The AI Café Chronicles: what we learned about AI startups over coffee

Here we are again, sitting at my favorite café in Berkeley. It’s where so many of our best ideas and toughest conversations take shape, over cappuccinos and the hum of the city just behind the wall of this greenhouse-turned-café-patio. Today, it’s no different. We’re diving into the whirlwind of the past few months—demo days, late-night tech cocktails, and endless conversations with brilliant, passionate founders trying to change the world with AI.

This time, though, there’s something we can’t ignore. A pattern that keeps surfacing, like a refrain in a song: the incredible energy in the AI space right now, but also the red flags that keep tripping up founders on their path to success. We’ve spent hours unpacking what we saw and heard—trends in tech, the challenges of founder mental health (a huge one! many feel isolated, without a sounding board) and the recurring themes that surfaced in almost every pitch we heard.

And if there’s one thing we kept coming back to, it’s this: applied AI.

It’s everywhere.

Every conversation, almost every pitch.

But not all applications of AI are equal and the hard truth is, some startups are missing the mark.

So, here’s the thing: I want to share the feedback we’ve been giving to founders and some smart remarks we heard from investors and advisors. Not because we think we know it all (we definitely don’t!) but because these conversations sparked real insight and, frankly, a lot of gratitude. They helped founders see their blind spots, refine their vision, and grow.

This is for you, founders. Take what resonates, ignore what doesn’t, and—most importantly—keep building with courage and clarity. We love your courage, perseverance and passion.

Why some AI startups struggle to stand out

1. Your technical moat is a mirage

Let’s start with one of the hardest pills to swallow: whatever technical edge you think you have… it probably won’t last. Most startups overestimate how defensible their technology is. The reality? That shiny moat you built might only hold water for 3-6 months, tops.

Why? Because the enabling technologies you’re using—open-source libraries, public research papers, and even frameworks released by the big players like OpenAI, Meta, Anthropic or Google—are available to everyone else too. Competitors can replicate your “point solution” and integrate it as a feature. And let’s be honest, if Sam Altman tweets or Google drops a new product, your perceived lead could evaporate overnight. Harsh? Yes. True? Also yes.

2. It’s a bloody red ocean

AI isn’t just competitive—it’s overcrowded. Right now, there are nearly 20,000 AI projects floating around. That’s a staggering amount of noise. Remember when we all used to joke “There’s an app for that”? Well, now there’s an AI project for that—and probably a dozen others too.

The problem is that many of these startups are undifferentiated. They’re tackling problems in ways that are incremental at best. Founders often come to us with point solutions that might solve a small pain point, but they’re not addressing specific, underserved needs in specific, valuable markets. Worse, they’re vulnerable to being outpaced or absorbed by bigger players like OpenAI, Google, or Meta who can integrate their ideas as mere features in a larger product.

3. Your team might not be ready yet

Investors don’t just look at your product; they look at your people. And while many founders are brilliant and well-meaning, they’re not always building teams with the depth, diversity and agility to weather the storm.

Here’s what we mean.

Credibility matters. If your core team doesn’t have clear, demonstrable expertise in both tech AND the market you’re targeting, it’s a red flag. (Hint: advisors don’t count unless they’re total rockstars)

Diverse skill sets are non-negotiable. If all your co-founders have similar backgrounds, you’ll struggle to adapt when the market throws you a curveball. Complementary strengths are key.

Speed wins. You need a team that can not only build fast but iterate faster, especially when early users give you feedback. That ability to pivot is the lifeblood of a successful startup.

What does work: the strategies we noticed

Good news. Some AI startups are thriving, and here’s what they’re getting right.

1. Exclusive access to data

Data is the lifeblood of AI, but here’s the kicker: over 80% of the world’s data is locked away behind firewalls. Startups that secure exclusive, legal access to this data have a huge advantage.

The best founders don’t just talk about access; they show us contracts, agreements or partnerships that make it real. Even better? They have the technical smarts to extract value from that data—whether it’s through anonymization, synthetic data generation or building models that improve outcomes for every customer on their platform. Exclusive data isn’t just a nice-to-have; it’s a true moat that’s hard to replicate.

2. Product-led growth that feels organic

The most successful products don’t need flashy marketing—they grow because they resonate deeply with users. Founders who prioritize time-to-first-value (how quickly a user gets something meaningful out of the product) and design for viral effects – or at least some user's reason to involve or tell someone else – often see their growth take off. Think about products like DocuSign, Zoom or even Box: their users naturally bring in new users because the value is baked into the way they’re designed.

Want to know the secret sauce? These founders know their users intimately. They understand their daily struggles, their goals and the metrics they care about. And they weave that understanding into every part of their product. They have a map of their customer's job and know the desired outcomes for each step of it.

3. Network effects

Here’s where the magic happens: when every new user makes your product better for all the other users, you’ve hit on something special. That’s the power of network effects.

For example, marketplaces like Etsy or Uber thrive because every new user (whether buyer or seller) adds value to the platform. But even non-marketplace products can harness this: an AI financial platform, for instance, can improve its benchmarking and insights as more customers use it. Founders who build for network effects are playing the long game—and it pays off.

Build the flywheel

The real magic happens when you combine all three: exclusive data access, product-led growth and network effects. Together, they create a defensible flywheel that’s hard for competitors to replicate.

This isn’t just theory—it’s what separates the good from the great. Speed matters, yes, but so does the depth of your strategy. The tech moat might not last, but a well-designed flywheel will.

Build boldly and build wisely

We’re rooting for you, founders. The AI space is a jungle, no doubt, but it’s also full of possibility. If there’s one thing we’ve learned over countless coffees and late-night conversations, it’s this: the startups that succeed are the ones that combine speed with strategy, passion with pragmatism.

So, keep building. Keep listening to your users. Keep refining your vision. The future’s still unwritten—and you’re the ones writing it.

Ad maiora,
Pan & friends of Product Expanse