AI Decides. People Push Back.

McDonald's lets software set what a Big Mac costs, and franchisees say they can't say no. OpenAI launched Dots, AI agents that keep working after you log off. And Apple shelved a plan to replace 5,000 support employees with AI, at least for now. This week, AI moved from advising people to deciding for them, and the humans started asking who's in charge.

The Bottom Line

Each story hands software a decision: a price, an invoice, a customer call. The winners won't be the companies that hand over the most. They'll be the ones that write down what the software may do alone and what needs a human yes. Tell me in the comments: would you let AI set your prices if you approved every change first? Or does pricing by algorithm cost you your customers' trust?

See what we build →

Top 3 Stories This Week

01 · McDonald's · Reuters · AI Pricing

Two Miles Apart, Two Big Mac Prices

A Big Mac cost $5.69 at one company-owned McDonald's in Fresno, California. Two miles away, it cost $6.89. That is a 21% gap for the same burger. Reuters reported on September 29 that software drives the difference. McDonald's runs machine learning (programs that find patterns in data) across nearly 14,000 restaurants. It weighs local sales, rival prices at Wendy's and Burger King, and what customers nearby will pay. Franchisees, the independent owners who run most stores, say the "optional" tool isn't. Since January, McDonald's has required them to "constructively engage" with its pricing tools, and five owners told Reuters the company pressured them to follow it. McDonald's calls the portal "a tool, not a mandate." The risks are visible. The system once recommended $18 for a single Big Mac in Connecticut, and the portal's own legal terms warn owners they "may be competitors" and should consult attorneys about antitrust. Your takeaway is not to copy the giant. It is that pricing by gut now competes with pricing by data.

Microservice Opportunity: Build a price guardrail, not a price robot. A microservice (one small AI tool doing one job) pulls your sales history and competitor prices every week, then proposes changes to the 10 items where you likely leave money on the table. It caps every suggestion inside a range you set, explains each one in a sentence, and waits for your yes. A café or salon gets chain-level pricing insight without an $18 Big Mac.

02 · OpenAI · DevDay · Always-On Agents

OpenAI's New Employee Never Clocks Out

On September 29, at its DevDay developer conference, OpenAI introduced Dots: AI agents that keep working after you close your laptop. Each Dot runs on its own cloud computer, is powered by GPT-6 Astra, and connects to more than 4,000 apps. OpenAI's examples read like a small-business to-do list. Notice an invoice nobody sent, draft it, and wait for approval. Chase a bug mentioned in Slack. Research a topic before you ask. You set the rules for when a Dot acts alone and when it must ask first, and you reach it through ChatGPT, Slack or Teams. Dots are live for Pro and Business Premium subscribers, and for Enterprise customers once an admin approves, with one Dot per plan for now. OpenAI also released GPT-6.1 Sol, which it says nearly matches Astra at one-fifth the token price, and a $500-a-month Pro 500 plan. The catch: a general agent knows your apps, not your business. It doesn't know the client who always pays late deserves a phone call, not a form email.

Microservice Opportunity: Give the general agent a specialist. Build a microservice for one workflow, such as collections. It reads your accounting system, knows each client's payment habits and terms, drafts the right reminder in your voice, and posts it for one-click approval. Keep it separate from any single vendor's agent so you can switch tools later. You get always-on follow-up that holds your rules, and cash arrives sooner without you nagging.

03 · Apple · AppleCare · AI Support

Apple Blinked on 5,000 Support Jobs

Apple planned to replace about 5,000 AppleCare support employees with AI phone and web agents. Then it stopped. Bloomberg's Mark Gurman reported on September 29 that Apple put those plans on hold indefinitely and is no longer "currently looking to take such a step." Bloomberg gave no reason, and the report says Apple could revisit the idea. The plan wasn't a fantasy. Apple already runs generative AI assistants on its AppleCare phone line that troubleshoot problems and pass callers to a person when they get stuck. Apple has the money, the engineers and years of customer data. It still hesitated. That should give a smaller company pause. Replacing a whole support team in one move is a bet that the AI never fumbles an angry customer at 6 p.m. on a Friday. The safer path is the one Apple already runs: let AI take the first pass, and keep a person on the handoff. Where does your team draw that line?

Microservice Opportunity: Build the handoff, not the replacement. A microservice answers your top 20 repeat questions (hours, order status, warranty, rescheduling) from your own documents. The moment it detects anger, a refund request or two failed answers, it passes the conversation to a person, who sees the full chat and a suggested reply. First response drops to seconds, and your people spend their time on the calls that decide whether a customer stays.

This week in numbers: 21%: the Big Mac price gap between two Fresno McDonald's two miles apart. · 4,000+: apps a single OpenAI Dot can connect to while you're offline. · 5,000: appleCare jobs Apple shelved replacing with AI agents, for now.

Sources

TL;DR — Key Takeaways

Top 3 Stories This Week

Two Miles Apart, Two Big Mac Prices

OpenAI's New Employee Never Clocks Out

Apple Blinked on 5,000 Support Jobs

Related answers