Her fear was specific: "What if the AI goes rogue?"
My instinct as a technician kicked in. I wanted to tell her about Private LLM Instances. I wanted to explain that we don't use the public ChatGPT, that her data doesn't train the model, and that everything stays within her Azure tenant.
I started to say, "Marta, technically we use isolated instances that..."
She cut me off with a wave of her hand.
"Mario, I don't care about instances. I care about a supplier getting paid twice because the machine got confused. I care about my reputation if we leak data."
Logic doesn't cure fear. Especially when that fear is based on loss of control.
Marta didn't need specs. She needed to understand who was actually in charge. She thought the AI was the Captain of the ship.
I had to change the narrative completely. I had to use a horror story she knew well to explain the solution.
"Marta," I said, putting my laptop away. "Do you remember the fraud from last month?"
Her face changed. I had touched a nerve.
(To be continued tomorrow...)
