Last Updated on July 16, 2026 by Craig Allen Keefner
Taco Bell’s Voice AI Drive-Thru: What Guests Really Want
Post on HT Hospitality Technology — We haven’t seen any of these in Colorado yet. We get a lots of road noise, low-quality speaker, a low-quality microphone, and a non-native English speaker.
What’s happening
-
Taco Bell is expanding its Voice AI drive-thru rollout via Omilia, now in 890+ U.S. locations.
-
Other QSRs are doing the same:
-
Wendy’s with Google “FreshAI”
-
Bojangles’ “Bo-Linda” in 450+ locations, claiming ~96% order accuracy
-
-
Goal: reduce labor pressure, free staff for food prep, maintain or improve throughput.
What operators see (the “pro” case)
-
Comparable transaction times vs. humans.
-
Labor efficiency gains (fewer staff tied up on headsets).
-
Scalable automation across large store networks.
-
Internal data suggests pilots are “successful enough” to expand.
What customers actually think
-
Strong preference against AI voice ordering:
-
14% prefer AI voice vs. 34% human, 31% mobile app (single-choice).
-
Only 5% prefer voice AI when multiple options are allowed.
-
-
Preferred channels:
-
Counter: 45%
-
Mobile app: 43%
-
Core friction points
-
Trust: fear AI will misinterpret orders or fail on customization.
-
Control: complaints about rigid, scripted interactions.
-
Experience mismatch: drive-thru is expected to be fast and human—AI doesn’t clearly improve speed.
-
Privacy concerns: unease about voice data and tracking.
-
Emotional reaction: ~30% feel uncomfortable using it (vs. ~45% comfortable).
Key insight (the disconnect)
-
Operators optimize for cost and efficiency.
-
Customers optimize for control, clarity, and familiarity.
-
Voice AI currently doesn’t deliver a clear customer-side benefit—just a backend one.
What it means (practically)
-
This is a classic “adoption curve vs. UX gap” situation:
-
Brands will keep deploying due to ROI.
-
Customer resistance is real and could impact repeat visits if UX is poor.
-
-
The winning model likely isn’t “AI only,” but:
-
AI + instant human fallback
-
Better handling of modifiers and natural language
-
Clearer value to the customer (speed, accuracy, personalization)
-
Quick example
A customer ordering a customized burrito:
-
Human: handles “no sour cream, add jalapeños, extra rice” naturally.
-
Current AI: may misinterpret, restrict changes, or rush the flow → frustration → lost loyalty.
- Not many video examples but here is one from Irvine California in 2025
- https://youtu.be/1Z0wZ7k-gwc?si=KZ5sc0CQebhskQJa
If you’re looking at this from a kiosk/automation lens, it mirrors early self-checkout: adoption happens, but only when UX friction drops below a tolerance threshold.
If you want, I can map this directly to kiosk UX lessons or ADA/accessibility implications—there are some strong parallels.
Evidence for the operational gains
The evidence is moderate on operational performance and weaker on broad consumer acceptance. The strongest support comes from real-world field studies and brand-reported pilot data, but the consumer-dislike conclusion rests on a survey plus a mystery-shopper study, not on long-term behavior or sales causality.
-
Taco Bell/Omilia claims the system is deployed in 890+ U.S. stores, and the company says it has seen on-par or better transaction times, higher retention at some locations, and improved workload for staff.
-
Intouch Insight’s mystery-shopping study found Taco Bell’s Voice AI cut total drive-thru time by 1 minute 54 seconds versus a 2024 benchmark, with 100% clarity reported at Taco Bell and Bojangles in that study .
-
That same study reported suggestive selling in 81% of Voice AI visits, which is a concrete operational metric, though it is still vendor-adjacent research rather than an independent randomized trial .
Evidence for guest dislike
-
The HospitalityTech article you shared cites a 1,004-person survey in which only 14% picked AI voice ordering as their top drive-thru method, while 34% preferred a human and 31% preferred the mobile app.
-
It also reports that 30.4% felt uncomfortable with an AI voice assistant, versus 45.4% comfortable, which supports the “many guests don’t like it” conclusion.
-
Intouch Insight’s 2025 consumer survey found 45% disliked voice AI and only 19% had even tried it, which suggests limited exposure and mixed sentiment rather than settled rejection.
How strong the conclusion is
-
The evidence is good for saying customers are not uniformly enthusiastic and that there is a real preference gap versus human or app-based ordering.
-
The evidence is weaker for saying “guests hate it” in a broad, definitive sense, because the data are cross-sectional, self-reported, and drawn from specific studies rather than a representative long-term market test.
-
It is also not enough to prove that voice AI reduces repeat visits systemwide; one quoted anecdote and some social-media complaints show frustration, but anecdotes are not generalizable evidence.
Best read of the article
-
The article’s main conclusion is supported: operators see ROI, guests show resistance, and there is a real experience gap.
-
The article goes a bit further than the evidence fully warrants when it implies a broad consumer backlash; “mixed feelings” or “significant resistance” is better supported than “hate”.
-
So the evidence base is solid for a pilot-to-scale business case, but only fair-to-moderate for a consumer-preference verdict.
Practical takeaways
-
Being tactful — Voice AI is expanding because operators see efficiency gains, but consumer acceptance remains limited and uneven.
-
If HT wants a stronger claim, we’d need longitudinal data on repeat visits, complaint rates, conversion, and order accuracy across matched stores over time—not just surveys and pilot readouts.
- Omilia is an enterprise‑grade conversational engine now powering Taco Bell’s Voice AI drive‑thru; SoundHound and Sodaclick compete in the broader voice AI and restaurant tech space, but with different strengths—SoundHound as a multi‑vertical platform and Sodaclick as a multimodal signage/kiosk player.
