Every week the AI headlines come at you like a fire hose nobody asked someone to turn on, each one engineered to make you feel behind before your coffee’s even cool. I don’t run my business on that feeling, and I don’t want you running yours on it either. So here’s the deal I keep with you every week: fifteen years in marketing, ten running my own shop, I read the week’s AI news so you don’t have to, I throw out the hype, and I hand you back the two or three things that actually matter when you sit down to work.
Lately the thread has been the money getting nervous: AI stocks tumbling, a billionaire down eight billion in a single day, and that quiet number showing most company AI projects make no money. This week the thread moved somewhere much closer to home with the kind of AI that doesn’t just answer your questions but actually does things for you, (sends the email, books the flight, makes the purchase). And it officially arrived in your pocket as Apple put one in your phone. Meta launched an AI that’ll spend your money, and it’s started unsettling conversation over who’s on the hook when one of these things gets it wrong.
1. Apple put a real AI assistant in your phone and it can act inside your apps
Here’s the one that landed in the most pockets. On September 14th, Apple released iOS 27, and with it the long-promised, rebuilt Siri AI: Apple calls it “a profoundly more capable and personal assistant,” and for once the marketing isn’t totally off. This isn’t the old “sorry, I didn’t catch that” Siri. The new one can read what’s on your screen, dig through your own messages, emails, and photos to answer you, and take actions inside your apps. You can be looking at a team’s website and ask it to find the home games and drop them on your calendar. You can ask it to pull the recipe out of an email and add the ingredients to your reminders. It’ll even draft an email and try to match how you normally write to that specific person.
You don’t have to care about Apple to care about this. You care because it means the “AI agent” (the version that acts on your behalf instead of just chatting) is no longer a Silicon Valley demo. It’s a software update that rolled out to a few hundred million phones overnight. (Fair warning: it needs newer hardware, roughly an iPhone 15 Pro or newer, and it’s launching in English beta first, so if yours hasn’t changed, that’s why.)
What it means for you: An assistant that reads your screen and drafts a reply in your voice is a fantastic way to kill the little friction tasks that eat your day: turning a message into a calendar event, getting a rough first draft of a boring email, pulling a detail out of a long thread. So give it a shot if you’d like! But notice the line it’s quietly asking you to cross: to be useful, it wants access to your messages, your email, your photos, the whole inside of your business. That’s a fair trade for the small stuff. It is not a reason to let it start writing the emails that actually matter: the client note, the apology, the pitch. Those are the ones people can tell you didn’t write. Let it clear the noise. Keep the words that carry weight. (Credit where due: Apple says the personal stuff processed in its Private Cloud Compute isn’t stored or made accessible to Apple or anyone else, a good example of the kind of plain data answer you should be looking for, which brings us to the next one.)
2. Meta’s Muse will send your emails and spend your money, and even Meta’s own people got nervous
Here’s the one to be a little careful with. On September 8th, Meta launched Muse, which it’s billing as a personal AI agent “built for everyone.” Muse goes further than Siri, it connects to your apps and services and will, in Meta’s own words, send emails, book travel, lower your bills, fill out forms, make plans, turn a recipe reel into a grocery list, send party invitations, and make purchases. It’s free to start (there are $20 and $100-a-month tiers for heavier use), and it asks for a payment card up front.
Read that list again, because it’s a lot. You’re handing an AI the keys to your inbox, your calendar, and your credit card, and telling it to go act like you. And here’s the part that made me raise an eyebrow: according to Forbes, Meta’s own staff flagged security concerns as it launched. Meta says Muse runs on a “dedicated, secure computer with its own browser” that can’t see your passwords or payment details, and doesn’t feed your data to its ad machine — but reporters were quick to note those claims still need real outside security experts to verify them, and this is the same company that’s paid billions in privacy penalties over the years.
What it means for you, and what to actually do: I’m not telling you to never touch it. But maybe run it through the boring questions before it runs your errands. When any tool wants to act on your behalf, ask: what does this company actually do with my data, can they say it in plain English, and has anyone outside the company checked? A tool whose own employees are nervous on launch day has not earned your client list yet.
My honest rule: I don’t let anything I can’t fully explain touch my inbox, my clients’ information, or my card, and I never let an AI make a purchase I didn’t personally approve. Let these agents draft, sort, and suggest all day long. Don’t let them spend and send without you in the loop. The convenience is real. So is the cost of a bot firing off the wrong email to the wrong person under your name.
3. Nobody’s quite sure who pays when the AI buys the wrong thing
Here’s the one that turns points one and two from “neat” into “pay attention.” Now that these agents can actually make purchases, a very unglamorous question has moved to center stage: when an AI agent buys the wrong thing or gets tricked into paying a fake invoice, who’s on the hook? As The Financial Brand laid out this week, the payments world doesn’t have a clean answer yet. Amex took a swing at it earlier in 2026 with an “agentic commerce” kit that promises to cover erroneous purchases made by registered AI agents, but experts pointed out the word conspicuously missing from the announcement was fraud. The systems that catch bad transactions today can’t reliably tell the difference between your legitimate AI agent and a hijacked one, which means every one of these automated purchases carries a little cloud of “…but who authorized that, really?”
If that sounds abstract, make it concrete: an AI agent that can be told to pay an invoice can also be fooled into paying a fake one. Scammers are already counting on it. The whole appeal of these agents (they act fast, on their own, so you don’t have to look) is the exact same thing that makes a bad instruction sail right through.
What it means for you and what to actually do: Automation is a gift for the repetitive parts of running a business: scheduling, reminders, sorting, first drafts. It is a genuine liability for the parts involving money and judgment. So draw the line clearly in your own shop: let AI prepare the payment, the order, the invoice – and keep a human (you) as the one who hits send. Anything with a dollar figure or a client’s name on it gets a pair of human eyes before it goes out.
The part I actually care about: the tools got hands this week, keep yours on the wheel
So here’s the question that actually matters for you: when a tool will happily do it all, what do you refuse to hand over? Because using AI responsibly means letting these agents clear the busywork, (the calendar wrangling, the first drafts, the forty little digital chores that pile up when you run everything solo), while you keep a firm grip on the three things that were never theirs to take. Your judgment, so you catch it when the agent is confidently wrong. Your data and your money, so a slick convenience doesn’t quietly become a breach or a bad charge. And your craft: the writing, the design, the relationships. None of that requires you to be technical. It requires you to stay the person who still knows how to do the work, so you can point these tools at the noise and instantly tell when one hands you back something wrong.
Learning to use AI unfortunately isn’t optional anymore. Learning to use it responsibly is the whole game.
Still don’t want to use AI but need help running the business?
If this left you thinking “okay, I get it, use the helpful parts, guard my data and my money, don’t hand over the stuff that’s actually mine, but I can barely keep up with the marketing I already do, let alone figure out which tasks are safe to hand off or how to set it up without wasting money or making a mess,” that’s exactly what I do.
I help small business owners build their online presence, with real humans that do the work, while also using AI tools and automation the smart way: strategy first, tools second, your skills and your voice protected, and you always in control of your own accounts, your own data, and your own work.
If you’re ready to jump into the modern age but don’t want to spend any time learning something new, let our team be the bridge that helps you get there.