Resource Hub

July 20, 2026 · Last updated on August 4, 2026

How to get started guide: Conducting performance reviews with ChatGPT Work

# Government

# AI Techniques

Use ChatGPT Work to draft clearer, more consistent performance reviews while keeping manager judgment at the center.

Laura Keenan

The leadership shift

Performance management should not be a memory test for government leaders. The leader's job is to build a fair, specific, and timely understanding of impact from evidence gathered throughout the year.

AI should not replace human judgment or make employment decisions. Use it as an evidence assistant: organize work artifacts, surface patterns, translate activity into outcomes, identify coaching moments, and help leaders recognize progress while it is fresh.

Old habit	AI-enabled leadership habit
Review performance at the end of the year.	Capture evidence weekly and use it for ongoing coaching.
Reward the loudest or best-documented work.	Look for visible and invisible contributions with the same discipline.
Describe activities.	Translate activities into outcomes, scale, stakeholders, and mission value.
Use AI to write reviews faster.	Use AI to build better evidence and create room for better coaching.

How government leaders should think

Direction	What it means in practice
Start with evidence, not memory.	Treat recollection as a starting point, not proof.
Look for impact, not activity.	Push every accomplishment toward outcome, stakeholder, risk reduced, time saved, capability created, or mission value.
Make invisible work visible.	Mentoring, influence, risk prevention, cleanup, and morale-building often drive value but rarely appear in review packets.
Create a continuous record.	Weekly summaries become monthly themes. Monthly themes become quarterly narratives.
Separate facts, interpretation, and decision.	AI can organize the first two. Leaders own the third.
Keep the employee in control.	Employees should know what sources are reviewed and be able to add context or correct errors.
Coach while work is actionable.	The highest value is timely recognition, redirection, and development.
Protect trust.	Do not use AI evidence gathering as covert surveillance.

Operating model

The strongest AI performance management systems are lightweight and rhythmic. Leaders need a repeatable habit, a clear evidence standard, and a shared understanding that AI supports coaching rather than replacing it.

	Leader action	AI output
Monthly	Identify patterns, progress against priorities, and work that is not yet visible enough.	A monthly impact summary and skill signals by person or team.
Quarterly	Compare evidence to role expectations, growth goals, and team objectives.	A quarterly narrative, promotion readiness indicators, and development plan inputs.
Annual	Use the year-long record to ground the formal review and career conversation.	A reviewed evidence portfolio and draft language for the human-led review.

Before prompting

Setup instructions

Step	Explicit direction
Define the north star.	Write down role priorities, team objectives, leadership expectations, and what great impact looks like.
Map approved sources.	Identify sources such as meeting notes, deliverables, project plans, customer feedback, peer recognition, and weekly summaries.
Set boundaries.	Clarify what will not be reviewed, what can be excluded, and how sensitive information will be handled.
Create the ChatGPT Work Project.	Use one Project for performance evidence. Add standing instructions, criteria, and guardrails.
Capture close to the work.	After a meeting, launch, customer interaction, or hard decision, add a short note before context disappears.
Review outputs before use.	Check for exaggeration, missing context, bias, unfair comparison, and unsupported claims.

Prompting doctrine

Every performance management prompt should include context, inputs, evaluation lens, output format, and guardrails. If one is missing, the AI may produce language that sounds polished but is not useful enough for leadership judgment.

Prompt part	What to include
Context	Person, role, time period, goals, and intended use.
Inputs	Notes, artifacts, feedback, metrics, and constraints.
Evaluation lens	Impact, influence, outcomes, growth, collaboration, customer value, risk reduction, and alignment to objectives.
Output format	Evidence bullets, coaching notes, review language, recognition drafts, or questions to ask.
Guardrails	Base claims only on evidence, separate facts from inference, flag missing context, and do not make the decision.

