Taco Bell × Omilia — voice AI deployed at 890+ US drive-thrus across 38 states (July 2026)

NA-USA

Content

Taco Bell × Omilia voice AI is the largest deployed voice-AI drive-thru footprint in the United States. As of July 2026:

• 890+ US Taco Bell drive-thru locations across 38 states are running Omilia voice AI for automated order taking.
• The partnership with Omilia dates to 2023. Taco Bell × Yum also tried a parallel deployment with NVIDIA voice-AI frameworks in March 2025 (~500 restaurants) — Yum slowed that deployment in August 2025 after public trolling incidents (orders for ~18,000 cups of water etc.) and customer complaints, and consolidated on the Omilia stack.
• Yum! Brands launched Byte by Yum! in 2024 — a proprietary SaaS platform integrating AI, POS, and online ordering across Taco Bell, KFC, and Pizza Hut; the voice-AI surface runs on this stack.
• The unit represents the largest deployed AI ordering surface in US QSR (as of mid-2026).

What this unit is doing in the taxonomy

The load-bearing deployed-at-scale example for generative AI in US restaurant ordering. Counter-anchor to the failed/discontinued McDonald’s × IBM unit. Together they form the public case study of voice-AI drive-thru: one chain scaled successfully, another pulled the technology.

Why it matters for talks

• Concrete deployment numbers (890+ units, 38 states) are unusually strong — verified across Restaurant Dive (Jul 7 2026) and Businesswire (Jul 7 2026).
• The Yum × NVIDIA → Omilia pivot is a teachable moment on production reliability vs technology stack novelty: the more novel NVIDIA deployment failed in market, the more conservative Omilia stack scaled.
• Public trolling incidents (18,000 cups of water) demonstrate the consumer-vs-AI surface that vendor presentations systematically underplay.

Critical context the unit does not sanitise

• The AI ordering surface is a consumer-facing product, but Taco Bell did not collect public consent for the underlying voice recording (drive-thru audio is captured for AI training and quality control). No publicly visible consumer data-governance instrument beyond Yum’s general privacy notice.
• Public trolling incidents are operational failures but also fun consumer behaviour — the surface has no clear escalation path beyond crew intervention.
• The same vendor deployment (Omilia) is being marketed to other QSR chains — the consumer experience is becoming standardised in ways consumers cannot see.