← All use cases
Use case 07

Manager's review tracker

Work through your team's pay review over several sittings. It shows the guideline, tracks your budget, flags what needs a reason, and checks for a skew between women and men before you submit.

Vibe-coded tool Remembers your work No AI needed to run it
Try itUse the real tool. Watch itFive steps, animated. Build itPrompts you can paste. Check the dataWhat is safe, and where. Take itDownload the finished file.

What it does and why it matters

A pay review is rarely done in one sitting. You start, get pulled away, and come back days later. A spreadsheet does not help much. The guideline sits on another tab, the running budget is a formula you have to trust, and a pattern across the team is easy to miss.

This tool keeps your place. It shows the rule beside each decision, adds up the spend as you go, and never makes the decision for you. You enter every number.

It is "stateful", which means it keeps your work between sittings, so it needs somewhere to keep it. This is the first tool on this site that keeps data, which is why the data check below matters more here than on any other page.

It pairs with use case 08, where the reward team loads what managers submit, and with use case 04, which helps you explain the outcomes afterwards.

1Try it

This is the real tool, running in your browser on an invented team. Enter some proposals, save your work to a file, and open it again.

Your file is read by your own browser. Nothing is uploaded. Your work is kept only in the file you choose to save.

Try this

  1. Enter proposals for a few people. Watch the budget bar and the fairness check update as you type.
  2. Push one proposal outside the guideline and see the flag. The person only counts as done once you give a reason.
  3. Press Save my work. Close the page, come back, and use Open a file to choose the file you saved. Your place is back.
  4. Finish everyone and download the submission file. Then load it into use case 08.

2Watch it

Five steps from opening your team to submitting it. Press Play, or step through at your own pace.

Step 1 of 5Invented data
Finance Operations (invented team)
E2041Senior Accountant95%
E2042Accounts Payable Officer86%
E2043Payroll Analyst98%
E2044Finance Manager112%
E2045Credit Controller105%

Each bar shows pay against the reference point. The dark line is 100%.

Open your team. Your reward team gives you a file. It lists each person by ID, with their pay, reference point and rating. No names needed.

E2041 Senior Accountant
Level 4 · Meets expectations
95%
Guideline 2.5% to 4.0%
Proposed increase (%)

The guideline beside each decision. The tool shows the rating, where the pay sits and the range that applies. The box stays empty. You decide the number.

Budget (invented figures)
Budget
$31,798
Spent
$19,700

Remaining: $12,098

E2045 Credit Controller: proposed 2.0%, guideline 0.0% to 0.0% Outside guideline: reason needed

Budget and flags as you go. The spend adds up as you type. A proposal outside the guideline is flagged, and that person is not done until you write a reason.

review-FIN-OPS-work.json
Saved to your folder
Opened review-FIN-OPS-work.json
4 of 9done

Save, and come back. Press Save and the file lands on your computer. Next sitting, open it and carry on from the same place.

Average proposed increase (invented figures)
Women
3.1% (4)
Men
3.0% (4)

All 9 people are done. Review the summary, then submit.

Download submission file

Check for a skew, then submit. Compare the two averages. A difference is a prompt to look again, not proof of a problem. When everyone is done, download your submission.

All figures in this walkthrough are invented, for illustration only.

3Build it

You can build your own version with any capable AI assistant your employer allows, for example Microsoft Copilot, Claude or ChatGPT. You describe what you want, it writes the code, and you check the answers.

Build it with invented data, never your real team. Expect a few rounds of "that's not quite right, change X". This one has more rules than most, so spell them out.

Step 1: Describe the tool and who uses it

Say who will use it, what goes in, and what the manager types. Ask the AI to plan before it builds.

I work in HR in Australia and I am not a developer. Please build me a tool for a line manager to work through their team's annual pay review over several sittings. It must be a single HTML file that I can save to my computer and open in a browser by double-clicking it. Requirements: - It runs offline. Use no external libraries, images or links, and send no data anywhere. - The user is one manager working on one team at a time, about 5 to 15 people. - Identify people by employee ID, never by name. - For each person, show: employee ID, role, level, gender (W, M or X), base salary, reference point (the middle of their pay range) and performance rating (Below expectations, Meets expectations or Exceeds expectations). - For each person, the manager enters a proposed increase as a percentage and a short reason in words. - Build in an invented sample team of 9 people so the tool shows something as soon as it opens, with no proposals entered yet. Label it clearly as invented data. - The tool must never suggest or fill in a number for the manager. It shows the rules, and the manager decides. - The audience is HR people and line managers, not developers. Use plain English and Australian spelling. Before you write any code, tell me in a few sentences what you plan to build, and ask me about anything that is unclear.

