AINEWS
McDonald's Tries Drive-Thru AI a Second Time, With Google and a 'Master Brain'
ArchIQ is being tested in five locations and pitched as kitchen operations software, not just an order taker, after the IBM system became a viral punchline.
ArchIQ is being tested in five locations and pitched as kitchen operations software, not just an order taker, after the IBM system became a viral punchline.
McDonald's is putting a new voice assistant in the drive-thru for the second time, and the company wants you to know this attempt is different. The system is called ArchIQ, the drive-thru persona is nicknamed Archy, and the partner this round is Google, not IBM. More importantly, in the system message introducing "McDonald's > NEXT", Chairman and CEO Chris Kempczinski pitches ArchIQ as something bigger than a voice box at the order point. He frames it as a "master brain" for restaurant operations, built to flag a freezer going down, surface kitchen bottlenecks, and give managers a real-time view of what is happening across the lane.
That is a structurally different product claim than the first time McDonald's tried this. The IBM-powered voice AI was rolled out to more than 100 restaurants before McDonald's pulled it in 2024, after viral videos showed the system stacking hundreds of chicken nuggets onto a single order and producing other order-mistake failures. The product that survived that episode is, in the company's telling, no longer a drive-thru voice at all. It is a back-of-house operating layer that also takes your order.
The deployment, for now, is small. Restaurant Business reports that the voice-activated drive-thru is being tested in five locations in partnership with Google, and that the test is a sliver of the broader ArchIQ rollout. The framing is deliberate. McDonald's learned that pitching AI to a public that just watched a machine mishear "chicken nuggets" three times in a row is a hard sell. Pitching an AI that tells a manager the freezer is failing before the lunch rush is an easier one.
The case for the new system leans heavily on a single source. The "McFranchisee" account on X, representing a major McDonald's franchisee, has claimed the AI has now processed more than 1 million orders, with roughly 90% requiring no human intervention. That figure is not independently confirmed by McDonald's investor relations, Google, or ArchIQ press materials in the public record. It is a social-media claim from one franchisee defending a system that franchisee is rolling out. Treat it as a signal of internal enthusiasm, not as a benchmark.
What is on the record is the scope expansion. ArchIQ is being positioned as the connective tissue between the order point, the kitchen line, and the equipment that keeps a restaurant running. In Kempczinski's framing, the system's job is to know, in real time, what is going wrong in the building so a manager does not have to find out the hard way during the dinner rush. That reframes the labor question in a way the IBM rollout never did. The first generation of drive-thru AI was sold as a way to remove the crew member at the order point. ArchIQ is being sold as a way to give the manager a second pair of eyes across the entire back of house. Whether that is a genuine product improvement or a relabeling of the same idea is the question the five-location test is supposed to answer.
The competitive backdrop is explicit in the > NEXT announcement. McDonald's is upgrading its menu, restaurant prototype, and hospitality program in part to fight off chicken, beef, and beverage specialists like 7 Brew, Dave's Hot Chicken, and Five Guys, who have been pulling in customers the chain's standard offering does not serve. AI is one of several bets in that response, not the whole strategy, and the company is clear about the menu and prototype pieces. The technology piece is the one still being argued over inside the system.
Customer and worker reaction captured on social channels runs heavily skeptical. Customers complain about voice systems mishearing them in noisy lanes. Workers and labor observers raise the longer-running concern that every generation of restaurant automation has been sold as a way to help crew, and the trend line of the last decade has not been kind to that framing. The franchisee counter-argument, again per the McFranchisee account, is that the harder part of the lane job is hearing orders correctly in noise, and that employees have been asking for help with it. Both of those claims are worth taking seriously. Neither one is settled by the available evidence.
What to watch next: whether the five-location test produces any operator-side data beyond the franchisee social-media claim, whether Google publishes a cloud or AI announcement that locks in the technical relationship on the record, and whether the > NEXT rollout reaches a second wave of restaurants with the same product, or quietly narrows back to a single voice-AI feature. If none of that materializes, the most likely explanation is that the master-brain framing is the marketing layer on top of a second attempt at the same voice-AI idea. If it does, the McDonald's AI story has a real chance of being a different story from the IBM one.