How to Write a Project Status Report With Traffic Light Indicators Using AI

Project Status Report Traffic Light Indicators AI, Explained

A traffic light status report — usually called RAG, for Red, Amber, Green — gives every project or workstream one colour that summarises its health, followed by the specific reason for that colour and the decision it needs from leadership. The colour is the headline. Everything else is supporting evidence.

It exists because senior readers do not read status reports. They scan them, in roughly a minute, looking for one thing: where do I need to unblock something. A report that buries a nine-day slip inside a well-written paragraph has failed at its actual job, however good the writing is.

The second version is shorter and it took less time to produce. That is the whole argument for this method.

That gap between how reports get written and how they get consumed is exactly what a project status report traffic light indicators AI workflow closes. The AI is not there to write more elegantly than you. It is there to sort, colour, and rank faster than you can, so the shape of the document matches the way it will actually be read.

Why the coloured dots matter

The visual part is not decoration. Coloured symbols are what let the eye jump straight to trouble without reading, which is why the convention is standard well outside AI tooling — Atlassian’s own support guidance for building RAG health fields in Jira recommends the same Unicode 🔴 🟡 🟢 characters this workflow asks AI to output, precisely because they survive being copied between tools.

The Threshold-First Status Method

The Threshold-First Status Method is the approach we teach at PromptPeakAI for this task: you write the numbers that decide each colour before you write anything else, so the AI is doing a calculation rather than forming an opinion. It has four moves.

Three of these four moves take seconds once you have written them down. The first one you do properly a single time, then reuse every week for the life of the project.

Set Your Red, Amber, and Green Thresholds

Amber and Red mean nothing until you attach numbers to them. Pick a slip in days, a budget variance in percent, and an age for open blockers — then write them down. These are defaults that work for most corporate delivery work; adjust them to match what your leadership actually escalates on.

The point of writing them down is not precision. It’s that the number is now external to the person reporting, so nobody has to decide whether a nine-day slip feels bad enough to admit to.

Save this block somewhere you can reach in two seconds — a pinned note, a saved chat, a project instruction. You will paste it above your notes every single week.

STATUS RULES — apply these exactly, do not soften them.

🔴 RED: milestone slipped more than 5 days, OR spend more than 10% over plan, OR a blocker open more than 5 working days, OR a decision is needed from leadership.
🟡 AMBER: milestone slipped 1-5 days, OR spend 5-10% over plan, OR a blocker open 3-5 working days.
🟢 GREEN: none of the above is true.

If the notes do not contain enough information to apply a rule, output "Not stated" for that item. Do not guess, and do not average two signals into a milder colour — the worst applicable signal sets the colour.

Build the Report in Four Steps

Four steps, start to finish, using any mainstream chat AI. The first run takes about fifteen minutes because you are writing the rules. Every week after that is closer to five.

Collect the raw material and resist the urge to clean it up. Half-sentences, contradictions, and the message where someone hedged are the most valuable input you have — the AI can spot the hedge, but only if you leave it in. Copy from Slack, Teams, email replies, ticket comments, and your own scribbled notes into one plain text block.

If your source is a recording rather than a chat thread, run it through a meeting transcript to minutes workflow first, then bring the action items here.

Order matters more than people expect. Rules first, then the report structure, then the raw notes at the bottom. The model reads it as instructions followed by data, rather than data it needs to figure out what to do with.

You are a PMO reporting analyst preparing a weekly status report for an executive audience.

[PASTE YOUR STATUS RULES BLOCK HERE]

Produce the report in exactly this order:

1. HEADLINE — one line naming the single worst item and what it needs. No preamble.
2. DECISIONS NEEDED — bullet list. Each bullet: the choice, the owner, the deadline. If none, write "None this week."
3. STATUS TABLE — columns: Workstream | Status | This week | Risk or blocker | Owner. Put the 🔴 rows first, then 🟡, then 🟢.
4. MOVEMENT — one line per workstream that changed colour since last week, and why.

Style rules: no adjectives of praise, no summarising sentences at the end, plain language a non-specialist can follow. Keep the whole report under 300 words.

RAW NOTES:
[PASTE YOUR NOTES HERE]

This is the step people skip and regret. Left alone, AI will smooth over a gap by inventing a plausible date or a risk nobody raised — the same failure mode behind every AI hallucination in professional documents. The fence is one paragraph and it costs you nothing.

Grounding rules — apply before you output anything:

- Use ONLY facts stated explicitly in the raw notes. Do not infer, estimate, or fill gaps.
- Any date, number, name, or percentage that does not appear in the notes must be written as "Not stated".
- After the report, list every item you marked "Not stated" under the heading GAPS TO CHASE.
- Do not soften language that describes a problem. If a note says "blocked", write blocked.

Then verify by hand. You do not need to check everything — check the three items most likely to be quoted back at you: every date, every number, and every Red. That is usually under a minute of work and it is the difference between a report you can stand behind and one you cannot.

