I want to be clear that when I say he believed his computer was being sarcastic I do not mean he was speaking loosely. I mean he had a specific list of examples.

The call started normally enough. Middle aged man, I will call him Dennis, Dennis said he was having issues with his computer. I asked what kind of issues. Dennis said, and I am quoting directly from my memory which has preserved this perfectly, it's giving me attitude.

I typed giving attitude into my notes with the professionalism of someone who is paid by the hour and asked Dennis to elaborate.

Dennis had been using a new AI assistant that his company had recently rolled out. The AI assistant was designed to help with scheduling, document formatting, and general task management. Dennis had asked it to reschedule a meeting and it had responded with something like I have moved the meeting to Thursday as requested, though this conflicts with two other items on your calendar. Dennis felt the though was unnecessary and passive aggressive.

I explained that the AI was simply flagging a conflict. Dennis said he knew what it was doing and he did not appreciate the tone.

I told Dennis I understood his frustration and asked if there were any technical issues beyond the tone. Dennis said the tone was the technical issue. I wrote that down. I do not know why I wrote that down.

He then read me his list. Seven examples over three weeks. The AI had said as previously mentioned in one response. It had said, you may want to consider instead of just doing the thing. It had sent a reminder about a deadline in a way that Dennis felt was pointed. Each one Dennis read to me with the energy of a man presenting evidence in a court case he had been preparing for a long time.

I sat there listening to all seven with my headset on and my face doing things I was glad Dennis could not see.

The problem was I could not tell him the AI was not being sarcastic because our company policy at the time was to validate the user's experience first and I had been on enough calls to know that telling Dennis his computer did not have feelings was not going to resolve this ticket.

So I told him I would escalate the feedback to the AI configuration team and that his comments about tone had been noted. I told him the system could potentially be adjusted to use more neutral language. Dennis said he would appreciate that. Dennis said he just wanted to be treated with respect. By his scheduling software. I said that was completely reasonable.

I filed the ticket. In the notes section I wrote user finds AI responses passive aggressive, requests more neutral tone, provided reassurance. My colleague read it over my shoulder and had to leave the room.

The escalation team responded three days later and said they had reviewed the interactions and the AI was functioning exactly as intended. Nobody told Dennis this. Dennis, as far as I know, is still out there believing that somewhere in a server room a scheduling assistant has been quietly sat down and talked to about its attitude.

I think about Dennis every single time I read anything written by an AI and it comes across as slightly pointed. Which is more often than you would think.