Step 2: Spell out the rules

Do not let the AI guess how a pay guideline works. Tell it exactly: how position in range is worked out, where the zones start and stop, what the guideline allows, and what counts as outside it, done and over budget. The table below is invented. Replace it with your own.

Now add the rules. Follow these exactly. Do not substitute your own, and do not guess anything I have not told you. 1. Position in range = base salary divided by reference point. Show it as a whole percentage, for example 96%. 2. There are four zones, chosen by position in range BEFORE the increase: "Below 90%", "90% to 100%", "100% to 110%" and "Above 110%". A position exactly on a boundary belongs to the higher zone, so exactly 100% is in "100% to 110%". 3. The guideline is the lowest and highest increase allowed, by rating and zone. This is an invented example. The four numbers in each line follow the zone order above. Below expectations: 0.0 to 2.0, 0.0 to 1.5, 0.0 to 0.0, 0.0 to 0.0 Meets expectations: 3.5 to 5.0, 2.5 to 4.0, 1.5 to 3.0, 0.0 to 1.5 Exceeds expectations: 5.0 to 7.0, 4.0 to 5.5, 2.5 to 4.0, 1.0 to 2.5 4. New salary = base salary x (1 + increase percentage / 100), rounded to the nearest dollar. 5. A proposal is outside the guideline when the increase, rounded to one decimal place, is below the minimum or above the maximum. Exactly on the minimum or the maximum is inside. 6. A person is "done" when an increase has been entered (0 is allowed) and, if the proposal is outside the guideline, the reason is not blank. 7. The budget is 3.5 percent of the team's total base salary. Spend is the total of (new salary minus base salary) for everyone who has a proposal. Show the budget, the spend so far, and what is left, or how far over budget the team is, in words and in a different colour. Show each person's guideline range beside their increase box, as "Guideline 2.5% to 4.0%". If a proposal is outside the guideline, show "Outside guideline: reason needed". Give each person a status of "Not started", "Needs a reason" or "Done". Put each rule in its own small function with a plain-English comment, kept separate from the code that draws the screen.

Step 3: Decide where the work is saved

A tool that remembers has to keep your work somewhere. There are two common choices. The first is in the browser, on one computer. That is quick to build, but the data sits in the browser where you cannot see it, it does not move to another computer, and on a shared computer it can be left behind for the next person. The second is a file you save and open. You can see it, move it, lock it down and delete it, and nothing is left behind in the browser. This site chose the file. It asks a little more of the manager, and gives the organisation a lot more control.

Now make the work last between sittings. Do NOT use the browser's own storage: no localStorage, no sessionStorage, no IndexedDB and no cookies. Instead: - Add a "Save my work" button that downloads a JSON file named like review-FIN-OPS-work.json. It holds the team details and every proposal and reason entered so far. - Add an "Open a file" button, with drag and drop as well, that reads a team file or a saved work file and puts the manager back where they were. - Show an "Unsaved changes" marker next to the Save button as soon as anything is edited, and clear it when the file is saved. After a save, show "Saved" with the local time. - If the manager tries to close the page with unsaved changes, warn them before it closes. - If the file is the wrong kind or cannot be read, show a plain-English message that says what it is and what the tool needs. It must never crash. - Show this line near the top: "Your work is saved only when you press Save. The file is kept on your own computer, wherever you choose to put it. Nothing is stored in this page or sent anywhere." When you finish, list every place in the code that could store or send data, and confirm there are none other than the downloaded file.

Step 4: Add the fairness check and the submission

The fairness check compares the average proposed increase for women and men, with a minimum group size so a small group cannot point to one person. The submission only switches on when everyone is done.

