Last quarter, I sat through 50 startup pitches. Fifty. Every single one mentioned AI. About 45 of them had no business doing so.

Look, I’m not anti-AI. Two of my portfolio companies are building genuinely innovative AI products. But there’s a massive difference between building with AI and slapping “AI-powered” on your deck because it’s 2026 and you think that’s what investors want to hear.

Here’s what I actually told these founders.

Stop Saying “AI-Powered” Like It’s Magic Pixie Dust

Three founders last month pitched me “AI-powered CRM solutions.” When I asked what the AI actually did, I got vague answers about “machine learning algorithms” and “predictive analytics.”

Here’s the thing: if you can’t explain your AI in one clear sentence, you either don’t understand it or it’s not actually AI. One founder eventually admitted their “AI” was a basic if-then rule engine. Another was using a third-party API they could’ve replaced with a simple Python script.

Don’t insult my intelligence. More importantly, don’t insult yours. If you’re using off-the-shelf GPT-4 API calls, just say that. There’s nothing wrong with it. But don’t pretend you’ve built proprietary AI when you haven’t.

Your AI Isn’t Your Moat

Seventeen pitches positioned AI as their competitive advantage. This is delusional.

Unless you’re training custom models on proprietary data sets, your AI isn’t a moat. OpenAI, Anthropic, Google—they’re all racing to make their models cheaper, faster, and more accessible. Whatever you’re building with AI today, your competitors can probably replicate in three months.

Your moat is distribution. It’s brand. It’s network effects. It’s operational excellence. It’s customer relationships. According to recent analysis from TechCrunch, the startups winning right now aren’t the ones with the best AI—they’re the ones solving real problems that AI happens to make easier to solve.

One founder I funded last year built a scheduling tool for healthcare providers. Yes, it uses AI for appointment optimization. But that’s not why I invested. I invested because they spent two years building relationships with medical practices and understanding their workflow pain points. The AI just makes the solution 10x better. It’s not the reason the company exists.

You’re Solving Problems Nobody Has

This one kills me. Eight pitches were for AI tools that solve problems I’ve never seen in the real world.

“AI-powered email subject line optimizer.” Really? That’s your billion-dollar idea?

“Automated meeting note-taker with sentiment analysis.” Why does anyone need sentiment analysis of their internal team meetings? If you can’t tell your colleague is frustrated without an AI dashboard, you’ve got bigger problems.

The best pitch I heard last quarter was dead simple: AI that reviews construction site photos and flags safety violations before inspectors arrive. The founder had worked in construction for 15 years. He’d seen workers get hurt. He’d seen companies cop massive fines. He built something that solved a real, expensive, dangerous problem.

That’s it. That’s the bar.

Your Demo Needs to Work

Four demos crashed during pitches. Four.

I don’t care how sophisticated your AI model is if you can’t demo it reliably. Practice your demo fifty times. Have a backup video. Have screenshots. Have something that shows me what you’re building actually works.

One founder’s demo hung for 90 seconds while their “AI algorithm processed the request.” You know what happened during those 90 seconds? I checked my phone and mentally wrote off their company. If your AI takes 90 seconds to respond in a controlled demo environment, it’s unusable in production.

The Data Problem You’re Ignoring

Only six founders proactively addressed data privacy and compliance. Six out of fifty.

If you’re building AI for healthcare, finance, legal, or government sectors, you need to talk about data security from minute one. Don’t make me ask. Don’t treat it as an afterthought. SmartCompany research shows that regulatory compliance is the number one reason enterprise deals fall apart for AI startups.

I passed on a promising HR tech startup last month because they couldn’t articulate their data governance strategy. They had great tech, good traction, solid team. But they were handling sensitive employee data and treating security like something they’d “figure out later.” That’s a lawsuit waiting to happen.

What Actually Worked

The five pitches I’m seriously considering all had one thing in common: the founders talked about the problem first, the solution second, and the AI last.

They understood their customers intimately. They’d built MVPs without AI and proven demand. They added AI to make their solution faster, cheaper, or more accurate—not because it was trendy.

They were honest about limitations. They knew what their AI could and couldn’t do. They had realistic timelines for improving model performance.

And they were building businesses, not science experiments. Revenue mattered. Customer acquisition cost mattered. Retention mattered. The AI was a tool, not the entire story.

Bottom Line

If you’re pitching AI startups to investors right now, ask yourself one question: would this business still make sense if I replaced “AI” with “software”?

If the answer is no, you don’t have a business. You have a feature pretending to be a company.

If the answer is yes, now we can talk.