Our Take
The AI drive-thru was never solving a customer problem, only a labor cost problem, and the SEC filing against Presto exposed the gap between marketing and reality: human workers in the Philippines were handling most orders.
Why it matters
QSR chains are discovering that deploying AI to replace voice interaction without consent or accuracy guarantees triggers both customer rejection and regulatory scrutiny. The broader lesson affects any sector betting on friction reduction through automation.
Do this week
Product teams: audit your AI rollout communication against what your system actually does in production before launch, not after an SEC investigation forces disclosure.
McDonald's and Taco Bell reverse course on AI ordering
Starting in 2021, major fast-food chains began deploying AI chatbots at drive-thru windows. McDonald's piloted the technology at 10 Chicago locations after acquiring Apprente (a voice tech startup) in 2019, then scaled with IBM. Checkers and Rally's moved to Presto-powered systems across all U.S. corporate drive-thrus in 2022. Wendy's launched "FreshAI" in 2023, trained on franchise terminology so it recognized "Frosty" as milkshake and "JBC" as junior bacon cheeseburger. Wendy's reported 86 percent accuracy without employee intervention (company-reported). Taco Bell, Panera Bread, White Castle, Carl's Jr., Hardee's, Panda Express, and Popeyes followed suit.
By early 2025, the momentum reversed. McDonald's ended its IBM partnership in 2024. A YouGov survey in January 2025 found 55 percent of Americans prefer human order-takers, versus 21 percent with no preference and only 4 percent favoring AI. Taco Bell's chief digital officer told The Wall Street Journal the chain is reevaluating its AI drive-thru deployment after customer complaints on social media and deliberate sabotage (customers ordered 18,000 water cups to jam the system).
The credibility blow came from the SEC. Last year, the agency charged Presto with misleading customers about its technology's capabilities. An SEC filing revealed that human workers in the Philippines were staffing most Presto orders, not the AI system itself.
Automation theater meets regulatory reality
Fast-food chains framed AI drive-thrus as solving three problems: labor cost reduction, order accuracy, and reduced wait times. None of these framing matched customer expectations or actual capability. The SEC filing against Presto exposed the third problem: the systems were not autonomous at all.
The chains are now exploring AI applications lower in the stack where the human handoff is invisible or irrelevant. McDonald's is piloting predictive maintenance for ice cream machines and AI-powered scales to catch missing items in to-go bags. Burger King tested "Patty," an AI assistant running inside employee headsets that helps workers recall recipes and evaluates their customer-service tone. Taco Bell and Burger King are testing dynamic menu boards that adjust content per customer, and Culver's and Zaxby's are deploying camera timers to measure traffic flow and service execution (Berry AI claims 20 to 40 percent service time reduction, company-reported).
The pattern is clear: chains are moving AI from the visible customer interface to back-office and internal operations. When the technology's limitations don't trigger customer frustration or SEC scrutiny, deployment continues.
What to do now
If you are building or deploying AI systems with human-in-the-loop fallbacks, require your team to disclose the percentage of requests that actually flow to the AI path in production before you launch. Don't wait for the SEC to force it. Wendy's disclosed its 86 percent accuracy and saw continued expansion; Presto did not, and faced enforcement action.
If you are evaluating vendor claims about AI capability reductions (cost per order, accuracy %, time saved), demand independent benchmarks or third-party validation. Vendor-only numbers without reproducers should be treated as marketing claims, not performance guarantees.