The last time McDonald’s handed its drive-thru lane to an AI, the experiment ended in 2024 after more than 100 test locations and a string of reported order errors. Now the company is back with a new AI voice assistant called Archy and a broader push into automated restaurant operations.
What Archy Actually Does
Archy is McDonald’s new drive-thru voice assistant, and it is part of a wider operating platform built with Google.
Archy takes your drive-thru order in English or Spanish and, according to McDonald’s own testing, has reached roughly 90% accuracy. It also handles upsell suggestions, nudging customers toward an additional item before the order closes.
Archy is the customer-facing piece of a broader system called ArchIQ, built in partnership with Google. ArchIQ is designed to connect front-of-house ordering with back-of-house operations. It handles inventory tracking, employee scheduling, and order verification tools that can flag missing items before the bag reaches your window.
The $8.5 billion figure attached to this rollout needs context. That commitment reflects projected franchisee support, including rent relief and capital assistance for restaurant modernization, through 2036, with about $5 billion expected through 2030. Archy is one component of that program, not the entire purpose of the spending.
Whether It Actually Works Is a Separate Question
McDonald’s internal accuracy claim and independent benchmark data tell meaningfully different stories.
McDonald’s 90% accuracy figure comes from its own early testing, not an independent assessment. Independent data offers a more complicated picture. Intouch Insight’s 2025 drive-thru study examined 120 AI-enabled orders across three quick-service restaurant chains and found an 83% accuracy rate for AI-handled orders, compared with 87% accuracy overall for all drive-thru visits in the study. It is worth noting that the benchmark measured AI ordering across multiple chains, not Archy specifically.
Customization errors were the most common failure point in that study: extra sauce, no pickles, a substitution on a combo. These are exactly the kinds of modifications that complicate your order and expose the limits of voice recognition in a noisy drive-thru lane.
Speed was the AI’s clearest advantage. AI-enabled drive-thrus in the study were about 21 seconds faster on average. When human employees stepped in to assist, accuracy climbed to 95%, which suggests the technology currently works best as a collaboration rather than a standalone replacement.
What This Means for Employees and Franchisees
McDonald’s says labor hours freed by Archy would be redirected, not cut, though that position has not been tested at scale.
Roughly 50 labor hours per week could be freed at a typical restaurant, according to McDonald’s. The company’s stated position is that those hours would be redirected toward food preparation and customer service rather than eliminated, though that claim has not been independently verified across a diverse franchisee network.
McDonald’s projects approximately 250 basis points of gross restaurant-level efficiency gains from its broader modernization program. That translates to roughly $100,000 in additional annual cash flow per average U.S. restaurant, according to the company’s investor presentation. These are projected targets, not realized results.
McDonald’s is also providing franchisee support to offset modernization costs, but the program still carries an expectation of measurable returns, not just a more polished ordering screen.
The Real Test Comes in 2027
A broader U.S. rollout is planned for 2027, and that is when Archy’s real performance will be measured.
A broader U.S. rollout is expected in 2027, though the primary materials support a planned technology deployment rather than a confirmed full launch at every location. What remains unproven is whether Archy can hold its accuracy numbers at scale. Thousands of locations, varied regional accents, noisy environments, and the full range of menu modifications customers actually request are the real gauntlet. McDonald’s has introduced a new system with a broader operational scope than its earlier IBM pilot. Now it has to perform in the wild, not just in a test lane.
