This week, AI stopped being a website you visit and started becoming something that works inside the tools you already use — your inbox, your text messages, your browser. At the same time, running powerful AI got noticeably cheaper for some businesses and far more private for others. If you run a $1M–$20M business, a few of this week’s headlines land right on your actual workflow, so here are the five questions worth asking right now.
- Should I let AI start reading and answering my customer messages?
- Why should medical spas, law firms, and accountants care about “local” AI that never touches the cloud?
- Is the price of AI about to drop for my business — or go up?
- What do “AI agents” actually mean for a real estate agent (or any lead-driven business)?
- If AI can now run a science lab on its own, what could it run in my business?
1. Should I let AI start reading and answering my customer messages?
This week OpenAI began rolling out a ChatGPT integration for Apple Messages on Mac — once you give it permission, it can read, search, summarize, draft, and even send texts. Perplexity did something similar, letting its “Computer” agent kick off tasks straight from your email. For a home services business — HVAC, plumbing, roofing, a pool company — where half the day is answering “can you come Tuesday?” texts and chasing quote follow-ups, this is the first AI news in a while that hits the work you actually do.
The upside is concrete: an assistant that drafts replies to routine inquiries, chases the estimates you sent last week, and summarizes a long back-and-forth before you call a customer back. xAI also shipped a faster voice model this week, pointing to the same thing eventually reaching phone calls — the part of a service business that’s hardest to staff.
The catch is real, though. The moment you let AI read and send messages, you’ve opened a door. Security researchers flagged that broad message access creates a “prompt-injection” risk — a cleverly worded incoming message could trick the AI into doing something you didn’t intend. So the answer isn’t “no,” it’s “start small.” Pick one inbox or one team, decide in writing what the AI may do on its own (drafting a reply, yes; hitting send to a customer, maybe not yet), and keep a human in the loop on anything that touches money or the schedule.
2. Why should medical spas, law firms, and accountants care about “local” AI that never touches the cloud?
Here’s the quiet but huge story of the week. Google’s open Gemma models passed one billion downloads, Meta released a new “Muse” family designed to run on a laptop, and European players Mistral and Cohere leaned hard into “sovereign” AI that a customer controls end to end. Translated for a small business: you no longer have to ship your data off to a giant cloud service to use capable AI.
Why does that matter to a med spa, a law firm, or an accounting practice in particular? Because you’re sitting on exactly the kind of information you can’t afford to leak — patient records, client financials, case details. Cloud AI tools can be a compliance headache in those worlds. A model that runs on your own computer, offline, means client information never leaves the building. It’s frequently cheaper, too, because there are no per-question fees.
You don’t need to become a tech company to take advantage of this. The practical move is to ask any vendor pitching you an “AI tool” one question: where does my data go? This week’s news means “it stays on our hardware” is now a realistic, affordable answer — and for a regulated business, that should push a tool straight to the top of your shortlist.
3. Is the price of AI about to drop for my business — or go up?
Both, depending on who you buy from — which is exactly why it’s worth ten minutes of your attention. xAI launched Grok 4.6 this week priced more than 60% below comparable tools from OpenAI and Anthropic. DeepSeek rolled out “off-peak” pricing, charging half rate for work that can run overnight. Meanwhile Anthropic, the maker of Claude, announced a price increase starting September 1.
The lesson for an owner isn’t “switch to the cheapest option.” It’s that AI pricing is now competitive enough that you have leverage. If you’re paying for AI writing, customer service, or research tools, this is a good month to ask your vendor what you’re actually paying per use and whether there’s a cheaper tier — or a competitor you can name. Even if you never switch, knowing Grok is 60% cheaper is a negotiating chip.
And if some of your AI work isn’t time-sensitive — generating a month of social posts, cleaning up a customer list, drafting bulk emails — DeepSeek’s off-peak model is a reminder that “run it overnight for half price” is now a genuine option. Small savings, but they add up over a year.
4. What do “AI agents” actually mean for a real estate agent (or any lead-driven business)?
“Agent” was the word of the week, and it means something more useful than a chatbot. Perplexity opened its AI-native “Comet” browser to businesses — it can run multi-step research and complete tasks inside a webpage, not just answer questions. xAI rolled out “Grok Bot” teammates that work alongside your staff. The shift is simple: instead of you asking the AI a question, the AI does a job.
If you sell real estate, picture what that looks like: an agent that pulls comparable sales, drafts a listing description, and lines up follow-up emails to every lead who toured over the weekend — while you’re at a closing. The same pattern fits anyone whose revenue depends on chasing leads: a landscaping company quoting jobs, an insurance broker, a mortgage office.
The honest caveat is that this is early, and these tools do best on repetitive, well-defined tasks. Don’t hand an agent your whole business. Do pick one annoying, repeatable job — lead follow-up is the classic — and test whether an agent can take the first 80% of it off your plate. That’s where the return is right now.
5. If AI can now run a science lab on its own, what could it run in my business?
The most jaw-dropping news of the week: Anthropic showed its Claude AI autonomously running a protein-design experiment — designing more than 1,300 candidates and getting working results on 14 of 15 targets, at hit rates roughly double the human norm. It didn’t assist a scientist; it ran the pipeline from end to end.
You’re not designing proteins. But the principle matters: AI has crossed from “helps with one step” to “can own a whole multi-step process.” So the question for your business gets concrete — what repeatable, multi-step process eats your team’s time every week? Onboarding a new customer, assembling a proposal, reconciling invoices, turning a finished job into a review request? Those are the workflows worth handing to AI first.
One more thing this week drove home: Anthropic is reportedly preparing a stock-market debut that could rival SpaceX’s, and its rivals are raising billions. The AI companies you rely on are becoming huge, fast-moving businesses in their own right — mostly good news, but also a reason not to bet your whole operation on a single tool. Keep a backup, and don’t wire anything in so deeply that you couldn’t switch.
What This All Means
Three things happened this week: AI moved into the tools you already use, it became cheaper and more private for the businesses that need those things most, and it proved it can run entire workflows rather than just answer questions. None of that requires you to overhaul your business tomorrow. But the “wait and see” window is quietly closing — the owners who pick one workflow, test one tool, and set one clear rule about what AI is allowed to do will be a full year ahead of the ones who don’t.
Not sure how any of this applies to your specific business? That’s exactly what our free audit is designed to answer. In 30 minutes, we’ll map out where AI can realistically save you time, generate leads, or give you visibility you don’t currently have — no jargon, no pressure. Book your free audit →