If you run a $1M–$20M services business — home services, real estate, professional services, retail, medical spas — this week of AI news mattered more than usual. Anthropic just passed OpenAI in revenue, voice cloning hit consumer-tier pricing, and ChatGPT quietly swapped the model running inside the tools you already pay for. Here’s what it means for the way your business buys, runs, and sells in the next 90 days.
The five questions to think through this week:
- Did anyone on your team notice that ChatGPT switched models on you this week — and do you have a way to find out next time?
- If you’re running on Excel or Google Sheets, when are you putting the new ChatGPT sidebar in front of one of your finance or ops people?
- With Grok’s Custom Voices going live, what’s your written policy on AI voice cloning before someone in your business uses it on a customer call?
- Should one of your Mac-using executives be the test case for Perplexity’s new Personal Computer agent?
- Anthropic just hit $30B ARR and locked up 220,000 GPUs — does that change which AI vendor you should be standardizing on for the next two years?
1. Did anyone on your team notice that ChatGPT switched models on you this week?
OpenAI quietly made GPT-5.5 Instant the new default model inside ChatGPT this week — no announcement banner, no popup. If you or anyone on your team built workflows, prompts, or custom GPTs against the old default, the model answering you today is not the one you tested against six months ago. OpenAI also shipped Codex updates and started testing ads inside ChatGPT in the same news cycle.
For a services business, this matters for a small but real reason: the prompt you wrote to draft your sales follow-up emails, summarize Zoom calls, or triage customer questions might be quietly behaving differently right now. A roofing company we hear from automated their estimate-summary process last fall — it would have started producing slightly different outputs this week without anyone touching it. Multiply that across your business and you have silent drift.
The action this week is simple: pick the three or four AI workflows you actually rely on, run a known input through each, and check that the output still looks like what you expect. Make this a monthly check, not a one-time scramble. If you’re on a Team or Enterprise plan, those workspaces give you better controls over which model runs and when it changes.
2. If you live in Excel or Google Sheets, when are you putting the new ChatGPT sidebar in front of your finance or ops lead?
The single most consequential rollout for executives this week was the global launch of ChatGPT for Excel and Google Sheets. It runs as a sidebar inside the spreadsheet itself, so your CFO, controller, or operations manager doesn’t have to copy-paste data into a chatbot to ask, “what’s driving the variance in May service revenue?” or “rebuild this pricing model for a 12% labor cost increase.”
This is huge for a few specific business types. Real estate agents and brokerages live in spreadsheets — comparative market analyses, commission splits, deal pipelines. Independent insurance agencies, bookkeepers, financial advisors, and small accounting firms run their entire client work out of Excel. Med spa and dental practice owners build out membership pricing models and inventory plans in Sheets. For those owners, the ChatGPT sidebar is not a “future of work” abstraction — it’s a tool sitting in the file you already have open.
The smart move is to designate one champion per function — one person in finance, one in ops — to use it every day for 30 days and report back what saved them time. Don’t try to roll it out across the whole company at once. The team that has a real workflow story will spread it for you.
3. With Grok’s Custom Voices going live, what is your AI voice policy?
xAI shipped Custom Voices this week alongside Grok 4.3, with native video and a 1M-token context window. The Voice Library is live and Grok is rolling into Apple CarPlay. In the daily executive brief, the call-out was direct: voice cloning at console-tier pricing creates governance gaps that need acceptable-use updates.
For service businesses with a phone-heavy front door, this is the question of the quarter. HVAC companies, plumbers, electricians, roofing contractors, pool service companies, landscapers, med spas, dental practices, and law firms all sit on the same dynamic — a receptionist answers the phone, schedules the work, and quotes the price. AI voice agents are now cheap enough that someone, somewhere on your team will try to clone your voice or your top tech’s voice and use it on outbound or inbound calls. Without a written policy, you’ll find out about it after the customer complaint, not before.
You don’t need a 20-page document. You need 30 minutes with whoever owns HR and IT to write down: who is allowed to deploy a voice agent in your name, what disclosure customers get, where recordings live, and what happens if an employee uses voice cloning without approval. Get that on paper before the first booking comes in from an AI receptionist you didn’t authorize.
4. Should you put Perplexity’s Personal Computer agent in front of one of your executives?
Perplexity opened its Personal Computer macOS app to all Mac users this week, ending the prior $200/month paywall, and ARR passed $450M with 45M users. The product is best understood as a persistent desktop worker — it sits on your Mac, watches what you’re doing, and can take actions across email, browser, documents, and local files. Perplexity has effectively repositioned from “answer engine” to a competitor with Microsoft Copilot, Claude, and traditional RPA tools.
Why this matters for a $1M–$20M business: the cheapest, most tangible productivity win for an owner-operator is offloading the 90 minutes a day you spend on email triage, calendar wrangling, lead-list research, and competitive intelligence. A persistent agent that works on your actual desktop is a much shorter walk to that win than a chat window in a separate tab. If you or your top operator is on a Mac, this week is the right week to spin up a 30-day test. Pick one workflow — inbound lead enrichment is a great one — and measure how many minutes per week it actually claws back.
If your business is on Windows, the same thinking applies, but the Microsoft Copilot stack is your starting point, not Perplexity. The point is the same: persistent AI on the desktop, doing real work, is now mainstream.
5. Anthropic just hit $30B ARR and locked up 220,000 GPUs — does that change your vendor strategy?
The headline number of the week: Anthropic surpassed OpenAI in annualized revenue, hitting $30B (3.3x growth from year-end 2025) versus OpenAI’s $25B. Anthropic also signed a $1.8B compute deal with Akamai and locked up the entire Colossus 1 facility from xAI/SpaceX — 220,000 GPUs and 300MW of capacity. Claude Opus 4.7 went generally available, Claude Design launched, and Claude Security is now in beta. Meanwhile xAI itself was folded into SpaceX as “SpaceXAI.”
Translation for your business: the AI vendor landscape just got more concentrated, and the historical assumption that OpenAI is the obvious default is no longer obviously true. For most service business owners, the practical implication isn’t to pick a winner — it’s to stop locking yourself into one tool. If you’re standardizing on ChatGPT Enterprise, ask a vendor like Claude (or a Claude-powered tool) to bid on the same workflow this quarter. If you’re using Microsoft Copilot, run a side-by-side test on a real internal task. The vendors are competing hard for your business right now, which means pricing and feature concessions are available if you ask.
Also worth noting: DeepSeek shipped V4-Pro and V4-Flash this week with a CAISI evaluation completed, and Mistral shipped Medium 3.5 at 77.6% on SWE-Bench. The point isn’t that you should run a Chinese or French model — it’s that closed-model vendors now have to justify their pricing premium to procurement, which is a great thing for your AI bill.
What This All Means
This was a consolidation week disguised as a product week. Anthropic and OpenAI are pulling away on revenue and compute, voice and desktop agents are becoming consumer-priced, and the model running inside the AI tools you already pay for is changing without telling you. For a $1M–$20M business, the right reaction is not to chase every announcement — it’s to pick one or two of the questions above, get a 30-day pilot in motion, and write down a basic AI usage policy before someone on your team beats you to it.
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 →