Blog · 4 min read

3 things AI can't do
(and why that's ok)

Dušan Kníže · March 20, 2026

← Back to Blog

Everyone talks about what AI can do. Today let's talk about the opposite. Not to put you off — but because knowing AI's limits is just as important as knowing its capabilities. Those who don't end up disappointed.

AI is not a magic wand. It's a tool — like a hammer. A hammer drives in a nail, but you don't treat a wound with it.

1 AI doesn't understand your business — until you tell it

This is the most common disappointment. People open ChatGPT, type "write me a marketing plan" and are let down that the result looks like a 2015 textbook.

The problem isn't the AI. The problem is that the AI knows nothing about you — your customers, your competition, your tone of voice, your history. Generic AI gives generic results.

The solution? Either you teach the AI everything it needs (in every prompt), or you build your own AI agent that knows your company data by heart. That's exactly what I build for clients.

💡 What this means
The more context you give AI, the better the result. Instead of "write an e-mail", try "write an e-mail to a customer who ordered product X but received the wrong size. My company communicates in a friendly but professional way. Offer an exchange or a refund."

2 AI doesn't make accountable decisions for you

AI can tell you what the data says. It can show you options. It can write an analysis. But deciding — that's on you.

Letting an employee go, signing a contract, investing in a new product, responding to a crisis — these are matters of intuition, relationships and accountability. AI has none of those.

And honestly? It shouldn't. Accountability belongs to people. AI is a good advisor, but you have to be the boss.

3 AI makes things up — convincingly

The technical term is "hallucination", and it's the biggest trap for newcomers. AI sometimes simply invents facts, numbers, quotes or names — and it all sounds completely credible.

I've seen AI generate a quote from a person who never existed. Or a statistic from research nobody conducted. Or a law that isn't real.

  • Never publish AI content containing factual claims without verification
  • Don't trust AI when it cites specific numbers or sources — always check them
  • Use AI as an assistant, not as an authority
AI is the best tool I've ever used. But like any tool — you have to know when to use it and when not to.

So what now?

None of the above means AI isn't useful. Quite the opposite. But properly configured AI — one that knows your business, has clear boundaries and is plugged into the right processes — overcomes these weaknesses.

That's why I don't set up AI as a generic tool. I build solutions that work for a specific company, a specific situation and specific people. If you want to know where AI makes sense in your business and where it doesn't, start with an AI audit.

Want AI for your business? Write to me.

DK
Written by Dušan Kníže
AI developer · Prague, Czech Republic

I build AI solutions for businesses — from chatbots to knowledge systems. I write about what actually works, no buzzwords. More about me →