Two Drafts of Feedback

Overview

Generative AI can turn a lesson transcript into a tidy feedback note in seconds, and some coaching platforms now offer to do it automatically. Coaches in this session read a short fictional transcript, score two unlabeled feedback drafts against a quality checklist, and guess which one a chatbot wrote. The reveal sets up the harder conversation: consent, student privacy, transparency with the teacher, and the line between an AI draft and a coach's judgment.

Objectives

• Participants will score feedback against criteria (specific, evidence-based, tied to the teacher's goal, one actionable next step) rather than by how polished it sounds.
• Participants will identify at least one claim in AI-drafted feedback that the transcript does not support.
• Participants will sort observation-related AI uses by the consent, privacy, and transparency conditions each requires and draft a disclosure statement for teachers.

Step by step

1. 10–4 min

   Hook

   Which one did a person write?

   Project the first two sentences of Draft X and Draft Y side by side. Ask everyone to vote with a hand: "Which draft came from a chatbot?" Don't reveal. Say: "Hold that guess. By the end, you'll see why the more important question is not who wrote it, but whether it's true and useful to this teacher."

   Facilitator noteRooms usually split or lean toward the more polished draft as the human one. Either way, the guess isn't the learning.

2. 24–16 min

   Explore

   Read the lesson, then score both drafts

   Everyone reads the transcript on Handout A and highlights, in one color, every moment where a student explains their thinking. Then pairs score Draft X and Draft Y separately with Handout B. For every "Yes," they must point to a line in the transcript that proves it. Pairs write their guess about authorship at the bottom of each checklist.

   Facilitator noteLook-for: pairs who go back to the transcript to check a claim in the feedback. That is the same verification habit we want teachers to teach with AI outputs.

3. 316–24 min

   Model

   The reveal and the trace

   Reveal that Draft Y was AI-drafted from the transcript and the teacher's goal. Model tracing its claims out loud: "It praises the turn-and-talk at the start. Let me find it. There is no turn-and-talk in the transcript." Then trace Draft X the same way and name its flaw too: specific and accurate, but it offers three next steps when a teacher can act on one. Ask: "Which draft would help Mr. Delgado more on Monday, and what would you change in each?"

   Facilitator noteDon't let this become "AI bad, human good." Draft Y also caught the short wait time, which is real. The point is that the coach owns every sentence the teacher reads.

4. 424–36 min

   Practice

   Sort the uses: consent, privacy, transparency

   Tables sort the Handout C strips onto three categories: Reasonable with consent and disclosure, Only with added guardrails, and Don't do it. For every strip in the middle column, the table must name the guardrail on a sticky note (for example, "remove student names first" or "teacher sees the AI draft and the coach's edits"). Then tables compare with the facilitator key.

   Facilitator noteExpect debate on recording video for AI analysis. Surface who has to consent: the teacher, and, when students can be heard or seen, the district's policy on student recordings and families. When in doubt, it belongs in the middle or right column.

5. 536–45 min

   Apply

   Write the feedback Mr. Delgado deserves

   Pairs write a new three-sentence note: one specific strength with evidence, one observation tied to his goal, and one next step. Digital path: the facilitator projects an approved chatbot, pastes the fictional transcript, and asks for a draft; pairs then mark which sentences they would keep, fix, or delete and why. Unplugged path: pairs revise Draft Y by hand into something they would sign.

   Facilitator noteAsk a pair or two to read their note aloud. Listen for whether the next step is something he can try in one lesson.

6. 645–50 min

   Reflect

   Our disclosure sentence

   Each person finishes this on the back of Handout B: "When I use AI in my coaching, I will tell teachers…" Invite three people to share. Close with the coaching question: "If you used AI to draft feedback, what would change in teacher practice that wouldn't have changed otherwise?"

   Facilitator noteStrong disclosure sentences name what data goes in, who sees the draft, and who makes the final call.

Adaptations

Campus leaders

Add a strip on using AI summaries in formal appraisal and discuss why coaching feedback and evaluation records need different rules.

Higher Ed

Use the same structure for peer observation of teaching. Add a strip on sending lecture recordings with student voices to an external AI service.

Teacher PLCs

Teachers score the drafts as the recipient: "Which would I want to receive?" Then write what they would want a coach to disclose.

Pairs well with

• Generative AI Adoption ChecklistUse it with leaders when a coaching platform offers built-in AI feedback.
• AI Custom Instructions LibraryIdeas for instructions that keep an AI drafting assistant tied to evidence.