Last month I talked four founders out of AI projects. Not because AI is bad. Because timing matters.

Here’s why waiting is often the right call.

You Probably Don’t Have Product-Market Fit Yet

If you’re pre-PMF, your entire product might pivot tomorrow. Building AI features into something that might not exist in three months is waste.

I’ve watched founders spend $50K on AI integration for products they later abandoned. That money should have gone into customer discovery.

Get to PMF first. Then add AI.

Your Data Isn’t Ready

AI needs data. Good data. Lots of it.

Most early-stage startups have:

• Messy, inconsistent data
• Not enough volume for patterns
• No documentation of what things mean

Feeding garbage data into AI gives you garbage outputs, faster. Fix your data foundation first.

The Tools Get Better and Cheaper Every Month

GPT-4 cost $0.03 per 1K tokens at launch. Now it’s cheaper and there are good alternatives. Anthropic’s Claude didn’t exist two years ago. TechCrunch tracks the pricing wars closely.

Whatever you build today will be outdated soon. The smart play is often to wait until the tools mature, then build faster and cheaper.

Early adopters pay the innovation tax.

You Don’t Have the Expertise to Evaluate

If you can’t tell good AI from bad AI, you’re vulnerable. Agencies will oversell you. Employees will overpromise. You’ll approve projects that shouldn’t exist.

Spend time using AI tools yourself first. Understand what they can and can’t do. Then you can evaluate AI projects properly.

When TO Move on AI

Despite all this, sometimes you should move fast:

AI is core to your value proposition: If you’re building an AI product, obviously you need AI now.

Competitors have launched AI features: Customers directly comparing you. You need parity.

You have a specific, validated problem: Not “AI could be cool” but “customers are asking for X and AI solves it.”

You have the data and expertise: Ready to execute, not just experiment.

If none of these apply, waiting is fine.

What “Waiting” Should Look Like

Waiting doesn’t mean ignoring AI. It means:

1. Using AI personally: Get fluent with Claude/ChatGPT. Understand capabilities.
2. Collecting data thoughtfully: Build systems that capture clean data for future use.
3. Watching the market: Who’s shipping AI features? How are customers responding?
4. Experimenting cheaply: Small tests, not big projects.

When the time is right, you’ll move from a position of knowledge, not FOMO.

The Conversation I Have

Founder: “We need to add AI.”

Me: “Why?”

Founder: “Everyone else is. We’ll look behind.”

Me: “What specific problem does it solve for your customers?”

Founder: “Well… it would make us more innovative.”

That’s not a reason. That’s anxiety. AI shouldn’t be a response to fear of missing out.

The founders who win are building things customers want. If AI helps with that, great. If it doesn’t, skip it.

You don’t get points for having AI. You get points for solving problems.

Wait until AI solves a real problem for you. Then move decisively.