Why I Stopped Using ChatGPT for Strategy

I admit it. I'm addicted to AI.

Like many of you, when GPT-4 first came out, I thought I'd never have to write a strategy document again. I fed it my messy notes, asked for a "SWOT Analysis," and watched the magic happen.

But last Tuesday, I was sitting in a board meeting, presenting one of those AI-generated strategies.

A board member asked a simple question: "Why do you think our Brand Loyalty is a Strength, given that our churn rate just doubled?"

I froze.

I checked my slides. There it was, under "Strengths": Strong Brand Loyalty.

Why was it there? Because the AI put it there. And I believed it.

That moment was my wake-up call. I realized that using generic AI tools was making me intellectually lazy.

The "Yes Man" Problem

The fundamental flaw of ChatGPT, Claude, and Gemini is that they are designed to please you. They are Sycophants.

If you tell them you have a great business idea, they will list 10 reasons why it will succeed. They won't look you in the eye and say, "This is delusional, and here is why you will go bankrupt in 6 months."

Real strategy requires friction. It requires someone (or something) to say "No."

The "Infinite Canvas" Trap (Miro & Jeda)

After my ChatGPT embarrassment, I tried the visual tools. Miro AI, Jeda.ai.

They are dazzling. You type a prompt, and poof — a canvas filled with sticky notes, arrows, and diagrams. It looks like work. It feels like productivity.

But zoom in.

It's just Confetti. It's 500 shallow ideas scattered on a digital whiteboard.

I spent 4 hours organizing the sticky notes, ensuring the colors matched, and aligning the boxes. I felt productive, but I hadn't made a single hard decision.

Finding the "Forcing Function"

I realized I didn't need a "generator." I needed a "constraint."

I needed a tool that would force me to answer the uncomfortable questions.

That's when I started looking for Structured AI tools. Tools that don't just "chat," but "interrogate."

Enter the Frameworks

This isn't just about SWOTPal (though I built it to solve this exact problem), but about a new class of tools emerging in 2025.

When I use tools like Perplexity for research, I don't ask it to "write a report." I ask it to find the data that contradicts my beliefs.

• "Show me 3 reasons why this market is shrinking."

• "Find me a competitor who failed doing exactly what I'm planning."

And when I use SWOTPal, I don't use it to "generate" the list. I use it to stress-test the connections.

The tool forces a TOWS Analysis:

• Okay, you say you have "Strong Engineering" (Strength) and "AI Trend" is an (Opportunity).

• The tool asks: "How exactly does your Engineering team capture that trend? Do they have the right skills?"

If I can't answer, the strategy fails.

The New Tech Stack for "Deep Work"

If you want to survive the AI age, you have to move up the value chain. You can't just be the person who prompts the AI. You have to be the person who judges the AI.

Here is my current "Anti-Yes-Man" stack:

1. The Cynic: Perplexity Pro. Use it to find facts that hurt your feelings.

2. The Architect: SWOTPal. Use it to force your messy thoughts into a logical rigid framework.

3. The Editor: Claude 3.5 Sonnet. Paste your final plan and give it the prompt: "Roast this strategy. Be ruthless."

Conclusion

Strategy is painful. If it feels easy, you're probably doing it wrong.

Don't let AI rob you of the struggle. The struggle is where the insight comes from. Use AI to make the struggle harder, not easier.