Last year, I visited a seventh-grade math classroom in a public school in the Bronx. About 20 students were working on laptops with an A.I. tutor, solving story problems about converting fractions to decimals. Their teacher walked around the room, checking a dashboard that showed how many tries each student needed to get the answer right.
At first glance, everything looked fine. Students were focused. The software was responsive. Most of them eventually got to the right answer. Gold star for productivity, right?
Not quite.
When I looked closer, many students were lost. They didn’t really understand fractions. The A.I. tutor kept backing up and offering another step, but it never found the real problem: the missing concept underneath. The teacher’s dashboard showed who was stuck, but not why.
That “why” is the heart of teaching. A good teacher can tell when a student needs a quick explanation, a visual model, a class discussion, or just a different way into the problem. In that classroom, though, the most important work had been handed to a tool that could only keep the students clicking.
The room was quiet, but not in a good way. No one was arguing through a strategy. No one was asking a classmate, “Wait, how did you get that?” Each child sat alone, staring at a screen, trying again.
That silence has a cost. Learning is not just about getting the right answer faster. Students need confusion, discussion, and even a little frustration to build real understanding. A.I. can remove too much of that struggle before students have a chance to work through it themselves.
This matters because the current trend in education technology is pushing hard toward one-to-one A.I. tutoring. Some early research shows gains in basic skills, but efficiency is not the same as understanding. A bot can help a student practice a procedure. It cannot build the relationships, trust, and conversation that make deeper learning possible.
When a student struggles, she often needs more than a clearer explanation. She may need to hear another student explain it badly, argue with that explanation, and slowly figure out what makes sense. She needs a teacher who can tell whether she is confused, bored, anxious, or on the edge of a breakthrough. A dashboard cannot read a room. A teacher can.
There is also a bigger concern. Kids are already spending too much of their lives isolated behind screens. If schools make A.I. tutoring the default, especially for students in under-resourced schools, we risk giving wealthy students rich, human-centered learning while poorer students get headphones, laptops, and software that never truly knows them.
That does not mean A.I. has no place in schools. The question is not whether to ban it or embrace it blindly. The question is whether it supports the kind of learning we actually value.
Students need to become original thinkers, thoughtful citizens, and strong collaborators. They need to ask better questions, defend their ideas, listen to people who disagree, and connect what they learn to the world around them. An A.I. tutor can help drill a formula. It cannot teach a student to debate the ethics of that formula’s use with a classmate.
Used well, A.I. could help teachers. It might organize student drafts, track feedback over time, or help teachers see how a student’s thinking is developing. It could make deeper projects easier to manage. That is a very different goal from replacing teachers with digital tutors.
Some teachers are already finding that balance. One Bronx science teacher first built an A.I. tool to act like a classroom coach. His students pushed back. They wanted to work through hard problems with him, not with a machine pretending to be him. So he changed the tool. Now it helps students prepare their thinking before they meet with him. The A.I. handles the setup. The teacher handles the deeper questions.
That is the right direction.
The seventh graders I saw did not need a better chatbot. They needed a teacher with the time and support to understand what was confusing them and help them work through it. Real learning requires friction, conversation, and connection.
A.I. can either give teachers more room to do what they do best, or it can quietly replace the best parts of teaching with more screen time. The choice we make will shape more than test scores. It will shape the kind of people students become.