I've seen this happen a lot: someone takes a messy manual process and bolts AI onto it. Now the mess just happens faster. Before you automate anything, write the steps on paper first. If you can't explain the process to a friend in plain words, the automation will be a mess too. Simple example: following up with leads. The messy version is "check my notes, maybe text them, maybe email them, depends on my mood." The clean version is "every new lead gets a text in 5 minutes, an email in 1 day, and a call reminder in 3 days if no reply." Once the steps are that clear, the automation is almost easy to build. My rule now: automate the second version of a process, never the first. The first version is where you learn what the process actually is.

Alright… this one's for anyone who does good work and then never asks for the review. Salons, trades, local service businesses — the job's done well, the customer's happy, and nobody ever types up the Google review because the window to ask quietly closed. So you build it once and it does the asking for you, the morning after every appointment. Here's what it does, in plain english. You keep a booking log in a Google Sheet — customer name, email, service, appointment date, a Status column. Every morning the automation looks at that sheet. Any appointment that happened yesterday and hasn't been asked yet, it hands to ChatGPT. ChatGPT writes a short, warm review request matched to the service they actually booked. Gmail sends it. Then the row gets marked "Asked" so the same customer never gets pinged twice. The services it touches, three of them: 👉 Make.com — https://www.make.com 👉 Google Sheets — https://sheets.google.com 👉 ChatGPT via the OpenAI API — https://platform.openai.com The build, start to finish: 1. Set up the sheet. Columns for Name, Email, Service, Appointment Date and Status. Put a drop-down on Status with Booked, Asked and Done. Your booking tool probably already fills most of this — point the automation at wherever that lands. 2. 2. In Make.com, create a new scenario. First module is Google Sheets, action Search Rows. Point it at your sheet and tab. 3. 3. In that module's filter, set Appointment Date to yesterday, and Status "not equal to" Asked. That's your "happened yesterday, not asked yet" list. 4. 4. Add the OpenAI module, action Create a Chat Completion. The prompt is where the tone lives. Something like: "Write a short, warm email to {{Name}} thanking them for their {{Service}} yesterday and asking if they'd leave a quick Google review. Keep it human, one paragraph, no pressure." 5. 5. Add the Gmail module, Send an Email. Map the customer email into To, and the ChatGPT output into the body. Drop your actual Google review link into the body so they can click straight through. 6. 6. Add a Google Sheets Update a Row at the end. Set Status to Asked on that row so it drops off tomorrow's list. 7. 7. Hit Run Once and check the email actually lands and reads like a person wrote it. Then set the scenario to run once a day in the morning and turn it on.

Alright… this one's for everyone who'd rather eat the loss than send the "just following up on invoice #0412" email. The chasing is boring, it feels rude, and so it slips. So you build it once and it does the asking for you. Here's what it does, in plain english. You keep your unpaid invoices in a Google Sheet — client, amount, due date, email, a Status column. Every day the automation looks at that sheet. Any row where the due date has passed and the status isn't "Paid", it hands to ChatGPT, which writes a reminder matched to how late it is. A day over gets a gentle nudge. Three weeks over gets firmer. Gmail sends it, and the row gets marked "Chased" so the same client doesn't get pinged twice. The services it touches, three of them: 👉 Make.com — https://www.make.com 👉 Google Sheets — https://sheets.google.com 👉 ChatGPT via the OpenAI API — https://platform.openai.com The build, start to finish: 1. Set up the sheet. Columns for Client, Email, Amount, Due Date and Status. Put a drop-down on Status with Unpaid, Chased and Paid. Add conditional formatting on Due Date so anything past today turns red — that way the sheet is readable even without the automation. 2. In Make.com, create a new scenario. First module is Google Sheets, action Search Rows. Point it at your sheet and tab. 3. In that module's filter, set Due Date "earlier than" now, and Status "not equal to" Paid. That's the "overdue and still owed" list. 4. Add a module to work out how overdue each one is. A Set Variable with a formula on the due date gives you a days-late number. You feed that into the next step. 5. Add the OpenAI module, action Create a Chat Completion. The prompt is where the tone lives. Something like: "Write a short, polite payment reminder to {{Client}} about invoice for {{Amount}}, now {{days late}} days overdue. Friendly if under a week, firmer if over three. Keep it human, no threats." 6. Add the Gmail module, Send an Email. Map the client's email into To, and the ChatGPT output into the body. 7. Add a Google Sheets Update a Row at the end. Set Status to Chased on that row so it drops off tomorrow's list. 8. Set the scenario to run once a day and turn it on.

Welcome to the four new people who joined this week — glad you're here. Something I notice when I talk to people about AI: everyone's sitting in a different spot. Some can't build fast enough. Some are quietly overwhelmed by all the noise. Some want it proven before they buy in. And some are still working out what the thing even does. All four are normal. None of them is behind. So here's the one I'm curious about — new people, this is aimed at you first, but everyone jump in: Which one are you right now? 1) excited, can't build fast enough 2) curious but a bit overwhelmed 3) a little skeptical, prove it to me 4) honestly still figuring out what it even does Vote in the poll, then drop a comment with why you picked it. Tell me and I'll point you at the best place to start 👇

Alright… this one isn't about making the image. It's about the part right before you hit publish — the part that saves you from a public mess. Here's the thing. AI invents text that looks right at a glance. Your product name, your phone number, the date — it'll happily spell them wrong in a way you won't catch unless you actually look. So before anything goes out, run this: 1. Read every word on it, out loud. Your own product name. The phone number. The date. "Speciallty Coffee" looks fine until you say it. 2. 2. Zoom in on hands, faces, logos and edges. That's where generation falls apart — six fingers, a melted logo, a doorframe that bends. 3. 3. Is it still your thing? Your product, your face, your premises — recognisably yours, not an improved stranger. 4. 4. Check it at real size. Phone screen for social, arm's length for print. Not on your laptop at 100%. 5. 5. Is any number or claim on it true? You're publishing it, so you're the one saying it. 6. 6. Would a customer feel misled seeing the real thing? That's the honest test for every enhanced photo. And if something's off, you don't start over. You say what to fix in the same chat: Fix the text to read "[exact correct words]". Keep everything exactly the same, just [one change]. Remove the [object] from the [location]. The full walkthrough, plus the repair prompts and the sizes you actually need, is here: https://www.skool.com/ai-for-beginners/classroom/fb16f17c?md=e252798b5b294c31b1405ea122abbba0 One thing to know — the video on that page was recorded a while back, so some menus and model names look different now. Follow the method, not the exact screen