Tech Explained: AI for Teacher Evaluations: Major Time-Saver, or Premature? in Simple Terms

Artificial intelligence is making its way into the teacher evaluation process, with some administrators using it to streamline classroom observations. But as the technology takes on a greater role in assessing educators, some critics warn that lagging guardrails and unclear policies could introduce new risks to an already contentious system.

While the scope of AI use in teacher evaluations is difficult to quantify, interviews with educators, administrators and school staff across the U.S. suggest the practice is already emerging in pockets of K-12 systems, often at the discretion of individual administrators. The trend raises new questions about transparency, data privacy and accuracy in a process tied to teacher performance and compliance — particularly as many districts have yet to establish formal policies or guidance around the use of AI.

A PRACTICAL TIME-SAVER, WITH LIMITATIONS

Melinda Garrick, an assistant principal at Pelham Lab High School in the Bronx, said K-12 teachers are already using AI informally, for tasks like lesson planning, differentiation and administrative work. Moreover, she said administrators like herself are beginning to apply AI tools to their own workflows, including teacher observations.

“I take my normal, low-inference observable notes,” Garrick said, referring to objective records of what occurs in a classroom, capturing exactly what teachers and students say and do without subjective interpretation or evaluative judgment. She then uploads the Danielson rubric — a framework that divides instructional practice into several domains — to ChatGPT and asks it to suggest a rating and give aligned feedback.

“I use it as a starting point, and then I take what they produce, and then I tweak it and use that in my final observation report,” she said.

For Garrick, who has over a decade of classroom teaching experience under her belt, the appeal is largely practical. She pointed out how complex traditional classroom environments can be, and described how generative AI has helped her synthesize all she absorbs in a single class period.

“I use AI as a tool,” she said, emphasizing that it does not replace the observation itself. “It definitely saves me time … It probably cuts my observation time in half, or the write-up part in half at this point.”

The allure of saving time comes amid growing workloads for school leaders, many of whom balance evaluations with a wide range of operational and instructional responsibilities. Evaluations often require detailed written feedback aligned to multiple components of teaching practice, making them both time-consuming and cognitively demanding.

However, while Garrick said her school’s principal and teachers are aware of her AI use in the evaluation process, there is no set policy in place for how it’s implemented.

The growing use of AI in this setting thus appears to be unfolding largely without formal school- or district-level direction, a dynamic that differs from more structured ed-tech rollouts, where tools are typically vetted, approved and implemented systemwide — a point of concern for some education leaders and agencies.

Rob Weil, CEO of the National Academy for AI Instruction, an initiative founded by union American Federation of Teachers, said he thinks using AI for teacher evaluation is a premature application of the technology. He pointed to fundamental limitations in current AI systems, particularly their inability to capture the intricacies of any given classroom.

“It doesn’t understand the relationships, it doesn’t understand the context … I just don’t believe the tools are good enough right now,” he said. “I don’t believe they can capture human context.”

Those limitations, he said, are especially significant in evaluation settings, where observations are used not only for feedback but for formal assessments tied to teacher performance.

Transparency, he added, is another critical issue, particularly in situations where teachers may not be aware that AI is being used to shape feedback or ratings. He said the issue does not appear to be that entire schools are piloting ed-tech tools for teacher observations, but rather that some administrators are using it at their own discretion.

“Using it without transparency is a real problem,” Weil said. “Nothing should be in the shadows.”

PRESERVING HUMAN JUDGMENT

Other educators see potential benefits in AI-assisted observations, particularly if the technology is used to reduce administrative burden while preserving human judgment.

Jennifer Lawson, instructional technology integration coordinator at Northville Public Schools in Michigan, said AI could allow administrators to be more present during classroom visits, rather than focused on low-inference documentation. Lawson’s thinking on the topic differs from Garrick’s: Whereas Garrick uses AI to synthesize her low-inference notes into actionable, rubric-aligned feedback for the teachers she supervises, Lawson believes administrators should allow AI to transcribe the events taking place in the classroom.

“I see it as a very valuable asset, depending on how it’s used [by administrators],” Lawson said. “If AI could do that [documentation], they could be more observant and really have a feel and understanding of what’s happening in the classroom.”

Still, Lawson stressed that human oversight must remain central to the process, particularly when translating observations into evaluative feedback. That balance, she added, is critical to ensuring that feedback reflects the nuances of classroom practice rather than purely automated interpretation.

“My big spiel is ‘human in and human out,’” she said. “What you put in should be uniquely human, and what comes out should be vetted by a human.”

The truth of the matter, teachers and administrators said, is that the workload of their roles is not decreasing, and that technology has the potential to significantly reduce time spent on compliance tasks, like mandated teacher observations. As districts and schools seem likely to continue using the technology this way, they may need to create formal policies and guardrails.

“I would love to do more mentorship. I would love to do more pop-ins … more than just the required ones that we have to do,” Garrick said. “It definitely does save me time, because of all the other responsibilities that I have as an administrator … because at the end of the day, I’m still trying to provide [teachers] feedback, and I’m still trying to help them improve their teaching, and so AI can help me, especially with some of our newer teachers. They have a lot of things that they need to work on.”

Lastly, both Lawson and Weil stressed the importance of data governance, especially given the sensitivity of classroom and student data.

“We have to make sure that it is a walled garden, that the data does not go out to anywhere but stays with us. We need to make sure it’s FERPA and COPPA compliant, and we should not be using any tools that are not, and so we really have to dig deep into that safety measure,” Lawson said. “Especially when you are working with children, you really, really need to make sure that the quality assurance is there.”