What Is an AI Instructional Coach?
An AI instructional coach is software that coaches a teacher on their own lessons. The teacher records or uploads a lesson, the system analyzes it, and specific feedback comes back in minutes, pointing to moments in that lesson. The teacher keeps the coaching going by replying and decides how deep to go. A well-designed one works alongside a district’s human coaches, supporting teachers between their visits.
That is the definition we use at Edthena, and it comes from building one. This post covers what instructional coaching means in the research, how an AI instructional coach works, and what a district leader should ask before choosing one. The choice matters, because a product that misses that definition spends teacher time and PD budget without delivering coaching.
What Does Instructional Coaching Mean in the Research?
A widely cited evidence base on teacher coaching is a 2018 meta-analysis by Matthew Kraft, David Blazar, and Dylan Hogan, researchers then at Brown University and Harvard University, in the Review of Educational Research, which pooled 60 studies with causal designs. Across those studies, coaching raised measures of instruction by 0.49 standard deviations and student achievement by 0.18 standard deviations.
The authors’ working definition describes coaching as instructional experts working with teachers in a way that is:
- Individualized: sessions are one-on-one.
- Intensive: coach and teacher interact at least every couple of weeks.
- Sustained: the coaching continues over an extended period.
- Context-specific: teachers are coached on their practice in their own classroom.
- Focused: the work centers on deliberate practice of specific skills.
The review predates AI-delivered coaching. Its virtual-coaching studies used human coaches working over web platforms. It still gives a district a fair test. An AI instructional coach earns the word “coaching” to the extent it delivers those five features. Feedback on a recording of the teacher’s own class is context-specific by design. A conversation the teacher can return to across several lessons is what makes it sustained. Individualized follows from the format, since each conversation is with one teacher. Intensive and focused depend on how often a teacher uses it and how its topics are built, which is worth asking any vendor.
Why Are Districts Looking at AI Instructional Coaches?
Even the best coach can only be in one room at a time. Many districts that want coaching to reach every teacher cannot fund enough coaches to do it, and the research shows why simply scaling up is hard.
In the same meta-analysis, programs serving fewer than 100 teachers improved instruction by 0.63 standard deviations, while programs serving more than 100 teachers improved it by 0.34. The authors concluded that “scaling-up coaching programs introduces additional challenges to those confronted by smaller-scale demonstration models.” They pointed mainly to organizational challenges. Larger programs had to recruit and train a sizable coaching corps and deliver a more standardized program to teachers with mixed levels of interest.
Our founder and CEO, Adam Geller, put the constraint this way in a 2026 letter on rebuilding AI Coach: “Coaching has always been rationed because it took a specific expert, in a specific room, at a specific time.”
We see an AI instructional coach as one way to loosen that constraint, though that is a perspective rather than a research finding. The expert stays. What changes is when and where coaching can happen.
How Does an AI Instructional Coach Work?
Products differ, so here is how the loop works in our product, AI Coach, which TIME named one of its Best Inventions of 2025. We describe it in one line as “Private coaching, in the inbox teachers already check.”
- Record. The teacher securely records or uploads a lesson directly in a web browser.
- Get feedback. Within minutes, Edie, the virtual coach, sends an email calling out specific moments for review, citing timestamps and exact dialogue from the lesson.
- Reply. The teacher replies via email or logs in to watch the key moments and patterns across lessons.
Delivering coaching by email has a practical effect. Teachers already check their inbox, so the coaching conversation asks for no new habit. A teacher can reply from a phone between classes or open the platform to look at the evidence. Nobody has to find an hour on a coach’s calendar, and there is nothing to schedule.
The teacher also sets the depth. One email can be a quick read on a single lesson. The same conversation can become one topic worked across several lessons, with strategies to try and follow-through on whether they took hold.
What Should an AI Instructional Coach Be Grounded In?
Coaching needs two things underneath it: the teacher’s own lesson, and a clear model of what good instruction looks like in that subject.
For AI Coach, the second part started with academic research. We had an AI system synthesize a large body of literature on how each field defines good instruction, then spelled out what you would see and hear in a classroom when that instruction is happening. We built and tested the methods on publicly available classroom datasets that education researchers coded. The content was then organized into small topics in a reasoned order, so the coach works on one pedagogical skill at a time.
A second-grade teacher working on reading, for example, can choose a topic on teaching the sound-to-symbol foundation systematically, anchored in the Science of Reading. The coach is built to recognize what systematic phonics instruction looks like and sounds like. It can point to the moments in the teacher’s video where it showed up and where it did not, and it can reference the research on what to try next.
