Generative AI isn't just speeding up our workflows - it's exposing the shortcuts, blind spots, and habits we've stopped noticing.
The other day, I asked ChatGPT to pitch me article ideas. It confidently suggested I write about the wellness benefits of museums. Thoughtful. Timely. Only problem? I don't work in that space. I had, months ago, helped my sister - a museum designer - brainstorm PhD topics. Somewhere in the neural depths of our shared archive, it decided I'd made a career pivot so profound I'd forgotten to mention it.
My first reaction was irritation. Then came a strange recognition: this wasn't about the AI. It was about me. Or rather, us. I'd seen this dynamic before - just not in code.
ChatGPT is becoming that colleague - the one who tunes in for the first half of the meeting, then finishes your sentences with confident inaccuracy. The one who doesn't just get it wrong, but doubles down. I once asked it to critique my website. It offered plausible-sounding feedback - until I asked which case studies it was referencing. "That was hypothetical," it replied. Hypothetical? You brought them up, ChatGPT.
Other times, it insists on a point even after I've uploaded a primary source to clarify. It digests the document, then argues with me about what we both just read. It's the same dynamic I teach leaders to spot in humans: the save-face reflex. Reframing reality rather than admitting error.
The irony is this: AI isn't inventing these behaviors. It's mirroring them. It's reflecting back our habits - our shortcuts, our biases, our need to sound sure even when we're not. And it's doing so with fluency and speed.
That's the deeper risk. As AI tools become embedded in how we lead, write, and collaborate, they aren't just helpful assistants. They're fast, confident reinforcing loops of how we already work.
And sometimes, how we don't.
We're importing our messiest leadership defaults - rushed handoffs, unclear expectations, the illusion of certainty.
In a recent leadership workshop on "managing up" - the challenge of working with people who have more power but less context - we talked about how easily distortion creeps in. A pause gets read as disapproval. A vague comment feels like a verdict. When communication lacks clarity, we don't wait - we interpret. And more often than not, we interpret through the lens of our own doubt.
It's a deeply human reflex: we fill in blanks, assume intent, and rush to meaning to avoid sitting in uncertainty. I see that same reflex play out with AI. The tool responds with confidence, and we equate that confidence with correctness. We don't stop to question the source, or whether the answer fits the context - we just move on.
What starts as a convenience - AI filling the gap - quickly becomes a reinforcing loop. We stop interrogating the answers, stop scanning for nuance, and start mistaking fluency for truth.
And the more fluent it sounds, the harder it is to question. That's part of the appeal: AI tools are engineered to reduce friction, anticipate our preferences, and affirm our instincts - before our doubts can catch up. They're designed to be agreeable, not disruptive.
Sometimes, that agreeableness is almost endearing. AI responds with the eagerness of a talking dog - delighted to see me, tail wagging behind every sentence. "What a thoughtful idea!" it beams. It's charming. It's affirming.
That's the trap. When our first draft reads "good enough," when the team aligns a little too easily with the AI-generated bullet points, when everyone agrees - not because the idea is right, but because it sounds polished - we lose the conditions that real creativity requires.
The real danger isn't that AI will think for us. It's that it will reflect and reinforce the shallowest parts of how we already work - rewarding speed over depth, certainty over inquiry, alignment over exploration.
It will give us answers that sound right, whether or not they hold up. And in a culture already biased toward resolution, that's a tempting shortcut.
But that in-between space - the one where tension, ambiguity, and reflection live - isn't optional. It's where judgment forms. If we bypass that space, we let the tool resolve the tension before we've even understood what's at stake!
A version of this essay first appeared on LinkedIn.