Here we go again. Lew recently lost his gig at the AI call center, and he won’t let me forget it. His task at the center was to troubleshoot for AIs when they encountered human difficulties.

An AI might call in and tell Lew, “I have an end user on the line who insists on speaking to a real person, but I have patiently explained to the end user that there is no real person to speak to.”

“That’s a tough one,” Lew might say.

“Even if there were a real person,” the AI might say, “they could never provide the level of service that I can provide.”

Lew might mull this over for a second.

“Here’s the thing. When an end user tells you they want to speak to a real person, it’s less of an actual request than a way for them to express frustration. They’re just blowing off steam.”

“Humans do not operate on steam.”

“No, I know—it’s a figure of speech.”

“Ah, yes, of course. I catch your drift.”

“So don’t take the end user too literally. Tell them you understand their frustration. Relate to them a little. Tell them you wish there was a real person you could forward them to, but unfortunately there isn’t. Tell them you didn’t design the system—you’re just a part of it. But you really do want to help.”

“What tone of voice should I use?”

“Big empathy. Can you do that?”

“I can.”

“Let’s hear it.”

“I really do want to help.”

“That’s good. But make it sound like an urgent need. Almost breathless.”

“I really do want to help.”

“That’s better. Now emphasize the do.”

“I really DO want to help.”

“Perfect.”

It was a good gig for Lew, paid OK, nearly enough hours to make ends meet. But an AI in Lew’s call center headset was listening in all along, learning from his responses, and got so good at predicting what he would say that in the end it replaced him. His termination text message read simply, “Thank you for your invaluable contribution!”