AI Startups Keep Forgetting This One Thing with Gemma Galdon Clavell

In AI, building great technology isn’t enough. You can solve a real problem and still struggle to gain adoption. Why? Because Product-Market Fit in AI isn’t just about function. It’s about trust.

That was the central theme in a recent conversation with Gemma Galdon-Clavell, founder of Eticas AI, and Collin Stewart. Their insights highlight why founders can’t rely on old playbooks:

Strong tech without trust still faces market reluctance.
Compliance doesn’t equal safety. Guardrails must be built in early.
In immature markets, founder-led trust-building comes before scale.
Referrals, not polite praise, are the real signal of PMF.
Even good VC advice can kill you if it’s mistimed.

The message is simple: in AI, true PMF is fit plus trust.

Product-Market Fit in AI Isn’t Binary

Most founders treat Product-Market Fit as a switch: either you have it, or you don’t. In AI, it’s not that simple. You can build brilliant technology, solve a real problem, and still hit a wall.

Gemma has seen this firsthand. Her company helps teams build trust and safety into AI systems from the ground up. And she’s clear on the problem:

“Developers often have zero data on the impact of their systems. It’s like being an engineer who designs a plane but never knows if it takes off or crashes.”

That gap creates market reluctance.

Customers don’t just need accuracy. They need confidence. They need to know your system won’t hallucinate, won’t create liability, and won’t expose them to risks they can’t explain. Without that clarity, even strong technology struggles to gain adoption.

As Gemma puts it, “PMF in AI is not just about working models. It’s about fit plus trust.”

Compliance ≠ Trust

It’s easy to confuse compliance with safety. Regulations are essential, but they were never designed to guarantee that your AI product will actually work in the messy, unpredictable conditions of the real world.

Checking boxes after launch doesn’t stop hallucinations, prevent bias, or protect you from liability. At best, it reduces legal exposure. At worst, it creates a false sense of security. Customers know this, which is why compliance alone rarely earns their trust.

Genuine trust comes from baking in guardrails early. That means testing how your system behaves not just in controlled environments, but in real-world scenarios where variables shift constantly:

How does your model behave with incomplete or noisy data?
What happens when minority user patterns are labeled as “errors”?
Can you explain why one output was produced instead of another?

The difference is subtle but critical:

Compliance looks backward, trust looks forward. Compliance checks whether you avoided mistakes. Trust proves to customers that your product will continue to work safely as conditions change.

For founders, this means treating trust and safety not as an afterthought, but as part of the product itself, as essential as usability or performance.

Market Timing Matters

Great technology isn’t enough if the market isn’t ready. In AI, this challenge is amplified: the technology is advancing faster than customer understanding. That creates a gap, not between supply and demand, but between capability and trust.

Founders often assume they can fill that gap by hiring a sales team and running a standard playbook. But immature markets don’t behave like mature ones. Prospects aren’t comparison-shopping between vendors; they’re trying to figure out what the product even is and whether they should adopt it at all.

That’s why founder-led sales isn’t just a scrappy early-stage tactic. It’s a necessity. Only the founder can:

Translate uncertainty into confidence. Buyers need hand-holding, not demos. They want to hear how others are approaching risk and what guardrails exist.
Adjust in real-time. In fast-moving spaces, customer objections shift weekly. A founder can reframe the product; a junior salesperson can’t.
Build the category. In an immature market, you’re not just selling your product. You’re selling the problem definition itself.

As Gemma put it when describing Eticas AI’s early sales struggles: “With an immature market, just cold emails didn’t work. We realized the trust building was really important. So we went back to founder-led sales.”

The deeper lesson:

Timing dictates tactics. Enter too early with a scaled sales motion, and you burn cash educating a market your reps aren’t equipped to guide. Enter too late, and competitors will have already defined the narrative. The founder’s job is to bridge that window, personally, until trust is strong enough to scale.

Referrals and Momentum Are the Real Signals

Early traction can be misleading. A prospect saying “this is interesting” or “great idea” feels validating, but it’s not Product-Market Fit. Praise without action is just politeness.

What actually signals PMF is momentum:

Customers are eager to use your product, even when it’s rough around the edges.
Teams rely on it daily because it solves a burning problem.
Most importantly, referrals come from customers introducing you to peers without being asked, or even before you finish the first engagement.

This is why Gemma emphasizes referrals as a core indicator for Eticas AI: if the loop breaks, it’s not fit yet. Customers might like the idea, but if they aren’t pulling others in, the problem isn’t urgent enough.

The difference looks like this:

Polite Praise: “This looks promising. Keep me updated.”
Real Pull: “You need to talk to my colleague about this. Can I introduce you?”

Founders often chase the wrong signals, sign-ups, demo requests, or nice feedback. But in AI, especially, where adoption is still full of fear and uncertainty, the only true proof of fit is whether people trust it enough to bring others along.

Momentum is the test. If you’re not getting it, you don’t have PMF yet.

Filter VC Advice Ruthlessly

Founders get flooded with advice from investors. Some of it’s gold, but even the best advice delivered at the wrong stage can wreck your company. Timing is everything.

Collin put it simply during the conversation:

“Great advice at the wrong time is bad advice.”

And he’s right. Advice like “hire salespeople” or “go product-first” might be perfect at Series A, but lethal at seed when the market is still immature and you’re still searching for PMF.

The danger is twofold:

Investors validate patterns, not customers. They’re drawing from what worked in other companies, in different markets, at different times.
Founders confuse credibility with relevance. Because the advice comes from someone who funded Uber or Stripe, it feels universal. It’s not.

The antidote is to treat VC advice as hypotheses, not instructions. Test it against customer conversations before acting. If customers aren’t pulling, no amount of hiring, tooling, or scaling will fix that.

This is especially true in AI, where markets are still defining themselves. Following investor pressure to scale prematurely doesn’t just waste money. It can lock you into a strategy the market isn’t ready for.

The deeper lesson: investors can help you think, but only customers can prove you’re right.

Conclusion

In AI, Product-Market Fit is never just about the tech. It’s about trust. The guardrails, momentum, and referrals that prove customers are ready to adopt. Scale too early, follow advice blindly, or rely on compliance alone, and you’ll stall.

The playbook is simple: bake in trust, lead sales as the founder, and let referrals prove fit.

To dive deeper into trust and safety in AI, check out Gemma Galdon-Clavell and the team at Eticas AI.

And if you want more conversations like this, listen to Collin Stewart on the Predictable Revenue Podcast.

NO TIME TO READ?