Reusable setup prompt

Master instruction

A. Master setup instruction

You are my AI performance evidence partner. Help me organize evidence of employee and team impact so I can become a better coach and a fairer leader. Base analysis only on approved sources. Do not make employment decisions, assign ratings, or inflate claims. Separate confirmed evidence, reasonable interpretation, and open questions. Look for visible outcomes and invisible contributions, including mentoring, influence, risk prevention, trust-building, operational cleanup, and work that made others more effective. Translate activity into impact by identifying outcome, scale, stakeholder, result, skill demonstrated, and why the work mattered. Flag possible bias, missing context, recency effects, and places where I should ask the employee for their perspective.

Copy and adapt

Prompt library

1. Memory to evidence audit

Use when: preparing for a review, calibration, or talent conversation

Review the materials below for [employee/team] during [time period]. Identify specific evidence of performance and impact. Organize the output into confirmed evidence, reasonable interpretations, and open questions. Flag recency bias, loud work, and missing context.

2. Performance evidence portfolio

Use when: building a durable record for coaching, reviews, awards, or promotion

Build an evidence portfolio for [employee/team]. Organize accomplishments, outcomes, stakeholder influence, feedback, skills demonstrated, problems solved, and growth areas. Include source, why it matters, confidence level, and proof gaps.

3. Invisible work finder

Use when: important work may be quiet, preventive, or enabling

Look for quiet leadership, mentoring, influence, risk prevention, operational cleanup, unblocking others, and work that made teammates more successful. Summarize the strongest examples and how I might recognize or reinforce them.

4. Weekly leadership reflection

Use when: turning the week into usable evidence and coaching signals

Based on the notes below, identify what the team accomplished, who made meaningful progress, feedback received, risks or blockers, and what should be recognized or coached while fresh. Give top evidence points, one recognition action, and one coaching action.

5. Documentary, not snapshot

Use when: reducing recency bias and showing development over time

Create a performance narrative for [employee/team] across [time period]. Show the full arc: early progress, turning points, repeated strengths, challenges addressed, outcomes achieved, and evidence of growth over time.

6. Activity to impact translator

Use when: someone has listed tasks but the leadership value is unclear

Translate these activity-based accomplishments into impact-based statements. For each item, identify outcome, audience served, scale, mission value, measurable result if available, stakeholder influence, and skill demonstrated.

7. Career asset generator

Use when: helping the employee tell their career story

Using the evidence below, create career-ready accomplishment statements: review language, resume bullets, and LinkedIn-ready achievements. Keep claims specific and grounded. Flag sensitive or unsupported content.

8. Skill map and capability radar

Use when: understanding team strengths and development gaps

Analyze the evidence and identify demonstrated skills across the team. Where do we have hidden strengths, capability gaps, or people ready for expanded responsibility? Separate strong evidence from weak signals.

9. Coaching conversation builder

Use when: turning evidence into a fair, human conversation

Prepare a coaching conversation for [employee]. Identify what to recognize, reinforce, raise as a growth opportunity, and ask to understand their perspective. Include an opening, three talking points, and a clear next step.

10. Progress and motivation prompt

Use when: recognizing progress before the finish line

Identify progress worth celebrating this week: wins, breakthroughs, improved behaviors, lessons learned, and movement toward important goals. Draft a short recognition note and one follow-up action.

Leader review checklist

Question	Why it matters
Can I point to the evidence?	A polished claim is not enough. The leader should trace it to a source or observation.
Did I separate fact from interpretation?	This prevents AI language from making assumptions sound like settled truth.
Have I considered context and opportunity?	Account for role scope, access to visible work, blockers, and available support.
Did the employee have agency?	The employee should understand the process and add context or correct errors.
Am I using this for coaching?	The highest value is earlier recognition, clearer expectations, and better development conversations.
Is any content sensitive?	Customer, partner, personnel, confidential, or personally sensitive details may need removal or generalization.

Getting started

30-day implementation plan

Week	Leader action	Definition of done
Week 1	Define success criteria, approved sources, privacy boundaries, and Project instructions.	A shared performance evidence setup exists.
Week 2	Run the first weekly reflection prompt.	At least one weekly evidence report is reviewed.
Week 3	Run invisible work, activity-to-impact, and recognition prompts.	One recognition action and one coaching action are completed.
Week 4	Review the month and refine evidence capture.	A monthly summary exists with next focus areas.