The AI Tutor in My Pocket

It Started With a 68%

My daughter brought home a 68% on a math assessment. My wife was aghast. I had almost no emotional reaction to the number itself.

I’ve seen that grade thousands of times as a teacher. My instinct was: okay. She is not there yet. We’ll figure out why.

My wife’s instinct was closer to: absolutely not.

There was frustration, talk about consequences, and the general feeling that something had gone very wrong. I tend to see a low grade differently. To me, learning is a process, not a verdict. A 68% tells us what she could do on that assessment, on that day, with the preparation she had. It does not tell us what she will be able to do next month.

My wife wanted an A. I wanted growth. The compromise was pretty straightforward: I committed to tutoring her, come hell or high water.

Two things were in my favor

Tutoring Was Not Magical

We worked hard. A lot. It was an exercise in patience for me and an exercise in accepting help for my daughter. She had to confront the horrifying possibility that her parents occasionally knew what they were talking about.

We went through homework together. We did extra problems. We revisited mistakes. We practiced the harder questions that she tended to avoid. We used old reviews and practice tests. Then we did new versions of those reviews and practice tests.

Over time, she started bringing home better grades. Then she started bringing home A’s.

That surprised me a little because I had been prepared to accept that maybe she simply was not an A student in math yet. Instead, the experience forced me to ask a harder question: how often does a student look like a “B student” when the real issue is that nobody has shown them what better preparation actually looks like?

The First Surprise: “Check Your Work” Is a Skill

One of the most useful things I learned had nothing to do with AI at first.

My daughter’s teacher assigned homework and had students check their own work against an answer key. I do the same thing with my own students all the time. “Go check your work.” It sounds like a perfectly clear direction.

Then I watched my daughter do it and realized she either did not know what “checking” actually meant or was not taking advantage of the opportunity. Looking at an answer key is not the same as learning from an answer key.

So we made the routine explicit.

What I actually meant by “check your work”

And she learned one additional rule: you better actually check your work because Dad is going through it afterward anyway.

Then AI Changed the Pace

I could already tutor her without AI. What AI changed was the pace and responsiveness of the tutoring.

Before, if she finished a set of practice problems and needed five more just like numbers 4, 7, and 9, I had to sit there and write them. If I wanted a new practice test using the same skills but different numbers, I had to build it myself. If I wanted a progression from easier problems to harder application problems, I had to design that progression.

None of that is impossible. It is just time-consuming.

With AI, I could take a picture of the work on my phone and ask for help. In seconds, I had something usable. While she worked, I could unload the dishwasher, help another kid, or do literally anything besides sit next to her manufacturing math problems.

What I used AI for

What Our Routine Actually Looked Like

At first, we went pretty hard because she needed the repetition.

We would complete the assigned homework, check it carefully, and then do extra problems — especially the harder ones she was likely to avoid. If she had a practice test or review sheet, we would use that too. Then I would ask AI to create another version covering the same skills with different values and slightly different wording.

We would check that second version, identify whatever was still weak, and then do targeted practice on those areas.

As she got better, I dialed back the volume. Some units were going fine. She did not need to grind through extra work simply because extra work existed. Other units were train wrecks. Those needed more.

How Another Parent Can Do the Same Thing

You do not have to be a teacher. You do not have to remember fifth-grade fractions. You do not need to know words like “scaffolding” or “formative assessment.” You need the kid’s actual work and the willingness to ask useful follow-up questions.

The basic workflow is simple:

1. Show AI the work.
   Take a clear photo or paste the problem. Include the student’s attempt, not just the question.
2. Ask what it notices.
   Have it identify errors or patterns before asking it to generate more work.
3. Ask for an explanation.
   Tell it the student’s age or grade level and ask for a simpler explanation if needed.
4. Generate targeted practice.
   Practice the exact weak skill rather than dumping 30 random questions on the kid.
5. Adjust the difficulty.
   Too easy? Say so. Too hard? Say so. Ask for a progression.
6. Check again.
   Use the next round of work to see whether the mistake actually disappeared.
7. Stop when the student has it.
   Extra work is useful when it addresses a need. It is not automatically virtuous.

AI Is an Assistant, Not the Adult in Charge

There are some obvious guardrails.

AI can make mistakes. Math tools are much better than they used to be, but “fairly reliable” is not the same thing as infallible. If something looks strange, check it. Ask the tool to verify its own work. Compare with the teacher’s materials. For higher-stakes or more advanced content, be more cautious.

Also: the goal is not to have AI do the homework. The goal is to use AI to create more opportunities for the student to think.

The student should still write the answer. Work the problem. Explain the reasoning. Correct the mistake. Ask the question.

The tool is most useful when it increases practice and feedback without removing the productive struggle that actually produces learning.

Use AI to increase the learning — not to remove it

What I Actually Learned

My daughter improved because she worked. Not because ChatGPT cast a spell on fifth-grade math.

She practiced more. She checked more carefully. She corrected mistakes. She did harder problems. She learned that if she did not understand something, the solution was not to quietly move on and hope it disappeared.

AI made that process easier for me to support. It gave me speed. Variety. More practice. Another way to explain something. It turned “I should probably make her a few more problems” into something I could actually do in ten seconds instead of ten minutes.

And maybe the most useful lesson had nothing to do with math at all: I realized how often adults give kids directions that sound obvious to us because we already know how to do them.

“Check your work.” “Study.” “Practice.” “Ask for help.”

Those are not always directions. Sometimes they are skills we have never actually taught.

AI can help with that too. But the adult still has the important job: noticing what the child needs, deciding what good practice looks like, encouraging them through the frustration, and knowing when enough is enough.