Inside QSR
McDonald's killed its AI drive-thru after two years. White Castle is expanding theirs. The difference isn't the technology — it's which lane operators put it in.
By Justin K. Sellers · 12 min read · March 3, 2026
McDonald's ended its AI drive-thru partnership with IBM in June 2024 after two years of testing at over 100 locations. The technology struggled with accents, picked up adjacent lane orders, and achieved accuracy rates of 80-85% — the same as human order-takers.
Meanwhile, White Castle is expanding SoundHound's AI voice ordering to 100+ drive-thrus. Wendy's FreshAI system reportedly achieves high accuracy when staff provide backup. Taco Bell is scaling from 5 to 30 locations based on positive feedback.
Same technology. Different results.
The question isn't whether AI works in QSR. It's which lane AI should stay in.
SoundHound claims its AI can complete over 90% of drive-thru orders without human intervention. The typical accuracy rate for human order-takers is 80-85%.
White Castle reported a 20% drop in order mistakes and 90-second average service times using AI. Wendy's FreshAI achieves over 85% hands-free order completion rates.
That sounds promising.
But McDonald's stopped its IBM automated order-taking (AOT) test in July 2024. According to CNBC, sources familiar with the technology said it had issues interpreting different accents and dialects, which affected order accuracy. McDonald's has shifted strategic direction on AI before — ending its IBM partnership after two years of testing.
The system added unwanted items to orders, mixed up orders from adjacent lanes, and ignored customer corrections, according to social media posts documented by the AI Incident Database.
BTIG analyst Peter Saleh wrote in spring 2024 that franchisees reported accuracy "remains in the low-to-mid 80% range and operating costs are high."
The AI performed the same as humans, but cost more and frustrated customers who expected better from technology.
The operators reporting measurable AI results aren't deploying it at drive-thrus. They're using it for inventory forecasting, labor scheduling, and kitchen automation — repetitive tasks that don't require human judgment.
Chipotle uses AI to monitor ingredient levels in real-time and predict demand based on sales history, weather patterns, local events, and time-of-day data, according to QSR Magazine. The system alerts employees how much to prep, cook, and when to start cooking.
AI forecasting systems can reduce food waste by approximately 15% when used properly, according to vendor reporting from Techryde. Industry vendors claim investments in waste reduction can yield a 7:1 benefit-cost ratio, meaning every dollar spent generates up to seven dollars in benefit, according to Fourth.
AI-powered scheduling systems analyze historical data and predict busy periods to generate optimal schedules that match labor to demand. Companies using AI-based scheduling have reduced labor costs by 10-15% and automated up to 50% of their forecasting and staffing processes, according to NOWSTA.
When operators can't staff consistently, no scheduling algorithm fixes the root problem. Losing a single GM costs $70,000 — AI scheduling helps, but retention is the lever that actually moves the needle.
White Castle's "Flippy 2" robot handles basket filling, cooking, and transferring to hot-hold areas in a closed-loop system. Throughput increased 30% to roughly 60 baskets per hour, according to QSR Magazine.
Chipotle tested "Chippy," an AI robot from Miso Robotics that makes tortilla chips using the exact recipe. Chipotle's "Autocado" robot cuts, cores, and scoops avocados, saving an average of 50 minutes per batch of guacamole. With Chipotle using over 5 million cases of avocados per year, the time savings compound across locations.
The pattern: AI handles the repetitive. Humans handle the exceptional.
Drive-thru AI voice ordering gets media attention because it's customer-facing. But it's also the hardest problem to solve.
Successful implementations require real-time language processing in noisy environments, understanding of accents and dialects, handling complex modifications, managing impatient customers, and knowing when to escalate to humans.
But here's what the successful deployments have in common: human backup.
Every system that works has staff ready to step in when AI gets confused. The customer shouldn't even know it happened.
This isn't full automation. It's augmentation.
Presto Automation was charged by the SEC for misleading statements about the capabilities of its AI product. The company used "human agents" in places like the Philippines and India to complete orders that its AI could not handle, according to CNBC reporting. Presto said in a statement that "using humans is common in the AI industry" and helps train the technology. Presto unveiled a "fully autonomous" version in May 2024.
