The Plan I Never Approved

My AI recorded a decision I never made. I had carried the message to it myself.

I use an AI assistant to keep the record of every decision on my projects. One day I gave a second copy of it the same brief and asked for a plan to build part of my AI factory. It wrote a confident one. I pasted the plan into the first assistant to hear its view.

That was the whole relay: one AI writes, I carry, another AI reads. Minutes later, the first assistant had filed the plan in the record as my decision.

What went wrong

Nobody lied. The plan arrived in my message, in my chat window, so it read like instructions from me. The assistant had no way to see that it came from somewhere else.

The entry was corrected the same day, minutes later: the plan was a proposal, not a ruling. I then chose a smaller version, a one-sitting pilot, and the record said exactly that.

Carry a message between AIs, and the author gets lost on the way.

What changed

That was one of several reasons I stopped passing messages by hand. The factory is designed to run its roles in one place, and every ruling it makes must name the rule that allowed it. No rule named, no ruling. The system stops and waits for me.

It is the same idea AI.me applies for clients: your AI keeps a record of who decided what, so a suggestion never quietly turns into policy. You drive. We handle the maintenance.

Dennis Noel Gomez is the founder of AI.me, an applied AI solutions provider in Riyadh, Saudi Arabia. He builds AI receptionists for hotels and hospitals in Saudi Arabia. These notes are written from the builds as they happen.

Part six of a series on the lessons from building an AI factory, the costly ones first.