Coaching ≠ evaluation in the age of AI
A bright-line framework for CAOs — where AI can help coaching notes, and where summative evaluation must stay human.
Teachers can smell when a write-up was generated. Unions can smell it too. That is not a technology problem. It is a trust problem wearing a chatbot costume.
Instructional coaching and formal evaluation are not the same job. AI makes it easier to blur them — faster notes, cleaner prose, a finished paragraph before the coach leaves the hallway. Speed is not the same thing as legitimacy.
This is an opinion framework for network and school leaders. It is not a product pitch and not a claim about any named campus.
Two different jobs
Coaching is formative. It exists to improve tomorrow’s practice. Evidence from a walkthrough should become a focused next conversation and a next step the teacher recognizes as fair.
Evaluation is summative. It affects employment, ratings, and due process. The bar for evidence, authorship, and human judgment is higher — and it should stay higher.
If you let the same AI workflow draft both, you train everyone to treat coaching as soft evaluation. That kills the coaching relationship faster than any tool ever will.
Where AI can help (coaching side)
Use AI as a drafting and organization aid for formative work, with a human still holding the pen:
- Turn sparse notes into a readable coaching log — then edit until it sounds like you were in the room.
- Cluster look-fors across several visits so patterns are easier to discuss.
- Suggest questions for the debrief — never verdicts.
- Remind the coach what was agreed last time so the next visit is not a cold start.
Rule of thumb: if a teacher could reasonably ask “did you actually see that?”, a human must own the answer.
Where AI must not run the show (evaluation side)
Keep summative work human-authored and human-owned:
- Final evaluative ratings and narrative judgments — no auto-generated “proficient” paragraphs pasted into the record.
- Evidence the observer did not collect — models invent details; evaluation cannot.
- Language that implies a formal finding when the visit was informal coaching.
- Anything your contract, handbook, or bargaining agreement treats as personnel action.
If your policy is silent, treat silence as “evaluation stays human” until counsel and labor partners say otherwise.
A one-page bright line for CAOs
Print this. Share it. Argue about it on purpose.
| Activity | AI assist OK? | Human must own |
|---|---|---|
| Raw walkthrough jots → tidy coaching notes | Yes, with edit | Accuracy of what was observed |
| Suggested debrief questions | Yes | Which questions get asked |
| Pattern summary across coaching visits | Careful yes | Interpretation and priorities |
| Formal observation write-up for rating | No auto-draft as final | Entire evaluative narrative |
| Summative rating / personnel language | No | All of it |
| Sharing notes with the teacher | Only after human review | Tone, fairness, specifics |
Policy is not a PDF trophy
Gallup’s early-2026 teacher survey made the boring problem loud: most teachers still report little or no formal AI guidance. Writing a policy is not the same as running a practice.
If you only ban student chatbots and never say how coaches may use AI in notes, you leave the trust gap open. Name the bright line out loud — coaching vs evaluation — before the first awkward write-up becomes a grievance.
What good looks like in a network
- Coaches can explain, in one sentence, what AI is allowed to touch in their workflow.
- Teachers know coaching notes are formative and human-reviewed.
- Evaluation documents are authored by the evaluator, not polished by a model into something nobody said.
- Walkthrough tools stay separated from HR systems in both process and culture.
Software can help hold the trail from observation to next step. It cannot buy you trust after you outsource judgment.
If you want a direct conversation about coaching-support practice — including where software helps and where it should stay out of the way — talk with us.