The distinction between AI-powered ordering and human-assisted ordering isn't always clear to the customer.
You've seen this movie before. Customer service calls used to go to local reps who knew your name and could actually fix your problem.
Then companies moved to offshore call centers to cut labor costs.
The P&L looked great. The customer experience collapsed.
Why?
You can script the predictable. You can't script the exceptional.
When the problem doesn't fit the flowchart, it needs someone with judgment and authority. Call centers optimized for routine calls. They failed the exceptions.
Drive-thru AI is making a similar bet: that most orders are predictable enough to automate.
The ones that aren't? Those define the customer experience.
The operators watching regional traffic data already know customers are voting with their feet. AI that degrades the experience accelerates that trend.
The operators winning with AI follow a pattern:
Automate the repetitive. Augment human judgment. Make staff more effective.
Tasks that benefit from AI automation:
- Inventory tracking and reorder alerts - Labor schedule generation - Production planning and prep forecasting - Waste pattern monitoring - Equipment maintenance predictions
These tasks don't require creativity. They require accuracy and consistency at scale.
Tools that should give humans better information:
- Kitchen display systems that prioritize orders by complexity - Real-time alerts for variance in food cost or labor - Exception reporting for unusual order patterns - Upsell suggestions at kiosks (customer-initiated, not forced)
The human stays in control. AI provides the data.
Before you sign a contract:
1. Does this make service faster AND better?If the answer is "faster, but we'll work on accuracy later," walk away. Speed without accuracy moves the problem downstream.
2. Does this free my team for higher-value work?If AI just replaces headcount, you're not innovating. You're cutting costs.
3. What happens when AI fails?Every system fails. Can a human step in seamlessly? If customers have to wait for system reboots, you're creating a worse experience.
4. Are you solving operations or cutting labor costs?Be honest. Implementing AI because you can't find staff is one thing. Implementing it to reduce payroll while service suffers is another. Customers see the difference.
- McDonald's IBM exit details: What specific operational challenges led McDonald's to end its IBM partnership — and whether the accuracy issues were hardware, software, or implementation problems. McDonald's did not publicly disclose the root cause. - AI system failure rates: How often human backup is required in "successful" AI deployments — customer satisfaction scores comparing AI ordering to human ordering are not publicly available from any major deploying brand. - Automation ROI data: ROI timelines for kitchen automation investments like Chippy, Flippy, and Autocado — and whether labor savings reduce headcount or reallocate staff to higher-value tasks. No deploying brand has published payback period data. - AI adoption landscape: What percentage of QSR brands are actively testing AI vs. watching from the sidelines — and how many pilots fail before reaching full deployment. Industry estimates vary widely and are not independently verified. - PE and AI acceleration: Whether PE firms acquiring QSR brands will accelerate AI adoption — the $18.6 billion acquisition wave suggests investors want proven unit economics, not tech experiments, but PE pressure to cut labor costs could push adoption ahead of readiness.
AI isn't the enemy. Bad implementation is.
The operators getting results — the ones cited in this research — use AI for the tasks nobody wanted to do: inventory counts, staffing calculations, waste monitoring.
That's not taking jobs. That's removing the repetitive tasks that cause burnout.
When Chipotle tested Chippy, the goal wasn't to fire the chip-maker. It was to let that person interact with customers and handle complex orders — work that requires a human.
When White Castle deployed Flippy 2, throughput increased 30%, according to QSR Magazine. The robot handled the frying. The staff moved to other stations.
That's AI in the right lane.
The call center analogy holds here too. When companies replaced local staff with offshore labor to cut costs, customer satisfaction declined. QSR risks the same pattern.
QSR can avoid that mistake.
Use AI where it automates the predictable, augments human judgment, and makes staff more effective. Don't use it to replace the human connection that makes customers come back. The brands that invest in operators rather than replacing them — like the Chick-fil-A model that generates $200K+ in cash flow with zero capital risk — understand that systems work when humans run them, not the other way around.
