Taco Bell finally put numbers on its drive-thru AI. Yum hasn't said what it compared them against.

Previously

Taco Bell's drive-thru voice AI went from 100+ stores in July 2024 to 500+ in August 2025, after viral glitches including an order for 18,000 waters. In October 2026, with the system in 890+ stores and supplied by Omilia, Yum said it helped keep workers but published no figures. A small January 2025 mystery-shop study found staff stepping in on 30% of Taco Bell's AI orders.

Nov. 3, 2026 — Yum! Brands put its first outcome numbers on Taco Bell's drive-thru voice AI. Here's what changed since Part 3 (October 2026), what it cost, and what went wrong.

Cost and outcome (as of Nov. 3, 2026)

Footprint: "more than 900 Taco Bell restaurants now use voice AI in the drive-thru," with "more than 1,500 restaurants by year-end" expected, CFO Ranjith Roy said. (Yum Q3 call transcript)
Satisfaction: "a 6% improvement in overall drive-thru satisfaction" in voice AI stores. (Yum Q3 call transcript)
Sales: "approximately 1.5% sales uplift driven by higher check." (Yum Q3 call transcript)
Staff: "sustained improvements in team member retention compared with our control group." No turnover figures. (Yum Q3 call transcript)
Cost: Not disclosed.

What Yum said

In prepared remarks, Roy said that in locations where voice AI is deployed, the brand has seen better drive-thru satisfaction, a sales lift from higher checks, and "sustained improvements in team member retention compared with our control group." As spoken, the control group is attached to the retention claim; he didn't say what the satisfaction and sales figures were measured against. (Yum Q3 call transcript; CIO Dive)

An analyst from Melius Research picked up on it: "You gave some useful data on like the uplifts from voice AI and the better employee retention." Roy answered: "So on the voice AI front, we're seeing uplift right now," adding that Yum is "focused on executing against the 1,500 deployments in Q4 and getting those right." (Yum Q3 call transcript)

The quarter around it

The figures came in a hard quarter for Taco Bell. CEO Chris Turner said U.S. same-store sales fell 4% "owing to the industrywide food safety issue." (Yum Q3 call transcript) The 1.5% uplift describes voice AI stores, not Taco Bell's overall sales, which were down. Yum didn't say whether it is measured against the same stores before AI or against stores without it. (MarketBeat)

What's changed

Part 3 ended with "hasn't shown the numbers." Now there are three, from the operator rather than the vendor, plus a target that would add about 600 stores in roughly two months. What's still missing is the method behind them.

What went wrong

No method. Yum didn't say what the 6% and 1.5% are measured against, how the retention control group was picked, how many stores or weeks it covers, or how "overall drive-thru satisfaction" is scored.
Uplift from bigger checks. Yum attributed the 1.5% lift to "higher check." It didn't say whether that comes from the AI's upselling or something else.
Retention without numbers. "Sustained improvements" in retention, but still no turnover rates for AI and non-AI stores.
Humans in the loop. Yum gave no figure for how often staff step in. The only outside number we have is still the 30% from a 40-visit study in January 2025 (Part 3).
A fast push. Going from 900+ to 1,500+ stores by year-end is the steepest expansion in this series. Roy called getting those deployments right the Q4 focus.
What we don't know yet: the control-group design, intervention rates at scale, and what franchisees pay.

Our follow-up question for Yum! Brands

What are the 6% satisfaction and 1.5% sales figures measured against, how is the retention control group built, and what is the staff intervention rate across the 900+ stores?

Sources

The Motley Fool, “Yum Brands (YUM) Q3 2026 Earnings Call Transcript,” Nov. 3, 2026. fool.com
CIO Dive, “Yum Brands flags AI as 'transformative to the business,'” Nov. 2026. ciodive.com
MarketBeat, “Yum! Brands Q3 Earnings Call Highlights,” Nov. 4, 2026. marketbeat.com

Disclosure: [not in copy] Editor placeholder