How can AI help Michigan evaluators write teacher evaluations?
How can AI help Michigan evaluators write teacher evaluations?
Michigan reshaped teacher evaluation in 2024, and any honest answer about AI has to start there. Under Public Acts 224 and 225 of 2023, effective July 1, 2024, the state cut the student-growth share, changed the rating labels, and handed tool selection back to local bargaining. What did not change is the shape of the work: a Michigan evaluation still rests on classroom observation measured against a state-approved evaluation tool. AI can genuinely help with that observation write-up — mapping evidence to the right component and drafting feedback in the framework's language, toward Michigan's current rating levels. What it can't do is measure student growth, assign the year-end rating, or replace an evaluator's judgment. This post walks through where AI helps, where it doesn't, and what to look for.
What Michigan's system actually is
Under MCL 380.1249, every Michigan district, intermediate school district, and public school academy must adopt a rigorous, transparent, and fair evaluation system. The Michigan Department of Education (MDE) maintains a list of state-approved teacher evaluation tools, and districts choose one from that list, modify one, or build their own within state requirements. As of 2024, that choice is subject to collective bargaining.
The state-approved teacher evaluation tools are the Charlotte Danielson Framework for Teaching, the Marzano Teacher Evaluation Model, the Marzano Focused Teacher Evaluation Model, the Thoughtful Classroom, and the 5 Dimensions of Teaching and Learning (5D+). They differ in structure, but they share a common purpose: describing effective professional practice in observable terms across levels of performance.
What changed in 2024
Three changes matter most for an evaluator.
Student growth was cut and capped. Before 2024-25, 40% of the year-end evaluation had to be based on student growth and assessment data. Beginning in 2024-25, that share can be no more than 20%, with the exact percentage set through collective bargaining. The remainder — now the clear majority of the rating — is based primarily on the teacher's performance as measured by the district's evaluation tool.
The rating labels changed. The former four ratings — highly effective, effective, minimally effective, and ineffective — were replaced, as of July 1, 2024, with three: Effective, Developing, and Needing Support. The change was meant to reduce competition and emphasize growth. A teacher who isn't given a year-end evaluation is deemed Effective, and a teacher rated Developing or Needing Support receives an individualized development plan with progress expected within a set period.
Bargaining came back. Tool selection, the student-growth measures, and the evaluation process are now subject to collective bargaining at the local level, and evaluations may no longer be used to inform tenure or certification decisions.
The four Danielson domains
Because EvalScribe is built around the Danielson Framework, one of the approved tools, it's worth naming the four domains that anchor the observation write-up:
Planning and Preparation. Knowledge of content, students, and resources, and coherent instructional design and assessment planning.
Classroom Environment. A culture of respect and rapport, clear expectations, managed routines, and a space that supports learning.
Instruction. Clear communication, strong questioning and discussion, engaging students in the work of learning, and using assessment to adjust in the moment.
Professional Responsibilities. Reflection on practice, accurate record-keeping, communication with families, professional growth, and contribution to the school community.
The other approved tools organize practice differently, but the underlying domains of teaching they describe overlap heavily.
The observation requirements that drive the workload
Michigan is specific about observation. An evaluation year requires at least two observations, each a minimum of 15 minutes, with at least one unscheduled, and written feedback provided within 30 days of each. All evaluators must complete rater reliability training. Effective teachers can move to a biennial or triennial cycle after three consecutive effective ratings, subject to bargaining — but for teachers on the annual cycle, and for anyone on a development plan, the observation-and-feedback rhythm is constant.
That 30-day written-feedback requirement, repeated across a caseload, is where the documentation hours pile up.
Where AI helps with the Michigan write-up
The strongest fit between current AI and Michigan's system is that observation documentation.
AI handles three parts of it reliably. First, it turns fragmentary notes into coherent, evidence-anchored prose in the framework's language — the difference between "kids explained their reasoning to each other" and a sentence tied to the instruction domain. Second, it maps evidence to the right component, including evidence that supports more than one. Third, it holds consistency across a caseload, so the same quality of evidence lands in similar territory from teacher to teacher rather than drifting with the hour of the night.
For Michigan specifically, the value is drafting in the Danielson framework's own component language and toward the current rating levels — Effective, Developing, Needing Support — so the written feedback is ready inside the 30-day window rather than something you have to translate first.
Where AI doesn't help — and what stays with the evaluator
The honest scope follows from how Michigan built the system.