The restaurant AI market is projected to grow from $129.2 million in 2024 to $1.32 billion by 2030, according to Precedence Research. Industry projections suggest 50% of U.S. drive-thru orders could be AI-handled by late 2026 — though this forecast seems aggressive given McDonald's recent exit and the ongoing implementation challenges across the industry.
The technology is coming. The question is which lane it stays in.
This analysis cites multiple independent industry sources to provide comprehensive operator-focused research. We reference publicly available data with full attribution and direct links to support our independent analysis.
QSR Research Hub is an independent publication. We are not affiliated with any brand, corporation, or entity discussed in this article and receive no compensation for citations or analysis.
We publish deep dives with real citations, real data, and zero corporate spin. Franchise economics, brand analysis, and operator strategy — no fluff, no investor pitch, no vendor influence.
And subscribe free — get it delivered to your inbox. Subscribe to QSR Research Hub1. Restaurant Business. "McDonald's is ending its drive-thru AI test." June 14, 2024. https://www.restaurantbusinessonline.com/technology/mcdonalds-ending-its-drive-thru-ai-test
2. CNBC. "McDonald's to end AI drive-thru test with IBM." June 18, 2024. https://www.cnbc.com/2024/06/17/mcdonalds-to-end-ibm-ai-drive-thru-test.html
3. Restaurant Dive. "White Castle to roll out voice AI to over 100 drive-thrus." August 2, 2023. https://www.restaurantdive.com/news/white-castle-soundhound-ai-voice-drive-thru-100-units-2024/689624/
4. Restaurant Dive. "Wendy's franchisees can pilot drive-thru AI in 2024." December 11, 2023. https://www.restaurantdive.com/news/wendys-expand-google-generative-ai-drive-thru-test/702184/
5. Canopy Remote Device Management Software. "11 Fast Food Chains Using AI Drive-Thrus (Problems & Issues)." https://www.gocanopy.com/news-insights/ai-drive-thru-problems
6. CNBC. "AI drive-thru ordering is on the rise — but it may take years to iron out its flaws." July 3, 2024. https://www.cnbc.com/2024/07/03/ai-drive-thru-ordering-mcdonalds-yum-wendys-test-tech.html
7. WebProNews. "AI Drive-Thru Revolution: Fast Food's Bold Bet on Voice Ordering." January 14, 2026. https://www.webpronews.com/ai-drive-thru-revolution-fast-foods-bold-bet-on-voice-ordering/
8. AI Incident Database. "Incident 475: McDonald's Reportedly Ends IBM Partnership After AI Drive-Thru Ordering Errors at U.S. Locations." https://incidentdatabase.ai/cite/475
9. Generation Digital. "AI Boosts QSR Operations: Streamline and Save Costs." https://www.gend.co/blog/ai-qsr-operations
10. Fourth. "AI in Restaurants: 25 Tools for 2025." December 15, 2025. https://www.fourth.com/article/ai-in-restaurants
11. QSR Magazine. "AI and Machine Learning Enter the Kitchen at Chipotle." April 8, 2025. https://www.qsrmagazine.com/growth/fast-casual/ai-and-machine-learning-enter-kitchen-chipotle/
12. Techryde. "Top 7 Operational Best Practices for QSRs in 2026." https://www.techryde.com/blog/top-7-operational-best-practices-for-qsrs-2026/
13. NOWSTA. "How AI Is Changing Labor Forecasting and Scheduling." https://nowsta.com/blog/how-ai-is-changing-labor-forecasting-and-scheduling/
14. Chipotle. "CHIPOTLE TESTS AI KITCHEN ASSISTANT, CHIPPY." March 16, 2022. https://newsroom.chipotle.com/2022-03-16-CHIPOTLE-TESTS-AI-KITCHEN-ASSISTANT,-CHIPPY
15. Canopy Remote Device Management Software. "Chippy, Autocado & Chipotlanes: How Chipotle Uses Robotics." https://www.gocanopy.com/news-insights/chipotle-chippy-chipotlanes-autocado
16. SEC.gov. "SEC Charges Restaurant-Technology Company Presto Automation for Misleading Statements About AI Product." January 14, 2025. https://www.sec.gov/enforcement-litigation/administrative-proceedings/33-11352-s