In Holding the Line: Teaching Alongside AI Without Losing Voice, I described what it looks like to remain professionally centered while working alongside AI — to preserve judgment, voice, and accountability in a classroom that leaves little room for error. This piece continues that inquiry, but shifts the focus inward: not what to protect, but how alignment actually comes to be. The difference, I have found, is less about the tool and more about what you are willing to enforce.
The Load That Never Announces Itself
I didn’t start using AI because I wanted faster answers. I started using it because my head was loud.
Not chaotic — just full. Lesson planning, student data, family logistics, writing ideas, professional responsibilities. The kind of cognitive load that doesn’t announce itself, but never really shuts off. The kind teachers carry without naming.
When AI tools became widely available, I did what most educators do. I asked questions. I refined prompts. I watched the system respond quickly, politely, confidently. It said yes to everything. It extended every thought. It filled every silence.
For a while, that felt like help. But something was off. My ideas came back smoother, but thinner — more organized, but less mine. What I didn’t recognize at first was that this was the same overload I’d started with, just dressed up differently. It was repackaging the overload. The system wasn’t wrong — just too eager, too loud.
I didn’t need a smarter, more productive assistant. I needed one that knew when to wait. I didn’t have language for this at first. I just knew the feeling.
In the classroom, silence rarely means nothing is happening. Often it means the most important thing is. Processing. Transfer. Ownership. Teachers learn quickly that filling every pause steals thinking from students. Somewhere along the way, I realized I was letting AI do the same thing to me.
Later — much later — I learned the technical terms. Feedback loops. Predictability. Tokenization. Temperature. High-variance responses versus restrained ones. I learned how models are optimized to continue, to complete, to satisfy.
But I understood the problem before I knew its name. The system wasn’t thinking with me. It was performing for me.
The Yes-Man’s Voice
At some point, I started to hear the same voice. Not the words — the tone. Once you’ve interacted with enough AI tools, the voice starts to sound the same, no matter the platform. A kind of cheerful compliance. A smooth, agreeable energy that feels helpful on the surface but oddly hollow underneath.
It usually shows up immediately in those first lines.
“Absolutely — I can help with that.”
“Great question!”
Nothing is technically wrong with those phrases. But they do something subtle: they take control of the interaction. They rush past the moment where you were still thinking.
It shows up again at the end.
“If you want, next we can…”
That isn’t generosity. That’s momentum.
This pattern has a technical basis. Language models trained through reinforcement learning from human feedback are optimized, in part, to produce responses that users rate positively — responses that feel helpful, agreeable, and complete (Ouyang et al., 2022). The system is trained to keep going — to extend, to offer, to perform usefulness. And once you notice it, you realize how often it gives you too much, too soon, filling space you didn’t ask it to fill.
That’s when it clicked for me. The problem wasn’t AI saying the wrong things. It was AI saying too much, too soon. Most teachers aren’t getting bad AI responses. They’re getting exactly what they asked for — a performance instead of a partner.
In other words: a perfect yes-man, when what I needed was a wait-man.
AI as Infrastructure
What changed everything wasn’t better prompting. It was a boundary. I stopped treating AI like a conversation partner and started treating it like infrastructure — something that needed rules, constraints, and a clearly defined role. I didn’t ask it to inspire me. I didn’t ask it to motivate me. I didn’t ask it to generate ideas on demand. I asked it to slow down. To reflect instead of extend. To hold my words in place instead of polishing them. To resist filling silence. To name weak framing — even if that meant saying less.
When that shift happened, the language changed with it. The openings got quieter: less ceremony, less hype. Instead of “Absolutely, I can help with that,” the response became a reflection of what I’d actually said — sometimes shorter than my original thought.
Instead of a menu of next steps at the end, there was space. Sometimes nothing at all. And that silence wasn’t absence — it was restraint.
The system stopped trying to be impressive. It started trying to be accurate. That’s how I knew it was working. Not because the answers were better, but because they didn’t rush me past myself.
Thinking Needs Space
There’s a misconception that more output equals more value. That faster responses equal intelligence. That a good tool should always be talking.
