Feedback Drafting Agent
A feedback drafting agent is an AI agent for manager enablement & coaching that helps a manager turn a vague impression into specific, behavioral feedback: what was observed, what the impact was, what would be better next time. Flags vague or loaded language and suggests how to say it so the person can hear it. Works for praise as well as correction.
How does the feedback drafting agent work?
What flows in, what the agent does with it, where a person decides, and what comes out.
Reads from
Manager's own observations · Feedback framework and examples · Company values and language guide
AI agent · runs when a manager brings an observation to turn into feedback
Feedback Drafting Agent
A person decides
Manager edits and delivers the feedback themselves
Produces
Draft feedback message · Talking points · Flags on vague or loaded language
What does the feedback drafting agent do?
Helps a manager turn a vague impression into specific, behavioral feedback: what was observed, what the impact was, what would be better next time. Flags vague or loaded language and suggests how to say it so the person can hear it. Works for praise as well as correction.
What does it produce?
A draft feedback message or talking points the manager edits before delivering
Who decides?
The manager decides what feedback to give and delivers it themselves. The AI drafts and coaches on wording; it does not send feedback or write anything into a performance record.
What systems does the feedback drafting agent connect to?
Examples of the kind of systems this agent would read from or write to, so you can picture it in your own stack. The actual set is whatever you run.
-
Chat
where the manager drafts
-
Knowledge
framework and language guide
-
Performance tools (kept separate)
not connected; the manager copies only if they choose
What data does it need?
- the manager's own observations
- feedback framework and examples
- company values and language guide
How would you measure it?
feedback instances per manager, monthly, self-view and aggregate; recipient-rated specificity and usefulness, quarterly pulse; share of drafts flagged for loaded language, monthly, aggregate
What does a first proof look like?
Thirty managers use it for a month, for praise as well as correction, and self-rate each draft before and after editing. Ask a sample of recipients how specific and useful the feedback was.
You'd call it working when
Drafts get more behavioral and less loaded, and managers give feedback more often than the month before.
What usually goes wrong?
- Drafts flowing straight into a review tool; keep the agent off the record
- Overwriting the manager's voice; the message must sound like them
- Using it only for correction; praise builds the habit
What are the guardrails?
- Never sends feedback or writes into a performance record
- Drafts private to the manager and not retained as HR data
- Reads no HRIS, ratings or prior reviews; only what the manager types
- Because drafts can end up in files, notify employee representatives before rollout and route formal-process wording through HR
- Flags language that could read as discriminatory and explains why
What leaves your boundary is set per build; the inputs above are the ceiling, and where the model runs, what it retains, and the DPA are agreed with your security team before anything is connected.
Our read
Widely deployed and well understood. Low-risk to build well.
Some ATS, HRIS or LMS suites ship a version of this. Where yours already does the job well, switch it on. The agent earns its place when the native feature is missing, rigid, or does not respect your rules; the strategy month is where that call gets made.
Produces written statements about a named individual's performance that can end up in reviews or disciplinary files
Where it sits in the order
Low data need but high stakes; roll out after the coach has proven privacy, with clear rules on what stays out of the record
Is a Feedback Drafting Agent worth building for your function?
That depends on your numbers, your data, and what else is on the map for you. The strategy month works that out.
Thirty minutes. Bring the number this would move.