Toolkit · Article 06 · AI at Work

Clear Update Kit

A short weekly update people can trust. It is 300 words, with a source beside every number and one clear ask.

3Tools
4Steps
30Minutes
FreeTo use
300 42 SOURCE SEND
01

At a glance

What it is

An update people can trust

AI can write a long update in seconds. It often looks good and says little. Someone else then has to fix it.

This kit keeps the weekly update short and true. It says what changed, what needs care and what you need.

Each number has its source next to it. The update ends with one clear ask.

Who it is for

The people who write updates

Team leads, managers, HRBPs, and anyone who writes weekly updates.

When to use it

Three moments

  • Your update gets a bit longer each week.
  • Leaders reply, "So what do you need from us?"
  • A number in a report turned out to be wrong.
Kyun, the Mimansa loris
Kyun

Your update goes out on Friday. What should the reader know and do?

02

The four steps

About 30 minutes each week. Do all four steps before you send.

  1. 1Write the three parts15 min
  2. 2Add the sources5 min
  3. 3Write the recommendation5 min
  4. 4Do the send-check5 min
1
Step 1 of 4

Write the three parts

15 min

The person who sends the update does this.

  1. Write what changed this week.
  2. Write what needs care, and why.
  3. Write what decision you need, and from whom.
Kyun

What changed, what needs care, what do you need?

CHANGED ATTENTION DECISION 300
2
Step 2 of 4

Add the sources

5 min

The writer does this. A teammate can check it.

  1. Put the source in brackets beside each number.
  2. Open each source and check the number.
  3. Remove any number you cannot trace.
Kyun

Where did this number come from?

42 SOURCE
3
Step 3 of 4

Write the recommendation

5 min

The writer does this.

  1. Say what you recommend in one line.
  2. Say who should decide.
  3. Say by when.
Kyun

Say what you want. Then say by when.

DECIDE
4
Step 4 of 4

Do the send-check

5 min

The writer does this just before sending.

  1. Read the update aloud once.
  2. Ask yourself: can I explain every number with no notes?
  3. Cut any line you cannot explain.
Kyun

Could you explain it on a call, with no notes?

CHECKED
03

Example and how-to

Filled-in example · 300-Word Update Template

Weekly Update to US Leaders

Payroll team in a Bengaluru GCC, 14 people. The numbers are made up. Copy the shape, not the words.

Vague: what changed
"We saw strong progress across key payroll workstreams."
Clear: what changed
Error tickets fell from 42 to 29 this week (ticket tool, 18 Oct).
Vague: needs care
"Some risks may impact timelines going forward."
Clear: needs care
Two of five payroll staff are on leave next week (leave tracker).
Vague: decision
"We look forward to alignment on next steps."
Clear: recommendation
Approve one temp for two weeks. The US payroll head decides by Wednesday.
Length
Clear version: 180 words.
How-to guide · 5 minutes

How to rewrite a vague AI report

  1. Read the draft and underline every number.
  2. Find the source for each one. Delete what you cannot find.
  3. Cut empty words like key, robust and going forward.
  4. Write one line each: what changed, what needs care, what you need.
  5. Add one ask: what, who and by when.
  6. Read it aloud. Cut it to 300 words or less.
04

What you get

3 free tools. Each one is a single page or a single sheet.

Kyun

That's the kit. Each one fits on one page.

05

Why it matters, and what's next

40%

of desk workers got AI workslop in the past month. Each case took nearly two hours to sort out.

Source: BetterUp Labs and Stanford Social Media Lab, workslop research, Sept 2025 (1,004 US full-time desk workers)
66%

of people asked about AI at work say they use its output without checking if it is right. 56% say AI has caused mistakes in their work.

Source: KPMG and University of Melbourne, Trust, Attitudes and Use of AI: A Global Study, 2025 (48,340 people, 47 countries incl. India)
42%

of workers who got workslop now see the sender as less trustworthy. Nearly 1 in 3 would rather not work with them again.

Source: BetterUp Labs and Stanford Social Media Lab, workslop research, Sept 2025 (1,004 US full-time desk workers)
Kyun

Next, learn to spot workslop fast.