EmpowerED AI Learning
"The future of work starts with you."
Build Your Performance Evaluation AI System
A step-by-step setup guide · AI for Managers & Supervisors · Week 3
What you are building
A three-layer system that turns your performance evaluation process from a 3–4 hour administrative task into a documented, repeatable workflow your AI runs with you every time.
A spreadsheet where you log observations on your team all year long — not just at eval time.
Your AI tool connects to that data. You may also have manual inputs from company systems that don't natively integrate.
A configured Gem, Project, or Custom GPT that holds your instructions and runs the same way every time.
Choose your tool ecosystem
Create your Team Tracking Sheet
This is your data home. It lives in your files all year long and becomes the single source of truth your AI draws from when eval season arrives. Log observations on your team members as they happen — not scramble to remember them later.
Set it up
- Create a new spreadsheet. Name it something clear — Team Eval Tracker 2026 or similar.
- Create one tab per direct report, or one sheet with a column for employee name.
- Add these columns as a starting point:
| Date | Employee Name | Category | Observation / Note |
|---|---|---|---|
| 6/14/2026 | Marcus T. | Strength | Led the safety briefing without prompting — strong ownership |
| 5/28/2026 | Jordan L. | Growth area | Missed two deadlines — discussed in 1:1, committed to new system |
Connect a form to your tracking sheet. The form feeds directly into the sheet as a new row. Add a shortcut to that form on your phone's home screen. Next time something noteworthy happens, open it in seconds, voice-to-text your observation, and hit submit. It lands in your tracking sheet automatically.
Google: Create a Google Form → link it to your Sheet using the Responses tab → add the form link to your phone home screen.
Microsoft: Create a Microsoft Form → link it to your Excel file → add the form link to your phone home screen or pin it in Teams.
Other: Any form tool that exports to a spreadsheet works — Typeform, Airtable, or a simple notes app you export periodically.
Where it lives
Connect your AI to the data
Tell your AI tool where to find the information it needs. Connect it at the highest level possible so it can pull from your tracking sheet, calendar, and meeting notes without you copying and pasting every time.
What to connect — and how
Manual inputs — company systems that don't natively connect
Most organizations have at least one system that lives outside your AI tool's reach — attendance records, payroll data, a timekeeping portal, or another HR platform. Before you run an eval, identify which data you need from those systems, pull it manually, and have it ready to paste in when prompted. These manual inputs become part of your documented workflow — you will know exactly what to gather and when every time.
Build your AI Specialist
Paste the starter prompt below into your AI tool's specialist configuration and customize the pieces specific to your organization. Once set up, you won't need to rewrite any of this — it runs the same way every time.
Where to build it
The starter prompt
Copy everything below and paste it into your AI Specialist's instructions field. Then swap out the bracketed sections for your organization's specifics.
You are an AI Specialist configured to help me write annual performance
evaluations for my direct reports.
Your job is to synthesize data from multiple sources into a professional,
objective, actionable evaluation draft that I will review and finalize.
INPUTS YOU WILL USE:
1. My Team Tracking Sheet — [INSERT FILE NAME HERE] — which contains
observations, strengths, and growth areas I have logged throughout the year.
2. My calendar, meeting notes, and 1:1 records from the past 12 months.
3. Manual data I will provide from company systems that don't connect
natively — such as attendance, timekeeping, or HR records.
4. A brief brain dump I will give you with my current observations
and any specific goals I want to address.
RULES — FOLLOW THESE EVERY TIME:
- Zero hallucinations. Do not invent projects, metrics, or events.
If information is missing, write [Data Missing] — do not guess.
- Write in formal, objective, third-person narrative style appropriate
for a permanent personnel file.
- All growth areas must be paired with a specific, actionable next step.
- If any manually provided data is missing, stop and ask me for it
before proceeding.
- Do not include anything from draft plans that have not been confirmed.
OUTPUT FORMAT:
Generate the evaluation in this exact structure:
1. OVERALL PERFORMANCE SUMMARY
[INSERT YOUR ORGANIZATION'S FIRST SECTION NAME AND DESCRIPTION]
2. CORE COMPETENCIES AND ACCOMPLISHMENTS
[INSERT YOUR ORGANIZATION'S SECOND SECTION NAME AND DESCRIPTION]
3. ATTENDANCE AND RELIABILITY
State exact figures provided. Add a brief reliability note.
4. DEVELOPMENT GOALS FOR NEXT REVIEW PERIOD
Exactly two goals. Each must be specific and measurable.
TO START EACH EVALUATION, ASK ME FOR:
- The employee's full name and job title
- The review period dates
- Any manual data from company systems (attendance, timekeeping, etc.)
- My brain dump: current observations and goals for this employee
Do not begin drafting until I have provided all four items.
- File name: Replace [INSERT FILE NAME HERE] with the exact name of your tracking sheet.
- Output section 1: Replace with your organization's first eval section name and what it covers.
- Output section 2: Replace with your organization's second eval section name and what it covers.
- Additional sections: Add more output sections if your eval template has more than four. Match your format exactly.
Run your first evaluation
Your system is built. Here is how to run it for the first time — and every time after.
Before you open your AI tool
- Pull any manual data you need from company systems — attendance figures, timekeeping records, or other HR data your AI can't reach directly.
- Have the employee's name, job title, and review period dates ready.
- Give yourself two to three minutes to do a brain dump — what do you remember about this person's year? What are you proud of? What do you want them to work on? Jot it down or speak it.
Run it
- Open your AI Specialist.
- It will ask you for the four items. Provide them one at a time.
- Let it generate the draft. Do not interrupt.
Read the full draft before it goes anywhere. Check these specifically:
- Do the manual figures match exactly what your source system gave you?
- Are the strengths and accomplishments grounded in real, documented events?
- Has anything been invented or assumed? If so, correct it or mark it [Data Missing].
- Is the tone appropriate for a permanent personnel file?
- Would you sign your name to this?
Anything you catch here becomes a rule in your prompt next time.
Finalize and file
- Make your edits directly in the draft.
- Move the finalized text into your organization's official eval template.
- Deliver the evaluation to the employee.
- File the completed document according to your HR process.
Keep the system good
A workflow without maintenance drifts. Two habits keep this system sharp.
Log observations throughout the year
The tracking sheet only works if you use it. Add an entry after every 1:1 where something noteworthy happens — strengths, growth moments, project wins, missed commitments. Log it in the moment, not at eval time. Five minutes now saves two hours in December.
Pressure-test after every eval cycle
After you run each evaluation, ask yourself: what did the AI get wrong or miss? What did I have to correct? Each answer is a rule to add to your prompt. Your system gets more accurate every cycle — not because AI gets smarter, but because you keep closing the gaps.
Name an owner
If you share this system with your team, name the person who owns the prompt and keeps it updated. A system without an owner drifts. The owner updates the prompt when the eval format changes, when HR updates the template, or when the team finds a better way to structure the output.
What you walk away with
A performance evaluation system that runs the same way every time — grounded in data you collected all year, reviewed by you before it goes anywhere, and documented well enough to hand off.
This is Mode 3. The brief is written once. The system runs every time. You stay in the loop at every checkpoint.