← All use cases
Use case 08

Reward team's review tracker

Run the whole pay review from one place. Load each manager's submission as it arrives and see progress, spend against budget, every exception, and any skew between women and men across the organisation.

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

The reward team's view of a review usually lives in a pile of spreadsheets, stitched together late at night. Exceptions get lost in the pile. Patterns across teams are only seen afterwards, when there is little left to do about them.

This tool is the reward team's half of a pair. Managers use use case 07 to propose increases for their own team and download a submission file. You load each file here as it arrives, and the whole picture builds up while there is still time to act.

The fairness view matters most. A pattern that no single team can see can show up once everyone is put together. Better to find it now than in next year's pay gap figures.

Both tools are ordinary code, built by describing what you want to an AI assistant. Neither uses AI when you run it. This one is "stateful": it keeps your work between sittings, so it needs somewhere to keep it. Here, that is a file you save.

1Try it

This is the real tool, running in your browser on an invented review of five teams. Three have already submitted. Load a submission, record decisions, change the minimum group size, or save your work.

Your files are read by your own browser. Nothing is uploaded.

Try this

  • Look at the four views: Progress, Budget, Exceptions and Fairness.
  • Record a decision on an exception, and add a note to it.
  • Change the minimum group size on the Fairness view and watch which levels appear and disappear.
  • Save the workspace, close the page, then reopen the tool and use "Open a workspace" to carry on where you left off.
  • Finish Finance Operations in use case 07, download its submission file and load it here.

2Watch it

Five steps from a folder of submissions to a pattern across the organisation. Press Play, or step through at your own pace.

Step 1 of 5Invented data
Files arriving from managers
TeamFileStatus
People Partnerspeople-partners.jsonNot startedSubmitted
Technology Platformtech-platform.jsonNot startedSubmitted
Risk and Assurancerisk-assurance.jsonNot startedSubmitted
Finance Operationsno file yetNot started

Submissions arrive. Each manager sends one file. Load them one at a time or all together, and each team flips to Submitted.

Progress by team
TeamStatusPeopleOutside guideline
People PartnersSubmitted72
Technology PlatformSubmitted114
Risk and AssuranceSubmitted82
Finance OperationsNot started9 expected–
Operations ServiceIn progress12 expected–

Where everyone is up to. One table shows who has submitted and who has not. Teams still working show dashes instead of money figures, so nothing is guessed.

Spend as a share of each team's own budget
People Partners
87% (under)
Technology Platform
112% (over)
Risk and Assurance
91% (under)
Within budget Over budget Budget line

Spend against budget. Two teams are under. One is over by about $5,500. The organisation as a whole can look fine while one team is not.

Proposals outside the guideline, furthest first
Cloud Architect, Technology Platform6.0%, 2.0 points above
“The only person who can run the cloud migration design.”
ApproveDeclineQuery
Fraud Analyst, Risk and Assurance6.5%, 1.5 points above
“Two recent searches for this role failed to fill it.”
ApproveDeclineQuery
Site Reliability Engineer, Technology Platform5.5%, 1.5 points above
“Rebuilt the on-call roster and cut out-of-hours callouts.”
ApproveDeclineQuery

Every exception in one list. The manager's reason sits next to the proposal. You record Approve, Decline or Query, and a note.

Average proposed increase, women and men, all teams together
Level 2
Hidden: small group
2 women, 1 man
Level 3
Women 2.9%Men 3.6%
0.7 points lower
Level 4
Women 3.8%Men 3.6%
0.2 points higher
Women Men

The pattern no single team can see. No one team has enough women and men at a level to show this. Put together, level 3 stands out. It is a prompt to look closer, not proof.

All figures in this walkthrough are invented, for illustration only. They match the sample data in the tool above.

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.

This one is harder than a single tool, because two tools have to work together. The steps below put the hard part first: agreeing what travels between them. Build with invented data, never your real review files. Expect two or three rounds of "that's not quite right, change X".

Step 1: Agree the file before you build anything

When two tools pass a file between them, they need a written agreement on exactly what is in it: every field, what each one means, and how every figure is worked out. If they disagree, even slightly, the totals will not match and nobody will know which tool is right. Writing the agreement down first is the single most useful habit for anything bigger than a one-off tool.

I work in reward in Australia and I am not a developer. I want to build two small tools that pass a file between them. Tool A is used by managers. Each manager enters a proposed pay increase for everyone in their own team, then downloads a "submission" file. Tool B is used by the reward team. It loads the submission files and shows progress, spend against budget, exceptions and a fairness check. Before any code, help me write a short agreement document that both tools will follow. It must cover: 1. Every kind of file: a starting file for a manager, a saved work file, a submission file, and a saved workspace file for the reward team. For each one, list every field, what type it is, and give an example. 2. A "format" field and a "version" number on every file, so each tool can recognise the wrong kind of file and say so in plain English. 3. The rules both tools must work out the same way: position in range (salary divided by a reference point), which guideline zone applies, the guideline range for each person, the new salary, when a proposal counts as outside the guideline, budget, spend, and the average increase for women and men with a minimum group size. 4. No names in any file that leaves the manager's computer. Ids only. 5. Nothing stored in the browser and nothing sent anywhere. Use invented example values. Keep it to two pages. Ask me about anything that is unclear, then wait for me to approve the document before you build either tool. Once I approve it, build to it exactly. Do not add, rename or drop fields.

Step 2: Describe the tool

Say who uses the tool and what each view shows. Name the views, so you can ask for changes to them later.

