AI Is Coming for Teacher Evaluation and Most Principals Aren’t Ready for What That Means

Let me name the conversation that most principals are not having with their district yet but should be. AI tools are entering the teacher evaluation process. In some districts they are already there, generating rating recommendations from observation data, flagging patterns in walkthrough notes, summarizing evaluation portfolios. In most districts they are coming within two years.

And the vast majority of principals are navigating this without a clear professional position on what those tools can and cannot assess — and without having had the conversation with their supervisor about how their contextual professional judgment fits into a process that is increasingly mediated by an algorithm.

That conversation needs to happen before the first AI-assisted evaluation cycle runs, not after. After is too late.

What AI Evaluation Tools Actually Measure

AI evaluation tools work by analyzing structured observation data against rubric criteria and generating ratings based on what was documentable in the observation. They are, at their best, accurate and consistent processors of the information they were given.

That is also their specific limitation. They process what they were given. They cannot account for what they were not given.

Here is a partial list of what AI evaluation tools cannot account for: the teacher who is three weeks into a personal crisis that you are aware of and that is temporary. The relationship between that teacher and that classroom that took two years to build and that produces the specific quality of student engagement the rubric element is trying to capture. The instructional decision the teacher made in response to the specific student who walked in that day carrying something difficult. The professional trajectory the teacher is on — the measurable difference between what you observed this semester and what you observed last year because of a specific coaching investment you made together. None of that is in the observation data. All of it is in you.

The Documentation Practice That Protects You and Your Teachers

When an AI tool produces a rating or recommendation that influences a personnel decision, note in the professional record that the recommendation was considered alongside your own contextual knowledge of the situation, and name specifically how your professional judgment modified or confirmed the AI’s output.

“The observation platform flagged a low engagement rating for this teacher. My direct knowledge of the specific student dynamics in that classroom on that morning, and the teacher’s deliberate pedagogical decision to hold discussion rather than activity, led me to contextualize the rating and record a different professional assessment.” That sentence is your protection. Build the habit of writing it now.

The Conversation to Have Before You Have To

The proactive version of this conversation with your supervisor: “I want to have a conversation about AI-assisted evaluation before it becomes a reactive one. As these tools become part of how we process observation data, I want to establish clearly how my professional contextual knowledge of each teacher is going to be part of the evaluation record — not as a counterpoint to the data, but as an essential component of it. Is that built into the process, or is it something we need to build in?”

The principal who has this conversation first is the one whose professional judgment gets built into the process. The one who waits for the system to define the role of professional judgment is the one who finds it has been defined without them.

This is not resistance to AI in evaluation. It is the appropriate professional assertion that a principal’s contextual knowledge of their building is irreplaceable input that no algorithm has access to. That assertion needs to be made proactively and documented explicitly. Do it before the first cycle runs.