Practicing feedback, including with AI

Like listening or public speaking, feedback is a behavioral skill: you don't improve by reading, but by practicing under pressure. Yet chances to practice with honest feedback are rare — we almost never "rehearse" a challenging conversation before living it. AI offers an unprecedented testing ground here: available, patient, able to play roles and analyze your phrasing.

Why practice

Feedback fails mostly under the effect of emotion: facing someone who defends themselves, our reflexes (judging, assuming intent, attacking the person) take over. Only repetition installs new automatic responses. This is the principle of behavioral rehearsal: replaying a delicate scene in safe conditions before facing it. Bonus: you can test several phrasings and keep the one that lands best.

The classic methods

AI as a practice partner

An AI assistant can play three roles. Simulate a realistic recipient (a defensive employee, a touchy peer, a rushed manager) to practice with no stakes. Analyze your phrasing to detect pitfalls: judgment about the person, assumed intent, missing request, accusatory "you." Coach by offering rephrasings in SBI or DESC. The decisive advantage: replay the same scene ten times, raising the difficulty, with no embarrassment.

flowchart LR
    U[You describe the situation<br/>+ the recipient] --> IA[The AI plays<br/>the recipient]
    IA --> R[You give<br/>the feedback]
    R --> F[The AI evaluates:<br/>fact? impact?<br/>judgment? request?]
    F --> A[You adjust<br/>and replay]
    A --> IA

Three practice prompts to copy

1. Difficult-feedback simulation

"Play the role of a competent but defensive employee who systematically turns in reports late and tends to make excuses. Stay in character, respond as them after each line, don't analyze me during the scene. I'm practicing giving feedback in SBI. We'll debrief when I say STOP."

2. Pitfall detector

"Here is my feedback: '[my sentence]'. Spot the pitfalls: judgment about the person, assumed intent, accusatory 'you', no dated fact, no request. Explain each in one sentence, then propose a rewritten version in SBI + request format."

3. Receiving coach

"I'm going to give you a piece of feedback I just received that stung: '[…]'. Help me identify which of the three triggers is firing (truth, relationship, identity), find the usable grain of truth, and phrase an exploring question I could ask instead of defending myself."

Cautions and limits

AI is an excellent simulator, not a human: its reactions are more predictable, its emotions acted, and it does not perceive your real nonverbal cues (tone, gaze, posture) that weigh heavily in feedback. Three precautions: aim for authentic phrasing rather than "perfect" phrasing; always transfer to a human, the only real test; keep in mind that the goal is relationship and progress, not technical performance. Good AI coaching should make you clearer and more present, not more mechanical.

A three-week training plan

Week 1 — observation: self-debrief after every piece of feedback given or received, spot your dominant pitfall (judging? avoiding? justifying?). Week 2 — simulation: three AI sessions on difficult feedback, with a debrief on your phrasing. Week 3 — transfer: one real piece of feedback per day in SBI + request format, and when receiving feedback, one exploring question before any defense. Note each evening what changed in people's reactions.

Summary

Feedback is built through repetition: role-play, written script, self-debrief, role reversal. AI complements these methods by simulating difficult recipients, detecting your pitfalls (judgment, assumed intent, missing request) and also coaching the receiving side (triggers, grain of truth, exploring question) — provided you aim for authenticity, transfer to humans, and keep the relationship as the goal. A progressive three-week plan turns theory into reflex.