How do you tell a coachable behavior from someone's personality?
Ask whether you can state the change as a specific thing to say or do at a specific moment. "Ask for the appointment before discussing price" is coachable. "Be warmer" is not; it describes a person. If you cannot write the instruction as an action attached to a trigger, you have named a trait, and coaching to traits produces resentment without behavior change.
The action-plus-trigger test
Good coaching instructions have two halves: what to do, and when. "When the customer asks what it costs, answer the question first, then give one reason for the number, then propose a time." A rep can execute that on the next call and you can verify it on the recording.
Trait feedback has no trigger and no action. "Sound more confident" tells someone that you dislike an aspect of who they are and leaves them to guess at a remedy. Most people guess wrong, then perform a caricature of the trait, which is worse than the original.
Most traits have behaviors hiding inside them
Before discarding a trait observation, unpack it. Traits are usually accurate perceptions with the mechanism left out, and the mechanism is coachable even when the trait is not.
- "Sounds rushed" usually means talking over the customer's first sentence, no pause after asking a question, skipping the read-back.
- "Sounds cold" usually means going straight to data collection without acknowledging the problem that prompted the call.
- "Not confident" usually means hedging language around scheduling: maybe, I think, possibly, we might be able to.
- "Too chatty" usually means the appointment was never proposed until the customer forced it.
Some traits really are not coachable, and that is a different decision
Occasionally you unpack a trait and find nothing underneath but the person. Someone genuinely dislikes selling, or cannot tolerate conflict, or does not enjoy talking to strangers all day. No coaching cadence fixes that.
That is a role fit conversation for a manager who knows the person, made with judgment and history, not a number. No evaluation system should be making that call; scoring only supplies evidence a human uses. We are explicit about that boundary in how coaching data is presented.
Why automated coaching notes need this filter
A model asked to summarize a rep's weaknesses will cheerfully produce trait language, because trait language is what most written performance feedback looks like in the wild. Left alone, it will tell you a rep lacks warmth.
The fix is in the configuration, not the model: require every coaching note to name an observable moment and a specific alternative action, and reject any output that cannot. Constraining what the system is allowed to say is ordinary integration work, and it is the difference between operational AI and a text generator with your call data in it.
Topics: coaching · behavior change · feedback · rubric design · management
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.