Now add a fairness check and a submission. Fairness check: - Work out the average proposed increase for women (W) and for men (M). Use a simple average of the percentages, shown to one decimal place, and count only people who have a proposal. - People recorded as X are counted in headcounts but are not part of the women-to-men comparison. - Show the comparison only when BOTH groups have at least 3 people with a proposal. Otherwise show "Too few people to compare" with the counts. Keep 3 as a setting that comes from the team file. - Show two labelled bars with the numbers and the headcounts, updating as the manager types. - Under it, write: "A difference is a prompt to look again at each decision, not proof of a problem. Compare like with like: rating and position in range drive the guideline." - Beside it, show the average proposed increase for each rating, so the manager can see whether a gap is explained by ratings. Submission: - Add a "Review and submit" section. If anyone is not done, list who and why, and switch the buttons off. - When everyone is done, show a printable summary: team, date, headcount, budget and spend, how many proposals are outside the guideline and the reasons given, and the fairness result. - Add three buttons: download a submission JSON file, download a CSV spreadsheet with one row per person, and print the summary. - The submission must repeat the team file's details and add, for each person, the increase, the new salary, the guideline minimum and maximum, whether it is outside the guideline and the reason, plus a submitted-at time. Never include any private label the manager typed to recognise someone.

Step 5: Test it like you mean it

A tool that keeps people's pay needs more testing than most. Ask the AI to help, then check the answers by hand.

Please help me check this tool properly. 1. Make a tiny test team of 4 people with round salaries and reference points, so I can check every answer by hand. 2. Before I look at the screen, tell me each person's position in range, zone and guideline range. 3. Include one person whose salary equals their reference point exactly, and tell me which zone they should be in. 4. Tell me what should happen when I enter 0, when I enter an increase outside the guideline, and when the team goes over budget. 5. Walk me through saving a file, closing the page and opening the file again, and tell me what I should see. 6. Tell me what the tool should say if I open the wrong kind of file. If anything does not match, say so plainly and fix it. Do not change the test data to make it fit.

A checklist to run yourself:

  • Save, close the page, open the file, and compare every number with what you had before.
  • Find a person exactly on a zone boundary (salary equal to reference point) and check they are in the higher zone.
  • Enter 0 for someone. They should count as done if 0 is inside their guideline.
  • Push the team over budget. Check the colour and the words change.
  • Open the wrong kind of file, for example a spreadsheet. You should get a plain message, not a crash.

What you would add next

  • A change history, so you can see who changed a number and when.
  • Names shown only on the manager's screen, matched from a separate list that is never saved into any file.
  • Locking after submission, so a submitted team cannot be edited by accident.

Good habits

  • Spell out every rule in your prompt. If you leave a gap, the AI fills it with a guess.
  • Test with a small team where you already know every answer.
  • Keep the prompts you used so you can rebuild or audit the tool later.
  • Get someone else to check the rules and the numbers before managers use it.

4Check the data

This page gets more room for the data check than the others. It is the first tool here that keeps data, and the data is each person's pay, rating and the reasons you write about them.

What this touches

Each team member's pay, rating and the reasons you write. This is the most sensitive data on this site, and it is the first tool here that keeps it.

What is safe

The tool itself stores nothing and sends nothing. The risk is the file it makes: where you save it, who can open that folder, whether it gets emailed, and whether old copies linger.

Use IDs rather than names. Keep reasons about the role and the work, not health or personal circumstances. When you build your own, give the AI assistant invented data only.

Ask before you start

  • Where are managers allowed to save a file like this?
  • How do they send the submission to the reward team safely?
  • When must working files be deleted?
  • Who else can open that folder?

Privacy rules for employee records differ between private employers and government agencies, so check with your privacy officer. Start with the OAIC's guidance on the Australian Privacy Principles and on employment and privacy. General information, not legal advice.

Three rules for the file

  1. Save it where your organisation keeps confidential documents, not on the desktop or in downloads.
  2. Send the submission the way your reward team asks, not as an ordinary email attachment unless that is their rule.
  3. Delete working copies when the review closes.

5Take it

Download the tool Open it in a new tab Download the sample team file

The tool is a single file. Save it anywhere and double-click to open it. Use the sample team file to practise opening a file.

What next: Use case 08: the reward team's review tracker · Back to all use cases