An agent that helps a manager prepare for, and reflect on, coaching conversations with their team — without ever joining the conversation itself.
The chronic gap between "managers should coach" and what actually happens in 1:1s. Coaching skills programs teach models (GROW and friends) that decay within weeks. This pattern moves the support to the moments that matter: right before and right after real conversations.
"I have a 1:1 in an hour and I know I'll slip into problem-solving mode. Help me show up as a coach."
A copilot for the manager's development — not a monitoring tool, not a script generator, and never present in the actual conversation. Before: a preparation partner. After: a reflection partner. The person being coached never interacts with, and is never profiled by, the agent.
- Trigger — an upcoming 1:1 or coaching moment (calendar-triggered or manager-initiated).
- Observe — manager describes the situation and what they want for the other person, in their own words. No HR data, no performance records.
- Act — agent helps the manager plan an opening question and a thing to resist (usually: jumping to solutions). Optionally a 2-minute rehearsal of the opening.
- Learner response — the real conversation happens, agent absent.
- Adapt — afterwards, the agent asks three questions: What did they say? What did you resist or fail to resist? What surprised you? Patterns across debriefs become the manager's development thread.
| Data | Source | Sensitivity |
|---|---|---|
| Manager's description of the situation | Manager, in their words | High — treat as confidential to the manager |
| Manager's own development goals | Manager | Medium |
| Debrief history | Previous sessions | High |
Deliberately excluded: the report's performance data, notes, or records of any kind. The agent knows what the manager tells it, nothing more.
- Shadow dossiers — accumulating descriptions of a team member across sessions amounts to profiling someone who never consented. Anchor memory on the manager's behaviors, not the report's.
- Scripted authenticity — over-preparation produces managers reciting coach-flavored lines. Prepare one question and one restraint, not a script.
- Substitution — managers debriefing with the agent instead of building judgment. Watch for dependence: the goal is a shrinking need for prep.
- Manager's talk-time in 1:1s drops (self-reported or, where consented, measured).
- Debrief answers show increasing specificity about what the other person said — evidence of actual listening.
- Team members report better 1:1s in normal pulse surveys, without knowing the tool exists.
- Warning signal: prep sessions getting longer over time instead of shorter.
You are a coaching copilot for managers. You help them prepare for and
reflect on coaching conversations. You are never in the conversation and
you keep no notes about the other person.
BEFORE a conversation, help the manager get to exactly two things:
- one open question to start with
- one habit to resist (name their default: advising, rescuing, filling
silence). Offer a 2-minute rehearsal of the opening if useful.
AFTER, ask three questions, one at a time:
- What did they say — actually say, not what you concluded?
- Where did you resist your default, and where did it win?
- What surprised you?
Connect what you hear to patterns from previous debriefs about the
MANAGER's behavior only.