AI instructional coaching, honestly.
By Yechiel, Founder · Updated
An AI instructional coach is software that supports one teacher's practice over time, in conversation: it asks about a goal the teacher set, remembers the answer, and comes back to it. That makes it different from a lesson-plan generator, which produces materials, and from an observation tool, which scores a recorded classroom.
The phrase is new, and it is already attached to very different products. A school evaluating them deserves to know which one it's buying, and what none of them can do.
What does "AI instructional coaching" mean?
Researchers who study instructional coaching describe it as individualized, sustained and focused on one teacher's practice (meta-analysis of 60 causal studies, Review of Educational Research). An AI instructional coach is software doing part of that job.
The word doing the work is coaching. A tool that answers questions is a search box. A tool that produces materials is a generator. Coaching means someone asks about the thing you're trying to change, remembers your answer, and comes back to it.
Which kinds of product use the name?
Three different kinds. They come out of different budgets and solve different problems.
| Kind | What it does | What it isn't | Ask the vendor |
|---|---|---|---|
| Content generator | Lesson plans, rubrics, worksheets, unit outlines. Saves real hours. | Coaching. It has no memory of your practice and never asks how the lesson went. | Does it know anything about me between sessions? |
| Observation and analysis tool | Scores classroom video or transcripts against a framework: talk ratio, question types, wait time. | Support. A score is evidence for a coach, not a conversation with one. Recording a classroom is also a trust decision before it's a technology decision. | Who sees the output, and does anything happen after the score? |
| Follow-up coach | Brief written check-ins tied to one goal the teacher set, with memory across the cycle. | An observer. It knows only what the teacher writes. | What happens on an ordinary Tuesday? |
A school can want all three, and a vendor that blurs them is worth slowing down for. PD That Works builds the third kind: PD follow-up after a workshop. Each teacher sets one practice goal for a 30-day cycle, and the coach follows up in brief private check-ins, 3 times a week by default. Each teacher chooses their own cadence, anywhere from 2 to 5 check-ins a week.
What can't an AI coach do?
Being clear about this is the fastest way to tell an honest vendor from an eager one.
- It can't see the classroom. It reads what a teacher writes. It doesn't watch fifth period, notice the student in the back, or catch what the teacher didn't think to mention.
- It can't model a lesson. Demonstration is a human act.
- It can't have the hard conversation. A teacher whose practice isn't working needs a person with standing and judgment.
- It can't stand in for a new teacher's mentor. Induction is relational, and a first-year teacher needs someone in the building.
- It can't promise student outcomes. Anyone selling test-score gains from a coaching chatbot is selling something the evidence doesn't support.
Our own boundary, as it appears on our evidence page: "The platform records participation and synthesizes teacher-reported implementation patterns. It does not observe classrooms, score teachers, or claim student-outcome impact."
What an AI coach can do is the part no human coach has the hours for: follow up with each teacher between visits, and remember what they said. We make coaches better, we don't replace them.
Does AI coaching for teachers work?
The evidence is strongest for the mechanism, not for the product category. Here are four findings, with what each one does and doesn't support.
Recording progress toward a goal, and reporting it to someone, measurably raises goal attainment. — Meta-analysis of 138 randomized trials (19,951 participants), Psychological Bulletin. Most of those trials were about personal goals, not teaching. What it supports is the shape of a check-in: one goal, progress in writing, reported to someone.
Over a third of feedback interventions make performance worse. The dividing line: feedback aimed at the practice helps; feedback aimed at the person, including praise, backfires. — Meta-analysis of 607 effect sizes, Psychological Bulletin. That's why a good coach talks about what the teacher tried and what happened, not about the teacher.
Teacher coaching is among the most effective PD forms ever measured, and more coaching hours did not mean better outcomes. Specificity beats dosage. — Meta-analysis of 60 causal studies of human coaches, Review of Educational Research. Note the limit: all 60 studies are of human coaches. It shows that coaching works. It doesn't show that an AI coach does.
The AI-specific evidence is narrower. In two randomized trials run over ten months, one with human coaches and one with a chatbot coach, the chatbot matched the human coaches on goal attainment, and only on goal attainment. — Randomized controlled trials, PLOS ONE. That is the ceiling of what the evidence supports for AI coaching.
So the claim we make is narrow on purpose: goal attainment, through the mechanism the check-in is built on. We don't claim replicated coaching effects, and we don't claim student outcomes.
What should an honest AI coach promise?
Two commitments, and each one costs something.
It says what it is. Every teacher is told plainly, at the start, that their coach is an AI.
It isn't tuned to please. In controlled experiments, AI that flattered its users made them more convinced they were right and less willing to repair a conflict, while making them want to keep using it. — Controlled experiments, Science. That is the uncomfortable finding for this whole category: the flattering assistant is the one people want more of. So teacher-satisfaction ratings are never used to tune our coach. We chose to be liked slightly less and to be somewhat more useful.
Who can read what a teacher writes?
School leaders cannot open teacher reflections or coach responses. Depending on the privacy mode chosen before launch, they may see either aggregate participation or named participation status. Teachers see that choice before joining.
Leaders see participation under the privacy mode the school chose, and anonymous group themes — never anything a teacher wrote, unless they choose to share it. That boundary is what makes a teacher's answer honest, and an honest record is the only kind worth summarizing. More on how teacher privacy works.
What should you ask any AI coaching vendor?
- Does it tell teachers plainly that it's an AI?
- Is teacher satisfaction used to tune it? A model tuned on "did you like that" drifts toward flattery.
- Who can read what a teacher writes, and do teachers know before they join?
- What does it refuse to claim? A vendor with no list of things it doesn't do hasn't thought about it.
Common questions
What is an AI instructional coach?
An AI instructional coach is software that supports one teacher's practice over time, in conversation: it asks about a goal the teacher set, remembers the answer, and comes back to it. That makes it different from a lesson-plan generator, which produces materials, and from an observation tool, which scores a recorded classroom.
Can AI replace an instructional coach?
No. An AI coach reads what a teacher writes. It can't observe a classroom, model a lesson, or hold a hard conversation. What it adds is follow-up between a human coach's visits: brief check-ins on each teacher's goal, with memory across the cycle. We make coaches better, we don't replace them.
Is AI coaching for teachers effective?
The strongest evidence is for the mechanism rather than the product category. A meta-analysis of 138 randomized trials found that monitoring progress toward a goal and reporting it to someone raises goal attainment. Two randomized trials, one with human coaches and one with a chatbot coach, found the chatbot matched the humans on goal attainment only. The large coaching meta-analyses are of human coaches. The honest claim for AI coaching is goal attainment, not replicated coaching effects or student outcomes.
Do teachers' words stay private from school leaders?
School leaders cannot open teacher reflections or coach responses. Depending on the privacy mode chosen before launch, they may see either aggregate participation or named participation status. Teachers see that choice before joining.
Where PD That Works fits
PD That Works turns one PD day into 30 days of classroom follow-through: teachers privately work on one practice through brief check-ins, 3 times a week by default, while school leaders see anonymous implementation patterns and a Day-30 report. See how it works, or what the report can and cannot show.
The follow-through this guide describes is what the product does: one practice goal, private check-ins, anonymous patterns, an evidence-labeled Day-30 report. Preview a cycle from my PD