Hey everybody,
November has a way of filling the calendar before you notice. The days get shorter. The holiday lists start. The work that was supposed to slow down does not.
So here is a question worth carrying into the month: what if your AI got a little less eager to finish things for you, and a little better at waiting for your yes?
That is the thread running through this issue. Not a new model to learn. Not a race to keep up. Just a simple shift: let the tool do the first draft, the first check, and the first summary, and keep the final call with you.
The most useful AI this week is the one that pauses before it sends, pays, signs, or deletes anything on your behalf.
Adoption doubled. The training didn’t.
The tools arrived before the training did
The UK just measured what adoption actually looks like. The government’s AI Adoption Survey 2026, published October 29, found one in three UK businesses now uses AI, double the one in six a year earlier. Construction firms nearly tripled, from 9% to 25%. Real estate came in at 45%.
But the payoff is thin. Use is concentrated in simple text tools for drafting and summarizing. Half of adopters say productivity rose, but modestly. Nearly two-thirds say operating costs have not changed.
Yes, the survey is British. The pattern is not. So here is the question for your next leadership meeting: what is your company doing differently? How are you improving internally, in training, in workflows, in routine practice, because of these numbers?
A healthier way to judge what AI is worth
On October 28, Anthropic made three tools generally available for Claude Enterprise:
- smart reports
- in-app surveys
- pooled group budgets
A smart report groups a sample of a team’s Claude work into workstreams, shows what was produced and what it cost, and is off by default. Anthropic says the reports are meant for adoption and investment planning, not for evaluating individual employees.
Pooled budgets let a team share a monthly spending limit while each person keeps an individual limit, with alerts at 50, 75, 95, and 100 percent.
The jobsite camera becomes a second set of eyes
On October 21, EarthCam premiered the tenth generation of its Procore integration at Procore Groundbreak. The update brings jobsite visual intelligence deeper into Procore workflows, with new detections for trench safety and housekeeping hazards, plus delivery verification and dump truck analysis. Milestone verification can now be uploaded on target dates with the date, task name, and site conditions labeled on the image, and punch list items and inspections can be populated with EarthCam imagery directly.
Why this matters to you: the camera already on the jobsite becomes a second set of eyes. A superintendent still signs off. A project manager still decides. But nobody has to download, re-upload, and hunt through photos to prove what conditions looked like on a given day. That is AI doing the boring, useful part, and leaving judgment where it belongs.
Give one recurring job its own project
Microsoft’s October Copilot update adds Projects. A project keeps related chats, tasks, files, and instructions together, so you stop rebuilding context every time you open a new chat. Instructions set at the project level apply across its work, and files created by project tasks collect in one place. Projects arrived in the Cowork tab in October. The new Home screen, which brings Chat and Cowork together with projects across both, is rolling out to early-access (Frontier) users through mid-November, so timing will vary by tenant.
Why this matters to you: pick one recurring job, a board packet, a monthly report, a holiday plan, and make it a project. Put the instructions, the files, and the last version in one place. The next time you open it, you will not have to re-explain the whole job. That is the difference between chatting with AI and actually working with it.
Give your AI a pause rule
The problem is one you already know. You paste something into AI, it hands back a polished draft, and it looks ready to send. Then, later, you spot the guess. A date that was never confirmed. A name spelled the way the tool assumed. A promise you never meant to make.
This week, give your AI a pause rule before anything leaves your hands.
- Pick one thing you were about to send: an email, a proposal note, a family message, a plan.
- Paste it into the AI you already use: Copilot, ChatGPT, Claude, or Gemini.
- Use the prompt below, word for word, and fill in the brackets.
- Read what it flags. Fix the facts yourself. Do not let it fill a gap with a guess.
- Decide yourself whether it goes out. The tool drafts. You decide.
Here is a draft I am about to send: [PASTE YOUR DRAFT HERE]. This is about [TOPIC, FOR EXAMPLE A CLIENT UPDATE OR A HOLIDAY PLAN]. Before I send it, check it carefully. Separate facts from assumptions. List anything that is assumed, unconfirmed, or missing, and flag it instead of guessing. Do not invent missing information. Do not rewrite it to sound more certain than the facts allow. Do not send, change, or commit to anything on my behalf. At the end, give me a short list: what is solid, what needs checking, and what I should change before I send it.
Watch for the confident sentence. That is usually where the guess hides. A tool that sounds sure is not the same as a tool that is right, and a polished draft can carry a wrong date further than a rough one ever would. When it flags something, do not ask it to guess harder. Go to the source yourself, a calendar, a contract, the person who actually knows, and fix it in your own words.
That is the whole move. It takes ten minutes the first time, and far less after that, because you will start writing drafts that are easier to check. Shorter sentences. Named people. Dates you can point to. The pause does not slow you down for long. It saves you the email you would have had to send afterward to correct the first one.
Same move, five other places
- Use it on a work email before it goes to a client or a boss.
- Use it on a family group message before a holiday plan becomes a misunderstanding.
- Use it on a contractor quote before you agree to scope you did not mean.
- Use it on a volunteer announcement before names, dates, or times go public.
- Use it on a travel booking summary before you pay for the wrong dates.
Try this at home
Are you playing with Muse? (Yes, I’m still loving it.) Here is one of my favorite moves. If your Muse is connected to your Facebook and Instagram accounts, it can see the folders you have already been building for years. All those recipes you saved and never cooked… I have so many!
Ask it first whether it is connected. If it is, point it at your saved recipe folder and ask for an interactive guide or holiday guide to dinners under 20 minutes that fit a dietary need in your house. Gluten-free, dairy-free, low-sodium, whatever matters at your table. It will sort, filter, and lay it out. You check it, you keep what works, you throw out what does not. Your Muse becomes an interactive chat with all those saved recipes!
Are you connected to my Facebook and Instagram accounts? If you are, look through the recipes I have saved there and build me an interactive guide to dinners I can make in under 20 minutes that are [DIETARY NEED, FOR EXAMPLE GLUTEN-FREE]. Sort them by main ingredient. For each one, show the time, the ingredients I probably already have, and what I would need to buy. Flag anything you are unsure about, an ingredient or a time, instead of guessing. If you are not connected, tell me how to connect you.
Be the one who still checks
You do not have to be faster than the AI. You have to be the person in the room who still checks and asks for proof. Keep the pause, keep the questions, and let the tool do the heavy lifting in between.
I’ll keep sorting through the rest with you.
Cheers,