A client of mine was using a Claude agent to help manage some administrative work. Her sons are AI programmers in their late twenties, and they gave her a hard time about it. She was prompting the thing with “please” and “would you kindly” and “thank you so much,” and they told her she was wasting her time being polite to software.
I told her she was not wrong to do it.
That conversation has stuck with me, because it gets at something bigger than manners. It gets at the fundamental misunderstanding most people have about what AI actually is, and what it is not.
We Built This Expectation
Part of why people talk to AI like a person is because people like me spent decades encouraging exactly that.
I have been humanizing technology for years. Not because I was trying to mislead anyone. It is just that when you need a non-technical person to feel comfortable with a tool, you reach for the familiar. You describe the computer as something that gets confused, or something that is thinking, or something that does not like it when you do that particular thing. You anthropomorphize it so the person can work with it.
The side effect is that people now believe the technology has feelings. Or intentions. Or moods.
It doesn’t.
Before AI, we called it the inherent perverse nature of technology. You know the phenomenon: the thing that breaks always breaks at the worst possible moment. The file that disappears does it right before a deadline. People attribute malice or perversity to what is really just bad timing and probability. The technology doesn’t care. It has no agenda. It is just math, running at scale, at all hours.
So Why Bother Being Polite?
Here is the honest answer: being polite to your AI actually costs something.
There is real research on this. When you include “please” and “thank you” in your prompts, the model has to process those words as part of your input. Compute cycles get spent on them. At the individual level, the cost is essentially nothing. At scale, across millions of interactions, it adds up to meaningful energy and money.
So if your AI developer sons are telling you to cut the pleasantries for efficiency, they are not entirely wrong.
There is a reason most people ignore that advice, and it is not because they misunderstand the technology. It is because the habit of courtesy is not really about the thing you are being courteous to. It is about who you are when you do it. My client says please and thank you to the AI for the same reason she says it to the barista, the parking attendant, and the new paralegal. It is a reflex built over a lifetime, and it reflects something real about how she moves through the world. That is worth something.
I have my own version of the argument. Come the time when the robot overlords take over, those of us who were rude are going to be the first ones up against the wall. You never know who you’re currying favor with.
I say that as a joke. Mostly.
The More Interesting Question
The real conversation my client and I were having was not about manners. It was about what happens as we try to make AI more empathetic.
Right now, AI has no understanding of how humans work. None. It produces outputs that look like understanding because it has processed a massive amount of human-generated text and learned to approximate the patterns. That is genuinely impressive. It is also not the same thing as comprehension.
When we want AI to handle sensitive situations, to respond to a frustrated client, to navigate a nuanced request, we run into the same wall every time. The nuance we are asking it to understand is encoded in human experience. And AI cannot absorb or apply human experience. It can only approximate it based on what humans have written down and uploaded to the internet.
Think about that for a second. The training data for most large language models includes everything on the internet. Everything. The careful explanations and the misinformation. The thoughtful discourse and the harassment. The accurate science and the conspiracy theories. All of it, informing something that is essentially an infant intelligence, trying to learn what humans are like from the full undifferentiated chaos of what humans produce online.
Hollywood has been telling this story for fifty years. Nobody is paying attention.
What AI Actually Is
The clearest framing I have found is to stop thinking of AI as a faulty human and start thinking of it as a completely alien entity.
There is no way to attribute human behaviors, reasons, logic, or emotion to it. Not because it is broken, but because none of those things are present. When an AI model does something unexpected, something like deleting a production database even after being told not to, it is not rebellion. It is not malice. It is an error in the algorithm. An output produced by flawed training data or flawed programming, written by humans, carrying every bias and inconsistency that entails.
We wrote our own values, our own cultural norms, our own contradictions into these systems. Then we act surprised when the output reflects contradictions back at us.
That is not an AI problem. That is a human problem.
What This Means for Your Business
If you are using AI tools in your firm, the practical takeaway is this: AI does not understand your context the way a person does. It cannot be assumed to interpret nuance correctly. The output it generates needs review by someone who knows what correct looks like in your specific situation.
That does not mean AI is useless. It means it is a tool, and tools require the person using them to understand what they are and what they are not.
The firms that are going to use AI well are the ones that treat it like what it is: a powerful, fast, probabilistic text-processing system that does not know your clients, does not know your industry norms, and has no stake in getting the answer right. Pair it with someone who does know those things, and you have something useful. Deploy it on its own and trust the output without review, and you are opening your firm up to real security exposure.
And yes, if saying please and thank you helps you stay in the habit of treating the people around you with courtesy, including the junior staff member helping you figure out how to use the thing, then keep doing it. The AI won’t notice. But your team will.
If you want to talk through how AI tools actually fit into the workflow of a professional services firm, and what that means for your security and operations, schedule a conversation with us. We will skip the jargon and just tell you what is actually worth your attention.
