Hey there!

Happy first week of September. Can you even believe it? Summer always goes by in a flash. I really feel like I soaked up every minute this year though. My kids started school this week and my jaw hit the floor when I took their back to school pictures and I saw just how tall they’ve gotten. It’s like they both grew up this summer? I have a tween and a teen (weren’t they just toddlers?).

Anyway, time is flying, and in business, that means we are coming up on the best quarter of the year IMO. The Q4 Black Friday season is around the corner, baby! More to come on that soon. For now, let’s get into this week’s news.

It’s Sunday night, and you have 9 tabs open. Instagram insights in one. Your email platform in another. The sales dashboard, the spreadsheet, and finally the doc where the report actually lives. You’re copying the same 5 numbers into the same 5 boxes you filled in last week. Somewhere around tab 6, the thought lands: why am I still the one doing this?

Real talk: you don’t have to be. My team used to lose a chunk of every Monday assembling our weekly report. Same sections, same sources, same slog. Now AI assembles it and we review it, and I’m a little annoyed we didn’t set it up sooner.

This week’s blog post is the full walkthrough. The short version starts with a 3-part test. Your report can build itself if it’s:

That’s the entire qualification. Nothing in there about being technical, because automating a report isn’t a tech skill. It’s a definition skill. And a recurring report is the rare task that passes on day 1, because you’ve been writing its instruction manual by hand every single week. You just never handed the manual to anyone.

So go do this today. Here’s how:

And that review step isn’t a downgrade from “fully automated.” It’s the design. AI is your COO here, not your CEO. It runs a system you defined, and you stay the one who decides what the numbers mean and what happens next.

The full prompt, the project setup, the scheduling walkthrough and the “what if my report is a mess” answer are all in the post. Next Sunday night, those 9 tabs stay closed.

As of last week, Claude’s memory works everywhere: regular chats, Cowork sessions, all of it, and it updates while you talk. It’s on by default for everyone, free plan included.

Which means the era of opening Cowork and having to re-explain yourself is officially over. Yay.

You can see every single thing it remembers, and change it.

Where it's saved depends on which version you have, since it’s rolling out gradually. If you see Settings > Memory, you’re on the new experience: everything it remembers is listed under Topics. Open one, hit the edit icon to change it, or delete it outright. If you see Settings > Capabilities instead (common on Team accounts), click “View and edit memory,” then use the pencil icon to update the summary.

Either way, there’s a lazier option I fully endorse: just tell Claude in the chat what to remember or forget, and it applies from your next conversation.

Do this once, and every draft it writes next week starts from your business instead of a blank slate.

LinkedIn quietly gave everyone a “seems like AI slop” button at the end of July. By mid-August? Pressed over 1 million times.

And it’s not just readers being petty. Posts that get flagged are seeing about 40% fewer views, straight from LinkedIn’s own product chief. The AI detection firm they work with found that 41% of long-form LinkedIn posts are now fully AI-written. Almost half.

At the same time, LinkedIn removed its own rewrite-your-post feature. The AI writing tool they built and promoted to a billion members? Gone. The replacement only proofreads. It fixes your grammar and leaves your voice completely alone.

So the platform that sold you AI writing just decided AI’s job is to assist your voice, not replace it. Which, hi. What have we been saying all year?

Snapchat made the same call this month, by the way. Fully AI-generated videos no longer get recommended in Spotlight. This isn’t one company’s policy. It’s the direction every platform is heading.

To be clear, this is not your sign to quit using AI. It’s your sign to stop using it to create your content from start to finish. Voice dump your own ideas as your core starting point, and then use AI to help you execute, research, repurpose. Always make the final pass sound like you, because 1 million flags just proved your readers can tell.

Two Instagram updates, and together they say the same thing LinkedIn just did. Earlier this year Instagram stopped recommending photos and carousels from accounts that mostly repost other people’s work. It checks monthly, and if most of your posts in a 30-day window are reposts, your reach drops. Its own test for third-party content: do you “add real value beyond just simply restating” it?

Then on Monday it went a step further. The optional “AI creator” label is now called “AI-generated profile,” and if your account features an AI-generated person and you haven’t labeled it, Instagram stops recommending you. Add the label yourself and nothing changes.

The line I actually care about, straight from Instagram: creators who “simply use AI tools as part of their creative process” don’t need the label. Editing a photo, polishing a caption, building a graphic with AI: none of that is what’s being flagged. An account where the “person” is AI is.

Which is exactly the distinction I was asking for last issue, when the watermark couldn’t tell AI-assisted from AI-generated. This week 2 platforms drew that line themselves. AI can help you make the thing. It can’t be the thing.

That’s a wrap. If you do one thing this week, take one report you create and run the prompt in section 1. 20 minutes today, and Monday’s report is no longer your job? Win.

Hit reply and tell me what you automated. I read every one.

See you next Thursday Gemma. Gemma Bonham-Carter gemmabonhamcarter.com