The SBI Model
The Center for Creative Leadership's SBI model works because it anchors feedback to specific observable behavior rather than character, which reduces defensiveness by targeting actions rather than identity.
The specific context: when and where it happened. Grounds the feedback in reality rather than a generalization.
The observable action, not an inference about motive or character. What exactly did or did not happen?
The consequence: on the team, the project, or the stakeholder. Makes the feedback relevant and actionable.
The Human Edit
AI gives you a structured starting draft. You make it true, fair, and deliverable. These three steps are not optional.
Replace any "be more X" with a concrete behavior and situation. "Be more organized" becomes "In the Q2 project, the deliverable came in without the required appendices, which required a same-day turnaround from the reviewer."
Would you use the same language for a different person in the same situation? Does the draft judge personality rather than behavior? If either flag is yes, rewrite before delivering.
Adjust tone and specifics for this actual person and relationship. What is supportive for one person may feel clinical to another. The final edit is where the feedback becomes yours.
Bias Checkpoint
AI-drafted feedback can quietly carry the slant of its training data, and it will not tell you when it does. You are the accountability layer. Read every draft for language that targets personality instead of behavior, and for framing you would not apply equally across your team.
Sam's story
Sam needs to give feedback to an analyst who missed a deadline. Raw observation in, SBI draft out: Situation (the Q2 project), Behavior (deliverable came in three days late without an early flag), Impact (downstream team had to compress their testing window). Sam reads the draft and catches a line that drifted toward character ("seems disorganized"), cuts it, and replaces it with the specific behavior. Sam also adds the shared accountability: the scope crept and neither side flagged it early, so the feedback is fair. The AI gave structure. Sam made it just.
Apply It Today
Take one piece of feedback you owe someone, write three or four bullet observations, and ask AI for an SBI draft. Then run the vague-language and bias checks before you would ever use it.
Reflect