AI Distilled
Playbooks · Turn data into decisions
Turn data into decisionsHuman approves

A Monday summary of last week’s sales, drafted for you

Every Monday, a short summary of last week’s sales is ready to check, with each number traced to the cell it came from.

Setup
45 min
Rung
03 → 04
Tool cost
Free
Approval
A person

Before and after

Before

  1. You export last week’s sales
  2. You clean it up in a spreadsheet
  3. You work out what changed
  4. You write it up, if there’s time

After

  1. YouYou paste the export into the sheet
  2. ToolFormulas work out the numbers
  3. AIAI drafts the summary
  4. YouYou trace each number to its cell
  5. YouYou share it with the team

Five steps, 45 minutes

  1. 01Set up the sheet

    15 min

    Make two tabs. “Raw” is where each week’s export gets pasted as it is. “Numbers” uses formulas to work out the totals and the change from the week before, one metric per row, with its name in column A. Let the spreadsheet do the math; AI is for the words.

  2. 02Choose the numbers that matter

    5 min

    List the numbers you’d act on: sales, orders, average order value, best sellers, returns. Each week, the summary leads with the three that moved most.

  3. 03Write the summary prompt

    10 min

    Paste only the Numbers tab, never the raw export. It has everything the summary needs and none of your customers’ details.

    You write a Monday sales summary for [business], for [who reads it].
Use only the numbers in the table below. Don’t calculate anything new.
After every number, name the row it came from in parentheses, like (Orders).
Start with the three numbers that changed most compared with the week before, one plain-English line each.
Then add up to two possible reasons, each labeled as a guess.
If a number looks wrong or is missing, write [CHECK] instead of guessing.

Numbers tab:
[paste the Numbers tab]
    Open ChatGPTOpen ClaudeOpen Gemini

  4. 04Check every number

    10 min

    Read the draft with the sheet open. Find each number in its row, and treat the guesses as questions to ask, not findings to share.

    Human check: Every number matches its row in the Numbers tab, nothing appears that isn’t in the sheet, and guesses are labeled as guesses. Clear every [CHECK] before you share it.

  5. 05Let it draft itself

    Optional · rung 045 min

    Once a few weeks check out, make it routine. If your assistant can read files from your drive and run a task on a schedule, point it at the Numbers tab with the same prompt so the draft is waiting on Monday. If not, a Monday reminder and one paste do the job. The check stays manual.

What to measure beyond “time saved”

  • Numbers corrected at review

    Aim for none after the first few weeks. If they keep coming, fix the formulas or the paste, not the wording.

  • Mondays it went out

    The point is a habit. Count the Mondays in a row.

  • Decisions it prompted

    A reorder, a promotion pulled, a call made. If nothing ever changes, you’re watching the wrong numbers.

  • “Compared to what?” questions

    If readers keep asking, add the comparison to the sheet.

Common pitfalls

  • Letting AI do the math. Language models can add up wrong, confidently. Formulas calculate; AI only describes what the sheet worked out.

  • Guesses dressed up as facts. “Sales fell because of the weather” is a guess. Keep guesses labeled, or leave them out.

  • Pasting the raw export. Customer names and emails don’t belong in the prompt. Paste the Numbers tab only.

Need help building it?

AI Distilled can build this with you: a working pilot, tested on your real work.

Scope a build