AI cannot measure student growth or assign the year-end rating. Tools like EvalScribe draft the professional-practice piece — the observation portion. The student growth and assessment portion (no more than 20%) and the combination into the year-end rating are handled in your district's process.
AI cannot supply an evaluator's professional and local judgment — what the evidence really shows, how a teacher's year unfolded, how a rating fits the context. That stays with the evaluator, who remains the last set of eyes on every rating and comment. Michigan reinforces this by requiring rater reliability training for every evaluator.
And general-purpose AI in particular is a step behind on Michigan. Paste notes into a consumer chatbot and it will reach for the retired highly-effective-to-ineffective labels, assume the old 40% growth weighting, and invent components. For a document tied to a development plan and a teacher's standing, those errors are the kind that surface at the worst possible moment.
What to look for in an AI tool for Michigan
A few questions worth asking before committing a tool to this work.
Does the tool draft toward Michigan's current ratings — Effective, Developing, Needing Support — rather than the retired four?
Does it actually know the domains and components of the tool your district bargained, or does it produce generic "good teaching" language?
Does it stay in its lane, drafting professional practice and leaving the ≤20% student growth and the year-end rating to your district?
Does it help you hit the practical requirements — turning at least two observations into written feedback within 30 days?
Where does your observation data live? Is it stored on the vendor's servers, or used to train models?
How EvalScribe handles Michigan evaluations
EvalScribe is built around the four Danielson domains. An evaluator captures notes by typing, dictating, or photographing handwriting (Smart Scan OCR converts it to text), EvalScribe maps that evidence to the component it supports, and drafts a rating and evidence-anchored feedback for each — toward Michigan's current levels: Effective, Developing, Needing Support. Every rating and comment is fully editable before you finalize it, and the evidence stays traceable to the note it came from, which is what makes the 30-day written-feedback requirement manageable across a caseload.
Two scope notes, because Michigan's structure calls for them. First, EvalScribe drafts the professional-practice piece — the observation portion, now the majority of the rating. The student growth and assessment portion (no more than 20%) and the combination into the year-end rating are handled in your district's process, not the app. Second, Michigan districts bargain which approved tool they use; EvalScribe is built around the Danielson Framework and the professional-practice domains these tools share, so it fits districts using Danielson most directly. Confirm your district's bargained tool.
Beta testers report saving 30 to 60 minutes per evaluation versus writing the documentation by hand. Across a full caseload and Michigan's repeated observation-and-feedback cycle, that adds up to dozens of hours back — hours that can go to the conversation the ratings are meant to support. More detail on the Michigan workflow is available at evalscribe.com/michigan.
Frequently asked questions about Michigan evaluations and AI
What evaluation tools does Michigan use? MDE approves a list — the Danielson Framework, the Marzano and Marzano Focused models, the Thoughtful Classroom, and the 5 Dimensions of Teaching and Learning. Districts bargain one, modify one, or build their own. EvalScribe is built around Danielson.
What are Michigan's current rating levels? Since July 1, 2024: Effective, Developing, and Needing Support — replacing the former four. EvalScribe drafts toward these.
How much of a Michigan evaluation is student growth? No more than 20%, down from 40%, with the exact share set by collective bargaining. The majority is the observation-tool portion.
Does EvalScribe assign the year-end rating? No. It drafts the professional-practice piece; student growth and the year-end rating happen in your district's process.
What if my district uses a tool other than Danielson? EvalScribe fits Danielson districts most directly. For Marzano, the Thoughtful Classroom, 5D+, or a local tool, email [email protected].
If you're evaluating teachers in Michigan, see how EvalScribe drafts in the Danielson domains and Michigan's current rating levels at evalscribe.com/michigan. Questions, or a school or district license? Reach the team at [email protected].
References
Michigan Department of Education, Educator Evaluation (state-approved tools; guidance on the 2024 changes)
Michigan Legislature, MCL 380.1249 (rating levels of Effective, Developing, Needing Support beginning July 1, 2024; student growth requirements)
Michigan Department of Education, Summary of Changes: Educator Evaluation Law (Public Acts 224 and 225 of 2023; collective bargaining; observation requirements)
Related articles
Page maintained by Anthony D. Neely, Ph.D. — practicing K-12 educator with nearly 20 years in the classroom, 2025–2026 Walker County Distinguished Teacher of the Year, and co-founder of EvalScribe. Framework details verified against the Michigan Department of Education's educator evaluation guidance and MCL 380.1249, as amended by Public Acts 224 and 225 of 2023.
Last updated on August 2, 2026.
