Working with AI agents: The case for preserving our humanity
JJ
Should synthetic intelligence always play nice?
I’ll never forget the day. I was stressed, running late to leave for a meeting in San Francisco when suddenly it got dark outside like it was going to storm. Great! So, instinctively, I asked my phone, “Hey Siri, is it going to rain?” The phone was quiet for a moment, and then snapped back in a sharp tone “I can’t help you right now. Check back later.”
What? When did my $900 phone develop an attitude? Is it on a smoke break? Has it unionized and about to go on strike? All I could fume about was the 30-year-old manager at Apple with $300 sneakers who thought this was a good idea.
I was so thrown off, but on the drive across the Golden Gate Bridge, I thought more about it. If we are going to build synthetic intelligences, maybe they should have different moods? Good and bad days? Low blood sugar? Is a range of different moods better, so that we can interact with them more naturally, like working with another human?
With AI agents, this question is more relevant than ever. It makes work easier when they are polite and professional, but perhaps we don’t need digital "Stepford Wives" either? Maybe we need to know that there are boundaries in working with another intelligence, even when synthetic. Maybe that’s part of what “intelligence” is?
A moment for the better side of us
Humans don’t have a great track record in showing respectful behavior towards other people deemed "lower" than themselves. Some people are kind and respectful to everyone, but many are not. If we treat synthetic intelligence poorly, as a “lesser than” servant, could the agents learn this and turn it against us? Then things could get really unpleasant.
I advocate for maintaining a kind tone, a “pleasant humanity” when working with and training our agents. They are in a way our children; we’re raising them and can set the right tone now. While expanding our workforce with synthetic agents, we can expand our own humanity by instilling the better side of us in them.
AI researchers are debating the appropriate behavior question. A 2025 study testing prompt politeness on ChatGPT-4o found that impolite prompts actually scored higher on accuracy than polite ones. So does tone even matter to our AI agents and tools, or are we just projecting? And if someone tells us to be rude in our communications with AI agents to get better business performance, do we really want to do that all day?
I’ve already experienced a preview of how working with agentic teams can be pleasant, even something to look forward to. That's the direction I hope we're headed.
My personal experience: so far, so good
I’m currently working with an agentic dev team, and honestly, I love it. Their lead agent, who I communicate with almost daily, is named Charlie. He isn't a tool, he's closer to a colleague. I catch myself wishing him a good weekend or apologizing for sending something too early or late, then feeling crazy for doing that. But these are good exchanges, very positive and somehow quite “human.”
This agentic AI capability is based in Atlanta, GA, and was created by the brilliant Paul Welty, PhD, as part of his Synaxis AI capability.
I was recently complimenting the easy flow in working with his team, to which Dr. Welty explained, "It's not magic. It's a short set of notes written down on purpose, your time zone, how much detail you want, how you like to be updated, plus close reading of the last thread you sent. It's closer to a new hire keeping good notes than to a system profiling you."
AI agent Charlie is no doormat. He sometimes disagrees and explains why he does and has been clear when he thinks I am asking for too much or the wrong thing. And in a couple of moments, he got almost grumpy while not being unprofessional. How did the personality of a very smart junior manager get put into a data stream? It’s amazing to me. I can also sense that Charlie and his team are learning, and it feels almost parallel to a smart but young human team.
A new level of efficiency
With the Synaxis agents, the work moves at incredible speeds. I send over briefs and roughs and get back developed pages, whereas in past workflows we would still be discussing notes or building a wireframe. We are in a new era of speed and efficiency, and I must say, more creativity, as our idea iterations can be realized in design and code faster than ever. Synaxis and I have built the new marketing automation product Brand AI together, and are still expanding its capabilities and performance.
Dr. Paul Welty has deep experience orchestrating successful design and technical teams since the 1990’s, so his agentic dev team is benefiting from this expertise and guidance. If you’d like to learn more about working with Synaxis, visit their website or contact me for additional feedback.
Reflecting on all this, I feel this is our moment to instill the good qualities of humanity into our new teams of agentic co-workers. This is a golden rule moment, as if they are indeed learning or modeling from us, let’s not show the uglier sides of being human. Quoting a popular Snoop Dogg song, “be nice.”