As of October 2026, the AI Coach catalog spans 110 topics across eight content areas: math, literacy, science, social studies, tech integration, world languages, multilingual learner support, and core teaching practices for any classroom, with a ready induction pathway for new teachers.
That is the standard to hold any AI instructional coach to. The feedback should name a moment, connect it to a specific practice, and leave the teacher with something to try.
Does an AI Instructional Coach Replace Human Coaches?
It should not, and districts should be wary of any product pitched that way. If schools could give every teacher a personal coach in the classroom every day, the way professional athletes have one, that would be the right choice. That has never been possible.
AI Coach is designed to extend your coaches’ reach between visits, not to replace them. Teachers can bring what they worked on with the virtual coach to their next coaching meeting, so the human coach starts the conversation with evidence the teacher chose to share. Each teacher can also export a record of their coaching, with their PD hours, topics, and dates. That growth record belongs to the teacher. The same approach works for teams. A professional learning community (PLC) can pick an area of focus, each teacher works on it with the virtual coach, and the team comes back together to discuss, anchored to their actual teaching. We have written more about why AI coaching should augment coaches rather than replace them.
Who Sees What Happens in the Coaching Conversation?
Privacy matters to teachers, and it should shape any district’s choice. Kraft, Blazar, and Hogan point to research suggesting that “having the same person serve as both coach and evaluator can undercut the trusting relationships needed between coaches and teachers.” In our view, an AI instructional coach that feeds an evaluation process would face the same problem.
In AI Coach, the privacy rule is simple. A teacher’s conversations with the virtual coach stay between the two of them. No one at a teacher’s school or district can read their coaching conversations or watch their lesson video through AI Coach.
We are FERPA and COPPA compliant, and our privacy and security policies explain how we take care of teacher and student data. Teachers can delete a video at any time.
How Should a District Evaluate an AI Instructional Coach?
These six questions separate coaching from feedback software:
- Does it coach on the teacher’s own lessons? Feedback should cite specific moments from a real class, not general tips.
- What is the content grounded in? Ask how the vendor defined good instruction in each subject and how it tested its methods.
- Does it fit the routines teachers already have? Coaching that arrives in a teacher’s inbox asks for no new habit.
- Who can see the conversations? Get the answer in writing, naming every role that can see anything.
- Does it work with your coaches and PLCs? Look for a way for teachers to bring their work to the people who support them.
- How does the cost compare to what you already spend on professional development (PD)? For AI Coach, compare a year of the platform to the cost of one workshop day. It is a different use of the PD budget you already have, not a new line item.
What Else Do District Leaders Ask About AI Instructional Coaches?
Will an AI Instructional Coach Take the Place of Our Coaches?
No. A well-designed AI instructional coach supports teachers between a coach’s visits, so teachers have support between meetings as well as during them. AI Coach is designed to extend your coaches’ reach between visits, not to replace them. Teachers can bring what they worked on to their coach or PLC.
Who Can See a Teacher’s Conversations With an AI Instructional Coach?
That depends on the product, so ask every vendor in writing. In AI Coach, a teacher’s conversations with the virtual coach stay between the two of them. No one at a teacher’s school or district can read their coaching conversations or watch their lesson video through AI Coach.
What Is a Good AI Instructional Coach Built On?
Two things. The first is the teacher’s own lesson, so feedback points to real moments rather than generic advice. The second is a research-based model of good instruction in each subject. For example, AI Coach’s foundational reading topics are anchored in the Science of Reading, and the coach references research on what to try next.
What Does a Teacher Need to Start Using AI Coach?
A lesson to record and a school or district email address. Teachers record or upload a lesson in a web browser, and specific feedback on that lesson arrives by email in minutes. To get started, a teacher can email [email protected] from a school or district address.
Is There Research on AI Instructional Coaching?
A large body of research covers human coaching. A 2018 meta-analysis of 60 causal studies found coaching improved instruction and achievement. AI-delivered coaching is newer. We have worked with researchers from the University of Virginia and the University at Albany, with funding from the Gates Foundation, to explore automated feedback for math lessons.
What Should District Leaders Do Next?
An AI instructional coach is worth considering when it meets the research definition of coaching, keeps the conversation private to the teacher, and works alongside the coaches you have. Use the six questions above in your next vendor conversation, and ask to see a real feedback email on a real lesson.
To see how this works in your district, learn more about AI Coach. Teachers can also start on their own by emailing [email protected] from a school or district address.
