The Performance Review Pack

What this is for

You have a dozen reviews due, a self-appraisal you keep avoiding, and one direct report whose feedback has read "needs to communicate better" for two cycles straight. This pack turns scattered notes into specific, defensible writing — for reviews you give, the one you write about yourself, and the conversation you're dreading.

Before you start

• Gather your raw material: notes, Slack threads, old 1:1 notes, last cycle's draft.
• Read the anonymization note below and do that pass before you open any AI tool.

What never goes into a public AI tool

• Full names. Use role tags instead — "the senior analyst," "Employee A."
• Comp, raise amounts, bonus targets, level/band. Add these by hand afterward, in the real document.
• Anything that IDs the company, if the case touches legal risk, a PIP, or termination — client names, codenames, org details.
• Medical, family, or accommodation info. Summarize impact ("attendance was inconsistent"), not cause.
• Anything already in a formal HR/Legal process. Use your company's approved system, or write it yourself.

Do the anonymizing in a separate text file first: names become role tags, dates become relative ("in March"), dollar figures get removed. The AI needs the substance, not the identifiers.

The six prompts

1. Writing a review from rough notes — either model.

ROLE: Management writing coach turning raw notes into a formal review.
CONTEXT: Notes below cover [TIME PERIOD]. Role: [ROLE, e.g. "mid-level
product designer"]. Do not invent accomplishments or issues not in my
notes.
NOTES: [PASTE ANONYMIZED NOTES — e.g. "Shipped checkout redesign 2 weeks
late, unclear specs from PM. Mentored two juniors well, unprompted
mentions from both. Missed Q2 planning deadline twice. Strong client
feedback on March presentation."]
CONSTRAINTS: Use only what's given — flag vague notes as questions
instead of guessing. Balance strengths and growth honestly, no padding.
OUTPUT FORMAT: Sections for Key Accomplishments, Areas for Growth, and a
summary paragraph. Flag anything too vague to use as written.

2. Writing your own self-appraisal with evidence — either model.

ROLE: Helping me draft my self-appraisal for this cycle.
CONTEXT: Role: [ROLE]. Goals this cycle: [LIST 2-4]. Evidence below, in
no order.
EVIDENCE: [PASTE — e.g. "Led the migration, under budget. Onboarded new
hire faster than usual. Missed June deadline due to vendor delay,
recovered by July. Presented to the board twice."]
CONSTRAINTS: No inflated verbs ("spearheaded," "revolutionized") — plain
language tied to outcomes. Keep accountability on misses, but state what
I controlled and what I didn't. Tie each item to a stated goal.
OUTPUT FORMAT: One paragraph per goal with evidence, plus a closing
paragraph on next-cycle focus.

3. Turning vague feedback into behavioral examples — either model.

ROLE: Converting vague feedback into specific, observable examples.
CONTEXT: Vague version: [e.g. "Needs to communicate better with
stakeholders"]. Fragments I remember: [e.g. "Didn't loop in Finance
before April's pricing change. Sent a May status update to the wrong
list. Didn't flag a delay on the June client call until asked."]
CONSTRAINTS: Turn each fragment into a dated, observable behavior — not
a trait. No "always/never" unless the fragments show a real pattern. If
I've given only one example, say so before calling it a pattern.
OUTPUT FORMAT: List in "situation — behavior — impact" format, ready to
read aloud.

4. Prepping for a review you expect to go badly — either model.

ROLE: Blunt, experienced HR business partner prepping me for a hard
review conversation.
CONTEXT: Delivering to [ROLE/RELATIONSHIP, e.g. "a direct report who's
been defensive before"]. Core message: [ONE SENTENCE, e.g. "Output has
been inconsistent, two deadlines missed with no early warning"].
Expected reaction: [e.g. "will argue deadlines were unrealistic"].
CONSTRAINTS: Give me the opening line, not a full script. Anticipate the
most likely pushback and give one honest response — not a deflection.
Don't suggest I soften the core message.
OUTPUT FORMAT: Opening line, pushback plus my response, three questions
to ask before I leave the room.

5. Writing a promotion case — either model.

ROLE: Building a promotion case document.
CONTEXT: Proposing [ROLE TAG] for promotion from [CURRENT LEVEL] to
[TARGET LEVEL]. Evidence of next-level work below.
EVIDENCE: [PASTE — e.g. "Took over vendor negotiation solo, saved 12% on
contract. Two peers requested to work on their projects specifically.
Mentored a struggling teammate back to good standing in one quarter."]
CONSTRAINTS: Map each item to a competency expected at the target level,
don't just list achievements. Flag any thin evidence so I can find more
before submitting.
OUTPUT FORMAT: One page — Scope of Impact, Evidence of Next-Level Work,
Identified Gaps.

6. Documenting a performance problem defensibly — either model.

ROLE: Meticulous documentation writer — this needs to hold up if it
becomes a formal process.
CONTEXT: Factual record only. Role: [ROLE TAG]. Timeframe: [DATE RANGE].
Incidents: [PASTE — e.g. "March 3: missed client deliverable, no advance
notice. March 20: verbal warning given, documented in 1:1 notes. April
15: missed a second deadline, same project."]
CONSTRAINTS: State only what happened and when — no motive, no character
judgments. Note where feedback or warnings were already given. Flag any
missing date or witness as a weak point.
OUTPUT FORMAT: Dated incident log, then a short pattern summary, written
so someone else could defend it in a meeting.

Where this goes wrong

• You paste real names and comp before thinking. Build the anonymize pass into your workflow — do it in a separate file before opening the AI tool at all.
• You let thin notes get filled in with confident-sounding invention. Feed more detail, or answer the "flag as a question" prompts instead of ignoring them.
• You copy the output straight into the official system. These are strong first drafts, not finals. Read each one like you're editing someone else's writing about someone you actually know.

The 2-minute version

Anonymize first: names to role tags, dollars out, dates relative. Pick the matching prompt. Paste real notes into the brackets. Edit the output before it goes anywhere official.

---

Where you stand with AI, in 10 minutes. The free AI Readiness Quiz gives you a personalized roadmap instead of another tool list: genxcelerate.com

Weekly, on Tuesdays. The Gen Xcelerator — practical AI intelligence for people who already know how to do the job. Subscribe free

Experience Is the API. — GenXcelerate