I fed nearly 400 pieces of my own writing into a personal GPT I built. Articles, podcasts, frameworks, talks, posts collected over the years. I wanted to see how close an AI could get to thinking like me.
It got remarkably close on the surface.
It learned my tone quickly. It still cannot make one hard commercial decision the way I would.
→ It cannot feel that a deal is structured wrong before the numbers prove it → It cannot tell the difference between a customer who is testing us and one who is quietly using us → It does not know when to hold a hard internal conversation, and when to wait a week
None of that was in the 400 pieces. Because none of that ever made it to paper. That is what years of being in the room actually built.
Many companies are about to miss this.
AI can clone your output. It cannot clone your discernment.
The mid-market companies that understand this distinction will use AI properly. Volume, drafts, summarisation, first cuts, anywhere judgment is light and speed matters.
The ones that do not, will hand the wheel to a model that has never had to walk away from a client at 9 PM.
They will not get more productive, instead they will just automate their bad judgment faster.
If you trained an AI on your own thinking tomorrow, what would it still be unable to do?