I sat through twelve pitch decks last month at a startup event in Sydney. Nine of them mentioned AI. Of those nine, I’d say two had a genuine AI component. The other seven were using AI as a magic word — a pixie dust they sprinkled on their slides to make a regular software product sound like a moonshot.

I’m tired of it. And more importantly, investors are tired of it. The AI hype window is closing. In 2023 and 2024, you could stick “AI-powered” on a pitch deck and get a meeting. In 2026, it’s become a red flag.

The tells

After sitting through hundreds of pitches — giving them, watching them, and occasionally funding them — I’ve developed a finely tuned BS detector for AI claims. Here are the tells:

“Our proprietary AI algorithm.” Unless you have a team with ML research backgrounds and published papers, you don’t have a proprietary algorithm. You have a prompt wrapped around ChatGPT’s API. That’s fine! API-based products are legitimate businesses. But calling it proprietary is dishonest, and any investor who does basic technical due diligence will find out.

“AI-powered” with no technical slide. If your pitch deck uses “AI-powered” as an adjective but doesn’t include a single slide explaining what the AI actually does — what model, what data, what task — you’re treating AI as marketing copy, not technology.

The architecture diagram with “AI/ML” as a black box. You know the one. Beautiful system architecture, clean arrows, and right in the middle a box labelled “AI/ML Engine” with no detail inside. What model? What’s the input? What’s the output? How was it trained? If you can’t answer these questions in two sentences, you don’t understand your own product.

Metrics without baselines. “Our AI improves customer retention by 30%.” Compared to what? Compared to a random number generator? Compared to a simple rules-based system that took two hours to build? Without a baseline comparison, the number is meaningless.

“We’ll add AI in V2.” I’ve heard this one dozens of times. “The current product is a manual workflow tool, but in V2 we’ll add AI to automate it.” Translation: we don’t have AI. We have a plan to have AI. That’s not the same thing.

Why investors have stopped buying it

The smart money has been burned. The AI startup cohort of 2023-2024 raised enormous amounts and a significant percentage have failed to deliver on their technical promises. Investors who wrote cheques based on AI buzzwords are now writing cheques based on AI evidence.

Here’s what early-stage investors I know are now asking:

• “Show me the product without the AI. Does it still have value?” If yes, the AI is a genuine enhancement. If no, you’re completely dependent on third-party AI providers, which is a major risk.
• “What’s your AI cost per transaction/user/query?” This question separates founders who understand their economics from founders who are living in the free-tier fantasy.
• “What happens when your AI provider changes their pricing or terms of service?” If you don’t have an answer, your entire business sits on someone else’s platform decisions.
• “Can I talk to your technical lead about the implementation?” If this question makes you nervous, that tells investors everything they need to know.

A First Round Capital survey from late 2025 found that 72% of venture investors now consider “AI-washed” pitch decks — those that overstate AI capabilities — to be a negative signal. Not neutral. Negative. It signals either dishonesty or technical naivety, neither of which inspires confidence.

What a good AI pitch actually looks like

I’ve given pitches that worked, and I’ve given pitches that flopped. The difference isn’t polish — it’s substance. Here’s what an honest, compelling AI pitch includes:

A clear problem statement that doesn’t require AI. Start with the pain. “Businesses waste 20 hours a week manually categorising support tickets.” That’s a real problem with a real cost. Now you’ve got the investor’s attention, and AI becomes the solution method, not the solution itself.

Specific technical choices with reasons. “We use GPT-4o for complex ticket classification and a fine-tuned Llama model for routine categorisation. The dual-model approach reduces our API costs by 55% while maintaining 94% accuracy on our benchmark dataset.” That’s one slide. It shows you’ve made thoughtful decisions and understand the trade-offs.

Honest limitations. “Our system handles English-language tickets well. Multi-language support is on the roadmap for Q3 and requires fine-tuning on non-English datasets we’re currently collecting.” Admitting what you can’t do yet builds more credibility than pretending you can do everything.

A competitive moat that isn’t “AI.” AI is a tool, not a moat. Everyone has access to the same models. Your moat is proprietary data, domain expertise, customer relationships, speed of iteration, or workflow integration. If your only moat is “we use AI,” you don’t have a moat.

Cost modelling that scales. Show investors you’ve thought about what happens at 10x and 100x your current scale. If your margins go negative at 10,000 users, that’s a fundamental problem that more funding won’t fix.

The pitch deck I wish I saw more often

Here’s my dream pitch structure for an AI startup:

1. Problem: Clear, quantified pain point
2. Solution: How you solve it, with AI as one component
3. Product demo: Show, don’t tell. A 60-second video of the product working
4. How the AI works: One slide. Technical choices, data strategy, accuracy metrics
5. Business model: Unit economics including AI costs
6. Traction: Users, revenue, retention — the standard stuff
7. Team: Why you’re the right people, including relevant technical background
8. Market: Size, growth, competitive landscape
9. Ask: What you need and what you’ll do with it

Notice that AI appears in exactly two slides: the solution and the technical detail. It’s not on every page. It’s not in the company name. It’s a tool in service of a business, not the business itself.

A plea to fellow founders

Stop leading with AI. Lead with the problem you solve and the value you create. If AI is genuinely part of how you deliver that value, explain it clearly and honestly. If it’s not, leave it out. Nobody is going to fund you because you used the word AI seventeen times. They’re going to fund you because you demonstrated a clear understanding of a problem, a credible solution, and a team capable of executing.

The founders who are going to win in 2026 and beyond aren’t the ones with the most impressive-sounding AI claims. They’re the ones who build products that work, economics that scale, and narratives that are honest.

Be that founder. Your pitch deck — and your business — will be better for it.