A SMALL FIRST WIN
Before you start
Open ChatGPT and switch to Work if it is available in your account. Connect only the source app you need, or attach an approved export. Features depend on your plan and workspace.
- A small CSV with matching reporting periods
- A definition for each metric
- A report outline and its intended reader
Use sample or approved information for your first run. Keep the review step visible before you expand the workflow.
LET’S DO THE WORK
Follow these four steps
- 01
Choose one report and its inputs
Start a Work conversation. Attach the practice CSV or a permitted export. State the dates, metric definitions, and audience. For an app connection, identify the exact report or folder to read.
You’re ready when: You can identify both reporting periods and every required column.
Step 1 · illustrated workflow - 02
Hand over a complete brief
Copy the prompt below. Replace the example metric names only if your data uses different definitions. Ask for the finished report and its calculation table together.
You’re ready when: The requested output has a format, a source, and a clear stopping point.
Step 2 · illustrated workflow - 03
Check the result against the source
Open the output. Recalculate one row yourself, then check dates and totals. In the practice data the rates are 25% and 30%; the difference is five percentage points. Ask for corrections before reusing the method.
You’re ready when: The report matches the data and does not turn a correlation into a cause.
Step 3 · illustrated workflow - 04
Make the good version repeatable
Save your corrected brief. If scheduling is available, use the follow-on prompt below and verify the saved schedule and first run. Make sure each run can reach fresh source data; an old attachment is not a live feed.
You’re ready when: A saved task exists and its first run uses the intended reporting period.
Step 4 · illustrated workflow
COPY, ADAPT, AND TRY
Your first run starts here.
Fictional practice data Period,Inquiries,Booked calls Previous week,40,10 Current week,50,15Download the sample CSV ↓
Create my weekly inquiry-to-booking report from the attached CSV. Use only its data. Show inquiries, booked calls, and booked calls divided by inquiries for each period. Include the calculation, the percentage-point change, three observations, and two questions to investigate. Separate observed changes from possible explanations. Return a one-page report and a table I can inspect. If a column, period, or definition is missing, flag it. Do not email or publish the report.
Current week: 50 inquiries; 15 booked calls. Booking rate: 15 ÷ 50 = 30%. Previous week: 10 ÷ 40 = 25%. Change: +5 percentage points. Open question: did lead sources change? Cause: not established by this file.
Actual output will vary. Use this example to judge structure and accuracy, not as a promise of a successful run.
KEEP THE USEFUL PART
Make it repeatable.
After the first result meets your needs, use this follow-on prompt to carry the method forward.
Every Friday at 9:00 AM America/New_York, prepare this report from [exact connected source and report]. Use the newest complete week and compare it with the previous complete week. Return the report here for review. If the source is unavailable or stale, report the problem and stop. Do not email it. Confirm the saved schedule, source, and where I can review or pause the task.
If this feature is not available
If Work or scheduling is unavailable, paste the small data table into a normal ChatGPT conversation and run the report manually. That is assisted work; it does not refresh itself.
YOUR FINAL CHECK
Is the result ready to use?
- Same date range and definitions in both periods
- Calculations independently checked
- Every claim traceable to a source
- No sharing before you review
Sources & access
Provider documentation reviewed October 31, 2026. These are original instructional recipes, not account-tested builds. Plan access, labels, and features can change.
Prompts and examples are written by Automately. Workflow illustrations are schematic, not screenshots of the providers’ interfaces.