Instructional coaching is among the most effective forms of professional development. But doing coaching right is resource-intensive and hard to scale. Can artificial intelligence fill in the gaps?
A growing number of school districts and organizations are experimenting with using AI in instructional coaching cycles. AI-enabled tools can guide teachers through self-reflection and goal-setting, which is a key component of a coaching cycle. They can also deliver feedback and analyze data, removing some of the administrative workload from instructional coaches, who are often strapped for time.
“Our hypothesis is if you could increase the number of coaching touch points with the teacher, then maybe we can make the coaching more efficient and more scalable, and you could extend the reach of a human coach,” said Sarah Johnson, the chief executive officer of Teaching Lab, a curriculum-based coaching nonprofit.
Teachers need regular coaching throughout the school year to truly improve, she said. And a coaching cycle requires a lot of steps. A coach typically meets with a teacher, looks at a lesson plan, analyzes student data, gives initial feedback, observes a lesson, writes down observations, debriefs with the teacher, and then observes again to monitor any improvement.
AI can replace—or supplement—some of these steps, Johnson said, giving coaches more room in their schedules and teachers more opportunities for feedback.
Melinda George, the chief policy officer for Learning Forward, a nonprofit focused on effective professional development, said AI can be a “huge timesaver,” summarizing notes and taking care of administrative tasks so the human-to-human interaction is focused on the “quality stuff” of coaching.
It can also make room for deeper coaching, experts say, especially among teachers who are new to this type of PD and might be uncomfortable with being observed.
Giving teachers an opportunity to first get feedback from an AI tool “allows them to dig deeper on what kinds of questions they have that they want to then bring to their real-life coach,” George said. “It almost helps to enhance their real-life coaching experience by allowing them to go a little bit deeper, quicker, and to also do some of the challenging things in their own space without risk or fear of being judged.”
Still, districts need clear policies on the use of AI in coaching or other professional development, cautioned Stacey Alicea, the executive director of the Research Partnership for Professional Learning, a collaborative housed at the Annenberg Institute at Brown University.
Protecting student privacy should be the foremost consideration when using or sharing data in coaching cycles, she said, and educators should guard against any biases in AI-generated responses. School and district leaders should also review the outcomes and experiences of any AI-assisted professional development, she said.
“Once you’ve made all of your selections [for tools and programs], I think AI requires, because of its constant evolution, to be continuously monitored in its implementation,” Alicea said, adding that research on AI-driven professional development, and which models are most effective, is still in the nascent stage.
AI-assisted coaching can generate customized feedback and podcast learning tools
So far, AI is being incorporated in instructional coaching in a variety of ways.
New York City and other school districts are piloting a tool created by the Teaching Lab, Johnson said. An instructional coach can take a photo of their handwritten observation notes and send it to the AI tool. It will respond with feedback for the teacher, based on the coach’s notes. If the coach approves, the AI will create a customized feedback report for the teacher, including a two-person podcast to go over the feedback in a more conversational way.
In the St. Vrain Valley school district in Longmont, Colo., meanwhile, teaching and learning coaches are rolling out AI for use in their new-teacher induction program and for evaluations of effective or highly effective teachers. The AI tool is part of the Edthena video-coaching platform.
Participating teachers upload a video of their lesson and then go through the video alongside an AI chatbot. The chatbot helps the teacher name areas of focus for the coaching cycle—for example, checking for student understanding or clearly presenting and explaining content. Then, the teacher watches the video of their lesson and makes a note when they notice a specific moment related to those areas. The AI will guide the teacher through reflecting on strengths shown and opportunities for improvement.
The AI experience is just one tool in coaches’ arsenals—it’s meant to reinforce reflection and goal-setting done with human coaches, said Patty Hagan, a teaching and learning coach in the district.
A coach who’s able to be in a classroom and see a teacher teaching in his or her environment ... can’t really be replaced by a machine.
Budget constraints limit the number of coaches in the district, she said, so AI allows teachers to participate in coaching cycles more frequently than they otherwise would.
“It enables teachers to still have that reflective process and presence, and even for our administrators, it takes some of that heavy weight of evaluating off their plates,” she said.
Cleveland Smith, a 6th grade language arts teacher at Altona Middle School in the district, volunteered to use the AI tool as part of his evaluation this year. An avid user of AI in the classroom, Smith was eager to see how the tool could be used to further his own professional growth. But he was disappointed.
“I set my expectations at a level that was beyond what AI is capable of doing right now, because my hope was that AI would be able to review my video and that AI would be able to give me feedback,” Smith said.
