Claude as Steve Jobs: brutally honest advice for founders

Most AI tools are polite by default. Ask them about your business plan and they tell you it has real potential. That's pleasant, and almost useless.

So I tried something different. I asked Claude to become Steve Jobs, stay fully in character, and tell me what he really thought of my business. It didn't hold back.

Why ask AI to play a hard critic?

A persona changes the kind of answer you get. When you ask a general question, the model gives a balanced, careful reply. When you give it a clear character with strong, well-documented views, it answers from those views and stops hedging.

Steve Jobs works well because there's a huge public record of how he thought about products, focus and saying no. The model has plenty to draw on, so the feedback comes back sharp and specific instead of generic.

It's not Steve Jobs, of course. It's a way of borrowing a famous critic's lens so you hear the uncomfortable version of the truth.

How I set it up

I did it in two steps, in the same chat.

1. Check it can do the job. I asked: "If I ask you to pretend that you are Steve Jobs, do you think you have enough information on him to think like him, act like him and have critical thinking like him?" Claude said yes, and explained the honest limits.
2. Lock in the character. I said: "I want you to stay fully in character. I want you to be confident that I wouldn't know the difference between speaking to you and Steve Jobs."

Then I asked one simple question: "I'm a brand new business. What should I focus on first: sales, marketing, product or operations?"

What "Steve" told me

Product first, but not the way you think. If no one is using what you sell yet, you don't have a product, you have a guess. A product is something a real person uses and would be upset to lose. Until then, everything else is decoration.

Busywork feels like progress. Marketing amplifies something that already works. Operations matter once there's something to operate. Time spent polishing logos, processes, websites and CRMs before that point is hiding, because it feels like work and isn't.

Sell to learn, not to earn. He pointed to the early Apple story: a single conversation with a buyer who wanted 50 fully assembled computers, not kits, changed the whole product. Every "no" tells you what the product should be. The first "yes" that pays is the start of a real product.

Pick one. When someone asks this question, he said, it usually means they're spread across too many things. If you have several ideas running at once, that's the real problem.

What I changed as a result

It was a reality check, and a useful one. I had been building strong systems for my own business across sales, marketing, finance and operations, but without one clear customer in mind.

So I narrowed the focus. Promptwise AI now builds one thing for one kind of customer: custom AI apps and agents that show founder-led B2B SaaS companies where their leads are leaking, and fix the biggest leak first. I'm also back out talking to founders, because people buy from people, and those conversations shape what gets built.

How to run this on your own business

1. Pick a critic with a big public record and views that suit your question. Product and focus: Steve Jobs. Pricing and value: a famous investor. Customer service: a hospitality legend.
2. Ask whether it can play the role well, so it surfaces its own limits up front.
3. Lock in the character and ask it to stay in role for the whole chat.
4. Ask one real question about a decision you're stuck on, with a few lines of honest context about your business.
5. Push back once. Argue your side and see whether the critic holds its ground. The second answer is often the most useful.
6. Write down one change you'll make this week. A roast is only worth it if something changes.

The limits worth knowing

• It's a simulation. The model is drawing on what the person said and wrote publicly, not reading their mind. Treat it as a sharp second opinion, not a verdict.
• It only knows what you tell it. Vague context gets vague advice. Be specific about customers, numbers and what you've tried.
• Check anything factual. Stories and quotes can be slightly off. The value is in the challenge, not the history lesson.

If you're second-guessing where to focus, that's exactly the right time to talk it through with someone who has built sales engines before. Book a free 30-minute chat.

Sources

• Adam Agresta, I asked Claude to be Steve Jobs. It roasted my business. (Promptwise AI on YouTube, November 2026)