The finished document leads with the worst news and the decision it needs, then shows the table, then everything else. This is the opposite of how most people write, which is why most status reports get skimmed and abandoned. If you also owe a longer written brief, the same ordering logic applies to an executive summary written with ChatGPT.

Read the headline once before you send. If it does not name a thing and an ask, the report is still describing rather than reporting, and one more prompt will fix it.

One variant worth saving

If your raw notes come from engineers and your reader is not one, add a translation pass. Same rules, same structure, different vocabulary — and it heads off the round trip where a director asks what a failing integration test means for the launch date.

Rewrite the report above for an executive with no technical background.

- Replace every technical term with its business consequence. "Refactoring the payment service" becomes "rebuilding how payments work behind the scenes".
- Every risk must state what it costs: days, money, or a customer commitment.
- Keep the 🔴 🟡 🟢 symbols, the table structure, and the decisions list exactly as they are.
- Do not remove bad news to make it read better. Translate it, do not soften it.

Make the Table Paste Into Word Cleanly

Here is where most people give up. The report looks perfect in the chat window, you copy it into Word or Outlook, and the table arrives as a row of pipe characters and dashes. Nothing is broken — you copied the raw markdown instead of the rendered version.

Three fixes, in order of how often they work:

• Select the rendered table, not the code block. Drag-select the visible table on screen and copy that. Word and Outlook both accept it as a real table.
• Ask for HTML instead. Say “output the table as an HTML table” — pasting HTML into Word preserves cells and borders far more reliably than markdown does.
• Keep the coloured circles as Unicode characters. 🔴 🟡 🟢 are text, not images, so they travel through email, Word, Slack, and Google Docs without turning into broken boxes.

Reformat the status table above as a clean HTML table.

- Use standard HTML table markup (table, tr, th, td tags) with visible cell borders.
- Keep the 🔴 🟡 🟢 characters exactly as they are — do not replace them with the words Red, Amber, Green, and do not use images or coloured cell shading.
- Do not change any of the content. Formatting only.

Output only the table code, with no explanation before or after it.

Which AI Tool Fits Your Notes

Any mainstream assistant can run this workflow — the method is the durable part, not the tool. What genuinely differs is how much friction sits between where your notes already live and where the report has to end up. The broader trade-offs are covered in our Microsoft Copilot versus ChatGPT comparison for office work; here is the short version for this one task.

ChatGPT and Gemini

Both handle pasted notes well and both are free to start. Gemini earns its place if your updates arrive as Google Docs comments and Gmail threads, because the material is already in reach. ChatGPT is the fastest path if you are copying from Slack or Teams anyway.

On either one, save your threshold block and the report structure as a reusable instruction rather than retyping it. A saved project or custom instruction means week two starts with a paste of raw notes and nothing else — which is where the five-minute figure actually comes from.

Claude

Claude renders formatted output into a side panel you can read and iterate on before copying, which removes most of the guesswork about whether a table will survive the paste. If you have not used that feature, our guide to Claude Artifacts covers it — and it is available on the free plan, not just paid.

Microsoft Copilot

Copilot’s advantage is that the report is born in the right file. Microsoft’s official guidance for Copilot in Word explicitly lists project updates with status, milestones, risks and decisions as a drafting use case, and you can reference existing files as source material. Our walkthrough of using Copilot inside Word covers the mechanics. The catch is cost: this needs a paid Microsoft 365 Copilot licence, currently $30 per user per month as an add-on. The free Copilot Chat tier can still run the workflow, just not from inside your document.

What Never to Paste Into AI

Status reports are unusually sensitive documents. They contain launch dates, vendor costs, headcount problems, and the name of the client who is unhappy. Before you paste, sort your notes into three buckets.

Redacting is faster than it sounds: swap real names for “Client A” and “Vendor B”, replace exact figures with percentages, and write the report against those. Swap the real names back in Word afterwards, where the document was always going to live anyway. On business and enterprise plans the training question is usually settled by contract rather than by a settings toggle — worth confirming with your IT team rather than assuming, and our guide to whether ChatGPT is safe to use at work goes through what to check.

What This Workflow Cannot Do

Every honest version of this method has the same boundary: AI handles the assembly, and you keep the judgement. Three things do not transfer, no matter how good your prompt is.

There is also a quieter limit worth naming. This workflow makes a bad project visible faster; it does not make it better. If your reports start turning Red, that is the system working, and the uncomfortable conversation that follows is the actual point.

The whole setup cost of a project status report traffic light indicators AI workflow is one afternoon of deciding what your numbers are. Everything after that is paste, check, send — and the checking is the part that stays yours.

When the Report Comes Out Wrong

Four failures account for nearly everything that goes wrong here, and each has a one-line fix.

If two colours conflict — schedule says Amber, budget says Red — the worst signal wins. Build that line into your rules block once and you will never argue with yourself about it again.

Run this for three weeks and something quietly shifts: your team leads start writing their updates in a way that anticipates the thresholds, because they know a vague note produces an Amber. The report gets better because the reporting got honest, which was always the harder half of the problem.