For example, one of his goals is for his students to be more engaged in class, and he told the AI that it could measure engagement by measures like raised hands and eye contact. But instead of the AI naming points in Smith’s lesson where students seemed disengaged or offering specific feedback, the tool pushed Smith to reflect on his own teaching.
“It might say, ‘At this point in the video, do you feel like your students were engaged?’” Smith recalled. “I would either say yes or no, and then depending on what I said, it would give me another follow-up thought or consideration or a research article to read and consider. ... It was another way to do a personal reflection.”
The experience wasn’t as useful as it would have been if a human coach or a colleague watched Smith’s lesson and made observations, he said. But it was still valuable to reflect on his own by watching a recorded lesson.
“It’s like watching game footage—if you’re playing a sport or something, you get to watch yourself do the thing and you’re definitely going to learn,” he said. “I didn’t realize I say ‘um’ all the time. I didn’t realize that kid had his head down or I didn’t realize that I call on these two tables 10 times more than I call on those two tables. You’re still going to notice things, but I don’t think that I noticed things more as a result of AI’s prompting.”
Still, Hagan said, it’s the principal’s job to tell teachers how or what to teach. It’s a coach’s job to help teachers learn how to reflect on their practice and meet their goals.
“The AI is not evaluating the teacher—you’re evaluating yourself,” she said. “This is driven by you, because that’s what we want to create: We want to create teachers who are self-reflective and build their own goals.”
Will teachers be on board for AI-assisted coaching?
In December 2024, the EdWeek Research Center asked a nationally representative sample of teachers how they would feel if asked to participate in professional development or instructional coaching that incorporated artificial intelligence. More than half—57 percent—said they would feel “somewhat” or “very” comfortable.
About a quarter said they would feel “somewhat” uncomfortable, and 20 percent said they would feel “very” uncomfortable.
“You’re always going to have teachers who are going to be like, ‘Oh yeah, I’m in! Give me more chances to do it,’” Learning Forward’s George said.
Even so, “I think with anything new, you’re going to see some initial resistance,” she added. “A term we use is a gradual release of responsibility—introduce things in a way that builds confidence, and you do it over time.”
In other words, coaches and school leaders should introduce AI-assisted coaching slowly, she said, letting teachers watch videos and see examples before they try new approaches themselves. Teachers should have the opportunity to ask questions and experiment in a low-stakes way.
And as ChatGPT, Google’s Gemini, Microsoft CoPilot, and other AI tools become more widely used in society, teachers will likely become more comfortable having AI incorporated into their professional development, George said.
Right now, though, teachers—and their coaches—would benefit from professional development on using generative AI, specifically the to give useful, relevant answers, George said. Yet many districts haven’t devoted time to professional development on AI, past surveys show.
In New Jersey, Rosemary Seitel offers professional development on AI to coaches and other educators through the state’s Learning Forward affiliate. Seitel, a consultant with the Foundation for Educational Administration, the professional learning division of the New Jersey Principals and Supervisors Association, said she’s shown coaches how to turn their written feedback into podcasts, upload data from a coaching session and ask for patterns and trends to show the teacher, and use ChatGPT to work through a tricky coaching scenario.
For example, a coach could ask ChatGPT what to do when a teacher is not implementing the instructional strategies that they had discussed. ChatGPT will offer some coaching strategies, and ideally, the coach will bring those strategies to a brainstorming session with a colleague, Seitel said.
“It taps into your own brilliance—what you already have inside you,” she said. “It just brings it out. But you do need somebody who’s been playing around with it enough to guide you because we’re all on this journey. It’s the beginning of the beginning of all this and how it’s going to evolve.”
AI experts insist there should always be a human in the loop during teacher PD
As AI becomes more advanced, and more districts develop plans to integrate it into operations, experts predict it’ll become a more common tool in teachers’ professional growth.
For a , researchers at the Annenberg Institute interviewed developers, professional learning providers, and educators about the use of AI in teacher PD. One emerging use they identified was classroom simulations for teachers to practice classroom-management techniques—a practice that some teacher-preparation programs are adopting.
Another was districts using AI systems to build differentiated learning plans aligned with district priorities that allow teachers to learn at their own pace and receive targeted feedback.
“For the first time, if a million teachers knocked on the door tomorrow, we could support them all, with these high-quality and completely differentiated experiences,” one developer told the researchers. “Of course, we always want as many experts in the room and on the ground [with] teachers, but there’s always that gap. Instead, we could turn on more computers and support more teachers.”
But there should always be a human in the loop, experts emphasized.
AI “doesn’t know your students,” George said. “You can prompt it, you can describe your location, you can get pretty specific—but I think a coach who’s able to be in a classroom and see a teacher teaching in his or her environment, that can’t really be replaced by a machine.”