Now build Tool B, the reward team's review tracker, as a single HTML file I can save to my computer and open in a browser by double-clicking it. Follow the agreement exactly. Use no external libraries and send no data anywhere. Use plain English and Australian spelling. It opens with an invented sample review of five teams, three of them already submitted, with a clear "Invented data" label. It has a "Load a submission" button that accepts one or several files at once, by file picker or drag and drop. It has four views, as tabs: 1. Progress: tiles for teams submitted out of the total, people covered, proposals outside the guideline, and decisions still to make. Below that, a table of teams with status, headcount, budget, spend, whether the team is over or under budget (in words and colour), and the number of proposals outside the guideline. Teams not yet submitted show dashes for money. 2. Budget: the organisation budget against total spend, then a bar for each submitted team against its own budget. A clear note that teams not yet submitted are not in the figures. 3. Exceptions: every proposal outside the guideline, furthest outside first, with the manager's written reason. For each one, buttons for Approve, Decline and Query, and a note field. Filters by team and by decision. 4. Fairness: the average proposed increase for women and men across all teams, then by level and by rating. Hide any group with fewer than a minimum number of women or men (default 5, editable) and show "Hidden: small group" instead of the figures. Finish with a summary I can print and a button to download every proposal as a CSV file. 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 3: Do not trust the file

A file can be out of date, edited by hand, or made by a different version of the other tool. A total you have not checked is a total you should not report, so the reward tool works everything out again for itself.

Change how a submission file is loaded. - Check the "format" and "version" fields first. If the file is the wrong kind, say which kind it is and which kind the tool needs, in plain English. For example: "This is a manager's working file, not a submission. Ask the manager to finish and download the submission file." Never crash on a file you do not recognise. - Do not use the new salary, the guideline range or the outside-the-guideline flag from the file. Recalculate all three yourself from the rules in the agreement, and use your own results everywhere. - If your result is different from the file's on any row, show a clear warning on that row that says "does not match the file", and say what differs. - Match the file to a team by team id. If that team already has a submission, ask before replacing it. If the team id is not in my list, offer to add it as a new team. - Treat all text from a file as untrusted. Escape it before putting it on the page. - After loading, say in one plain sentence what happened, for example: "Loaded Finance Operations: 9 people, 2 outside the guideline." When you finish, list every figure that is taken straight from the file and every figure that you calculate.

Step 4: Decide where the work is saved

A tool that "remembers" has to keep your work somewhere. A browser can keep it for you, but only in that one browser on that one computer. It can be cleared without warning, you cannot see where it is, and you cannot put it in a controlled folder. A file you save and open is different. You choose where it lives, who can reach it, and how long you keep it. This site chose a file, because pay data belongs in a folder you control, not in a browser you do not.

Add a way to save and reopen the reward team's work, using a file and not the browser. - A "Save workspace" button downloads one JSON file named review-2026-workspace.json. It holds everything needed to carry on: the list of teams, each loaded submission, every decision and note, and the settings. - An "Open a workspace" button loads that file again and restores the screen exactly as it was. - Show an "Unsaved changes" marker whenever anything has changed since the last save, and "Saved" with the time after saving. Warn me before I close or leave the page with unsaved changes. - Do not use localStorage, sessionStorage, IndexedDB or cookies, and make no network requests. - Show this line on the screen: "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 that the only one is the file download.

Step 5: Test the hand-off

Test the two tools together, not one at a time. Use a small invented team, and work out the answers by hand before you load anything.

Please help me test the hand-off between the two tools. 1. Write a test submission for one invented team of 6 people with round-number salaries. Make 2 of the proposals fall outside the guideline. 2. Before I load it, tell me what to expect, worked out step by step: the budget, the spend, the number of people outside the guideline, and each person's guideline range. 3. Give me four more test files: the same team again, a manager's working file, a team file, and a spreadsheet saved as CSV. 4. For each file, tell me what the tool should say or do. 5. If anything does not match, say so plainly and fix the tool. Do not change the test data to make it fit.

Your test checklist

  • A submission from the manager's tool loads, and the people, budget, spend and exception counts match what the manager's tool showed.
  • Loading the same team twice asks before it replaces anything.
  • The wrong kind of file is refused politely, in plain English, and nothing breaks.
  • A level with a tiny group is hidden, with no figures shown.

What you would add next

  • A shared database, set up by IT, so the whole team works on one copy instead of passing a file around.
  • A record of who decided what, and when.
  • A way to lock the review when it closes, so nothing changes afterwards.

Good habits

  • Write the file agreement down before you build, and keep it with the tools.
  • Check the totals by hand on a small invented team before you trust the tool with a real review.
  • Treat every saved workspace as a committee paper: name it clearly and keep it in a controlled folder.
  • Keep the prompts you used so you can rebuild or audit the tools later.

4Check the data

What this touches

Pay, ratings and managers' written reasons for everyone in the review. This is the largest collection of pay data on this site.

Written reasons can say more than a salary does. Treat them with the same care.

What is safe

The tool stores nothing and sends nothing. The risk is the files: the submissions coming in, and the workspace you save.

One saved workspace holds the whole review. Treat it like the remuneration committee paper.

Use ids, not names, in every file.

Ask before you start

  • Where does the reward team keep review working files?
  • How should managers send submissions?
  • Who outside the team can open that folder?
  • When are working files deleted after the review closes?

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.

When a file is no longer enough

If several people need to work on the review at the same time, or the review has to be audited, a file passed around stops being safe or practical. That is the point to talk to IT about a proper shared build. See the foundry use case and the ways to build for what that looks like.

5Take it

Download the tool Open it in a new tab Download a sample submission

The tool is a single file. Save it anywhere and double-click to open it. Load the sample submission to see what a manager's file looks like and how the tool reads it.

What next: Use case 07, the manager's tracker · Back to all use cases