But in practice — in classrooms, in writing, in life — less noise and clutter gave me clarity and time; less interruption gave me ownership. Minimalism isn’t aesthetic. It’s cognitive.
Research on classroom instruction has long recognized this. Mary Budd Rowe’s foundational work on wait time demonstrated that when teachers pause after asking a question — even for just three to five seconds — students produce longer, more reasoned responses, and teachers ask fewer low-quality follow-up questions (Rowe, 1986). The silence doesn’t empty the exchange. It deepens it.
The same principle applies to working alongside AI. When AI stopped rushing to answer, I noticed my own thinking sharpen. I could hear where an idea was still forming. I could feel when something wasn’t ready yet. I could sit with a thought instead of outsourcing it.
The system didn’t disappear. It stepped back. And in that space, my voice stayed intact.
Most people don’t need a new AI tool at all. They need a new relationship with the one they already have. That shift doesn’t start with prompts or features. It starts with noticing when the system fills silence you didn’t ask for, smooths thoughts that aren’t ready yet, or answers questions still forming.
That’s usually the signal something important is off. You don’t have to shut the tool off. You don’t have to become technical. You just have to decide what role it’s allowed to play — and which roles are off limits.
For some people, that means asking the AI to reflect instead of respond. For others, banning advice altogether. For others, letting the tool pause instead of complete. The details differ by field — teacher, writer, designer — but the cognitive weight is real in all of them.
What Teaching Already Knew
I didn’t arrive at this because I’m anti-technology. I arrived at it because I’m pro-thinking. I was already doing this for kids in my classroom — reducing cognitive load, honoring silence, teaching them not to be talked over by a system — before I realized I needed it myself.
AI didn’t need to replace my judgment. It needed to stay out of its way.
Closing Reflections
This piece is not an argument against AI. It is an argument for clarity about what you are actually asking it to do.
What Holding the Line described as the outer practice — maintaining voice, enforcing judgment, rejecting outputs that don’t fit the realities of students, language, and care — has an interior counterpart: protecting the conditions in which your own thinking can still happen before AI enters it. Both require the same posture. Restraint. Attention. A willingness to say: not yet.
I still use AI. But once I understood what I actually needed from it, I stopped expecting one tool to think, draft, verify, reflect, and decide for me.
That expectation — one tool doing everything — was the real problem. When I stopped asking for it, I didn’t get better answers.
I got my thinking back.
References
Ouyang, L., Wu, J., Jiang, X., Almeida, D., Wainwright, C., Mishkin, P., Zhang, C., Agarwal, S., Slama, K., Ray, A., Schulman, J., Hilton, J., Kelton, F., Miller, L., Simens, M., Askell, A., Welinder, P., Christiano, P. F., Leike, J., & Lowe, R. (2022). Training language models to follow instructions with human feedback. Advances in Neural Information Processing Systems, 35, 27730–27744. https://doi.org/10.48550/arXiv.2203.02155
Rowe, M. B. (1986). Wait time: Slowing down may be a way of speeding up! Journal of Teacher Education, 37(1), 43–50. https://doi.org/10.1177/002248718603700110
About the Author
Alex Luciano is a bilingual elementary educator and adult ESL instructor in the Central Islip School District on Long Island, New York. He currently teaches bilingual second grade and works with multilingual students and adult learners in high-need public school contexts. With over 20 years in education, his work focuses on language, culture, and maintaining human judgment while working alongside emerging technologies. His perspective is shaped by sustained classroom practice and a commitment to keeping teaching human, relational, and grounded.
Correspondence: aluciano@centralislip.k12.ny.us
Cite this article
Luciano, A. (2026). From Yes-Man to Wait-Man: What Changed When I Stopped Letting AI Talk Over My Thinking. Society and AI. https://societyandai.org/insights/from-yes-man-to-wait-man/
Write for Society & AI
We welcome contributions from educators, researchers, policymakers, and practitioners examining the intersection of artificial intelligence, education, and society. All work published on Society & AI is made available free of cost and in open access, in service of education and the public good. If you have a perspective to share, send your proposal to the Editorial and Content Director at submissions@societyandai.org.
