For 25 years my Monday started the same way: find out what actually happened since Friday, then write it up for someone who would read the first line and maybe the second. At Level 3, Microsoft and Meta that was the job. Not the writing. The finding out.
That's what people miss when they type "write a client update" into AI and get back a page of nothing. The writing was never the hour. Pulling the thread back together was the hour: the last three emails, what you said you'd deliver, what slipped, what you need from the client to keep moving. AI can't gather what it was never given.
Here is what mine looks like now.
One chat per job, one file it reads first
I don't run my business in one long chat. The blog and newsletter have a chat. LinkedIn has a chat. Setting up my business email had its own. Each one keeps a short status file in the project, and the first line of my brief to any of them is "read your status file first." Post #1 covered that file. What matters here is what it makes possible: the update doesn't start from a blank page, because nothing in my business starts from a blank page anymore.
The last thing I ask every chat to do before I close it is update its own file. By Monday morning, the update is most of the way written before anyone asks for it.
What this Monday looked like
Post #1 went live at 6:00 AM. I'd asked Claude to check it over and tell me where things stood. This is the update it sent, trimmed:
Post #1 is live, but I accidentally paused Tuesday's newsletter while checking on it. It needs your OK to reschedule before 7:00 AM Tuesday, or it won't go out. Reply "go" and I'll click confirm in Mailchimp.
The pause happened at 6:37 this morning. Opening the campaign's page dropped me into Mailchimp's editor, and that paused it. Nothing in the email changed: same list, sender, subject and content.
The rest of the check came out fine. The post is up with the scorecard card and signup box under it. Blog is in the menu. The list has grown to 11 people, up from 2 last Thursday.
Read that the way a program manager reads it. Bad news in the first line. What broke and why. What didn't change. What I need to decide, and by when. Then the good news. Three paragraphs, and I knew everything I needed to know in under a minute.
I didn't write a word of it. I also didn't act on it until Wednesday, and Issue 1 missed Tuesday. The update was right. The reader was slow. No format fixes that, and I'm saying so because people buy AI expecting it to make the decision, and it doesn't. It puts the decision in front of you with everything you need. The 20 minutes is still yours.
Where the 20 minutes goes
If you send clients a weekly update, the setup is the same. A project per client. Inside it, one file with the engagement in plain language: what you were hired to do, what done looks like, what's been delivered, what's open, what they're waiting on from you and you from them. Update it at the end of each working session, not on Monday morning.
Then the ask is short. Read the status file and the last email thread. Draft this week's update. Lead with anything that slipped or needs a decision from them, then what shipped, then what's next. Flag anything you're not sure is true.
That last line carries more weight than the rest. Mine reads my own notes, and my notes drift. A client update built on a stale file is worse than one you wrote by hand, because it reads confident.
So the 20 minutes: read the draft against what you know. Fix what's wrong. Cut what the client doesn't need. Add the one sentence only you can write, the call on what the slip means for them. Send it from your own address.
The hour I spent gathering is gone. The 20 minutes deciding never will be. That was the part they were paying for anyway.
