Sounds provocative. I mean it.
If you tried AI and thought "well, quite nice, but the hype is overblown" — that is almost certainly not your fault. And not AI's fault either. It is about what you were working with.
Around half the participants who come to my AI courses either use the free version of ChatGPT or whatever their employer provides. Company AIs so heavily restricted for privacy or cost reasons that barely anything of the actual performance remains. That is not a criticism of companies. It is simply economics.
What AI apps don't show you
Someone on Reddit recently calculated what it costs to offer a simple AI assistant as an app — with a reasonably capable model, reasonable usage limits, a price people would actually pay. The conclusion: the numbers don't work.
Every AI app you use for free or cheaply is not showing you what AI can do. It is showing you what a company can afford to offer you. That is a throttled version. A restricted one. One that is enough to create the impression that the tool works — but not good enough to show what is actually possible.
"The difference is not a nuance. It is enormous."
What most people see — and what lies beneath
- Free or cheap AI apps
- Company AI with privacy filters
- Throttled models with usage limits
- "Quite nice, but the hype is overblown"
- Direct access to the strongest available model
- No app layer, no filters, no throttling
- Complex reasoning, genuine problem solving
- Most common first reaction: irritation
I work daily with the most capable available models — no app layer, no company filters, no cost throttling. When I show course participants what that means in practice, the most common reaction is not excitement.
It is irritation. "Why doesn't this work like that at home?"
Because you are using a different tool. Not a different version of the same tool — a different tool. The performance gap between the best currently available model and what most people use daily is not gradual. There are entire performance classes in between.
What I wish for every course participant
I cannot and do not want to assume that someone comes to the course with a paid AI subscription. That would be unrealistic and unfair. But I try to convince everyone that:
Probably the best single investment in your own professional development right now
A decent AI subscription per month after the course. Not because I owe anything to any provider — but because I see what happens to people who work with the right tool after the course. And what happens to those who go back to the free version.
The difference shows quickly.
The real question
If you tested AI and were not impressed: do you actually know which model you were working with? Do you know how much it was throttled?
If not — maybe the comparison is worth making before you form a judgement.
How is it for you — are you using a restricted AI solution at work, and have you ever directly compared the difference?