow up in my face?

I learned that as co-founder of Felloz. We sold to 100+ big organisations, and what closed deals was mostly that we didn’t look like a risk.

It got harder since then. MIT’s research says 95% of enterprise gen AI pilots fail. Your buyer read that too. They’re checking if you’re part of the 95%.

During a 6-month sales cycle you need:

• A website that shows you are real (and not generated in 20 minutes)
• A website and sales deck that handle objections before they come up
• Social graphics and print materials that look the part

For Almetra AI I designed a full booth for a convention. For Altor Partners I designed their LinkedIn graphics, and the same company looked way more professional in a matter of minutes.

What comes after

After those three steps, it comes down to a successful pilot. That’s a topic for another issue.

Still reading?

Then you probably just closed your round. Now investors want to see the product actually works, and you don’t know where to start.

That’s ok. We’ll figure it out together.

One question: what’s the objection that keeps coming up in your enterprise deals? Just hit reply, one sentence is enough. I read every one.

​Book a call with me here​

Nik
​Ex-founder, 8+ years building products from 0→1