One task splits into two jobs
The share of US fast-food locations using voice AI in drive-thru lanes was roughly 4 percent in 2024, rose to 5 percent last year, and early findings now put it at 6 percent. Taco Bell says it has reached 890 lanes, while White Castle reports the system called Julia taking orders at 12 percent of its restaurants and opening in every new location. Those figures describe a deployment large enough to affect many staffing schedules rather than a small demonstration. Deployment counts alone, however, do not say whose task changed or by how much.[1]
Intouch Insight finds AI lanes 21 seconds faster than the industry average, while order accuracy and friendliness scores are lower. When the system struggles with some accents and dialects, chains keep a worker on a headset ready to intervene. Order-taking is therefore not disappearing. The machine handles regular exchanges, while a person takes responsibility for complicated cases and customer recovery. The company can benefit from a shorter queue as the worker absorbs rarer but harder and potentially more tense exceptions. That is a change in the composition of work more than the disappearance of the work itself.[1]
Speed is measured; human handoffs are not
Earlier trials show that this task split does not work automatically. McDonald's sold its technology unit to IBM and abandoned the system, while Del Taco rolled back its own test in 2024; McDonald's is now running a new English-and-Spanish trial with Google at five locations. Dairy Queen, White Castle, and Taco Bell, meanwhile, are expanding. The divergent paths of similar technology across chains suggest that the 21-second difference cannot be assigned to software alone. Store selection, menu complexity, worker training, and the frequency of human handoffs can all change the outcome.[1]
To complete the distribution account, chains should place labour hours per lane, intervention rates, corrections and refunds, and schedule changes beside wait time. If interventions fall, accuracy holds, and workers do less repetitive work for the same pay, part of the gain can reach labour. If handoffs rise and fewer people carry more exceptions in a shift, the price of those 21 seconds is booked as work intensity. The answer to whose pocket receives the efficiency is hidden here not in sales totals, but in the unreported minutes spent on the headset.[1]