McDonald's tests ArchIQ, a Google-built AI assistant for drive-thru orders

McDonald's has begun testing a new AI system for taking orders and supporting restaurant operations. It is called ArchIQ and has the unofficial name Archy. The system was presented at the company's annual Worldwide Convention, Restaurant Business reports. According to the outlet, the test is running at five McDonald's restaurants in the US, but the company itself has not named the specific locations of the pilot.

What we know about Archy

ArchIQ is an AI system designed not only to take orders at the drive-thru window but also to support the restaurant's operational processes as a whole. Judging by the wording of the announcement, this is not a narrow voice assistant but a broader platform built into the daily work of a location, from talking to the customer at the pickup window to internal operational tasks for staff. Google is involved in the development. McDonald's is thus continuing to work with major technology partners on AI rather than building such systems entirely with an in-house team.

Key facts about the pilot:

• The system is called ArchIQ, with the unofficial name Archy.
• It was presented at McDonald's Worldwide Convention.
• It is being tested at five McDonald's restaurants in the US.
• The company does not disclose the specific locations.
• Google is supporting the development.

Why this is not McDonald's first attempt

McDonald's has already tried to automate drive-thru ordering with AI. The chain previously tested voice order recognition together with IBM, but that pilot was wound down: speech recognition in the noisy drive-thru environment and the variety of customer accents proved a serious technical barrier for the systems of the time. Returning to the idea with a new partner and, apparently, a new architecture suggests the company has not given up on automating this sales channel. It appears to treat its earlier experience as a basis for more cautious, phased testing, starting with a limited number of locations and without a broad public announcement of specific restaurants.

What this means for the fast-food industry

The drive-thru remains one of the main revenue channels for fast-food chains, and any technology that can speed up service or reduce staff workload at peak hours is of direct commercial interest to companies like McDonald's. At the same time, AI in fast-food restaurants is a sensitive topic for customers and employees: order recognition errors quickly go viral on social media, and talk of automation inevitably raises questions about the future of jobs for cashiers and drive-thru operators. A limited pilot at five locations without public disclosure of where they are looks like a cautious strategy: test the technology in real conditions and minimize reputational risk in case of failures before deciding on a wider rollout across McDonald's in the US and beyond.

McDonald's has tens of thousands of restaurants worldwide, and at that scale even a small reduction in the time it takes to serve a single drive-thru order can add up to a noticeable effect on restaurant throughput and customer satisfaction during peak load. The fast-food industry as a whole has long been moving toward digitizing customer touchpoints: self-service kiosks in dining rooms, mobile apps for ordering and payment, and personalized offers based on purchase history have already become routine. A voice AI assistant at the drive-thru logically extends that trend, but because it interacts directly with the customer's voice in real time, it carries markedly higher technical and reputational risks than digital tools that are less visible to consumers, such as those inside apps and self-service kiosks.

Testing at several locations rather than a single pilot site also lets McDonald's and its technology partner collect data on how the system performs in different conditions, from how busy a particular restaurant is to regional differences in pronunciation and how orders are phrased, before deciding whether to scale. If the pilot succeeds, the next logical step would be a gradual expansion of the number of locations using ArchIQ. If it does not, the company could wind down the experiment as quietly as it began, without drawing extra attention to the failure, unlike the earlier episode with IBM.