Many people are commenting on the State of AI in Business 2025 from MIT, which states that 95% of attempts to make money or efficiency with AI models have failed.
I have insisted on this. AI is a powerful tool, but it doesn't work on its own, without human agency. It's not a silver bullet. It's not enough to give everyone ChatGPT, Gemini, Claude, etc. and that's it.
Here's what the report says: "...the main reason for the failures lies in the learning gap. This is because companies still don't know how to use the new tools well or how to take advantage of them in their work routine."
There's no way around it, you have to get your hands into the code and develop specific agents for each situation. And "no code" solutions are only delaying this. They give the feeling of speed in development, but it's chicken flight. They're only good for small projects and proofs of concept.
Also, check out this article about "Parallel AI", which is the use of AI assistants without formal authorization from the company. In other words, the company is failing in its AI projects, while its employees are using it "on the side".
There is no easy path, no silver bullet and no magic formula. We carry on as always:
- Choosing a relevant problem
- Clear definition
- Data quality
- Good people in development
That won't change.
