Voice AI handled the entire order in 40% of drive-thru AI encounters in 2026, double the 20% measured a year earlier (QSR Magazine, 2026 QSR Drive-Thru Report). Taco Bell now runs voice ordering in 890+ U.S. restaurants across 38 states (Omilia press release, July 2026), and order accuracy at AI-equipped locations climbed to 90.8% from 82.5% (Intouch Insight, 2026 Drive-Thru Study). The awkward counterweight sits in an SEC filing: at one vendor’s most advanced sites, only about 46% of orders needed no off-site human agent (Presto Automation, Form 10-Q for March 31, 2024). We aggregated data from the Intouch Insight Drive-Thru Study, SEC filings from Presto Automation, SoundHound AI, Yum! Brands and McDonald’s, company statements from Taco Bell, Wendy’s and White Castle, and the National Restaurant Association. The gap between pilot claims and audited completion mirrors what we documented for voice AI in contact centers, only with a car engine running in the background.
TL;DR
- AI handled the full order in 40% of voice AI drive-thru encounters in 2026, up from 20% in 2025 (QSR Magazine, 2026 QSR Drive-Thru Report).
- Taco Bell voice AI is live in 890+ U.S. restaurants across 38 states, up from 100+ in 13 states in July 2024 (Omilia press release 2026; Yum! Brands press release 2024).
- Order accuracy at voice AI locations reached 90.8%, up from 82.5% in 2025 and above the 89.3% all-lane benchmark (Intouch Insight, 2026 Drive-Thru Study).
- Total time was 5:14 at AI locations versus 5:15 for the same brands with human order takers (Intouch Insight, 2026 Drive-Thru Study).
- At Presto locations running its most advanced AI version, approximately 46% of orders needed no human agent to enter them (Presto Automation, Form 10-Q, March 2024).
- The SEC found Presto made misleading statements about its drive-thru AI from November 2021 through May 2023 and imposed no civil penalty (SEC Administrative Proceeding File No. 3-22413, 2025).
- McDonald’s ended an IBM automated order taking test that had reached about 100 restaurants since October 2021 (McDonald’s statement to Nation’s Restaurant News, June 2024).
- Wendy’s FreshAI handled 86% of pilot orders without crew intervention and was in 160+ U.S. restaurants by May 2025 (Wendy’s blog 2023; Wendy’s Q1 2025 earnings call).
- 45% of consumers disliked the idea of voice AI ordering, and 54% of those 45 and older did (Intouch Insight consumer survey 2024).
- Shoppers rating the AI experience difficult rose from 7% to 19% in one year (QSR Magazine, 2026 QSR Drive-Thru Report).
- SoundHound AI reported $168.9 million in 2025 revenue, up 99%, and guides to $230 million to $260 million for 2026 (SoundHound AI Form 10-K 2025; Q2 2026 earnings release).
1. Chain Deployments: Where Voice AI Is Actually Taking Orders
One chain accounts for most of the publicly countable footprint. Taco Bell went from five California test stores in early 2024 to 890+ U.S. restaurants in 38 states by July 2026 (Omilia press release, July 2026), and every other brand that has disclosed a number is still in the low hundreds or declined to update it.
Scale also has to be read against the lane count. McDonald’s reports nearly 29,000 drive-thru locations globally, including over 95% of its approximately 13,700 U.S. restaurants (McDonald’s Form 10-K 2025), so the roughly 100 restaurants in its discontinued IBM test were a rounding error on its own system. Mystery shoppers found voice AI on 12.5% of visits to three AI-equipped brands in 2026, down from 14.1% in 2025 (QSR Magazine, 2026 QSR Drive-Thru Report): presence in the field slipped even as the headline store counts grew.
| Metric | Value | Source |
|---|---|---|
| Taco Bell U.S. restaurants with voice AI (July 2026) | 890+ in 38 states | Omilia press release, July 7, 2026 |
| Taco Bell voice AI footprint (July 2024) | 100+ drive-thrus in 13 states | Yum! Brands press release, July 31, 2024, reported by Restaurant Dive |
| Taco Bell voice AI orders processed (Q3 2024) | Over 2 million, in 300+ U.S. stores | Yum! Brands Q3 2024 earnings call (CFO Chris Turner), reported by Sixteen:Nine |
| Yum! restaurants targeted for NVIDIA-powered AI in 2025 | 500 | NVIDIA blog on the Yum! Brands partnership, March 2025 |
| Wendy’s FreshAI restaurants (May 2025) | 160+ U.S. restaurants; target 500+ by year-end 2025 | Wendy’s Q1 2025 earnings call, reported by Restaurant Dive |
| McDonald’s IBM automated order taking test | About 100 restaurants, from October 2021 | McDonald’s statement to Nation’s Restaurant News, June 2024 |
| White Castle planned AI lanes | 100 of 350 locations by end of 2024 | White Castle announcement, reported by Restaurant Business 2023 |
| Mystery-shop visits where voice AI was found (2026) | 12.5% combined: Taco Bell 21%, Wendy’s 13%, Burger King about 3% | QSR Magazine, 2026 QSR Drive-Thru Report |
Context: the White Castle figure is the most recent available data (announcement of August 2023) and was a plan, not a confirmed count. The 890+ figure comes from the vendor’s release, and Taco Bell’s own executive quote in it describes the platform as proven “in select U.S. restaurants”. After an August 2025 incident in which a customer tried to order 18,000 cups of water, Taco Bell’s technology chief told the Wall Street Journal the chain was rethinking where to deploy the system, according to Nation’s Restaurant News. The primary count is in the Omilia announcement.
2. Completion Rates: How Often the AI Finishes the Order Alone
Being installed and taking the order are different things. In half of all voice AI encounters in 2026, the system greeted the customer and immediately handed off to an employee (QSR Magazine, 2026 QSR Drive-Thru Report). That is better than the 70% handoff rate of 2025, and it still means the typical “AI lane” is a human lane with an automated hello.
The spread between cuts is the real finding. At restaurants built around AI ordering, 81.1% of orders ran end to end; across every encounter shoppers logged, 40% did (QSR Magazine, 2026 QSR Drive-Thru Report). On the 120 visits that Intouch Insight aimed at AI-equipped restaurants, the AI took some or all of the order on 92 and completed 77 from start to finish (Intouch Insight, 2026 Drive-Thru Study). Vendor and chain claims cluster between 86% and 93%+, and none of them use the same denominator.
| Metric | Value | Source |
|---|---|---|
| Encounters where AI handled the full order (2026 vs 2025) | 40% vs 20% | QSR Magazine, 2026 QSR Drive-Thru Report |
| Encounters limited to an AI greeting, then employee handoff (2026 vs 2025) | 50% vs 70% | QSR Magazine, 2026 QSR Drive-Thru Report |
| AI-led orders that still required an employee (2025) | 22% | Intouch Insight, Emerging Experiences 2025 |
| Wendy’s FreshAI pilot orders handled without crew intervention | 86% average | Wendy’s blog, December 2023 |
| Wendy’s FreshAI pilot success rate when orders fixed by a human are counted | Nearly 99% | Wendy’s, reported by Restaurant Dive, December 2023 |
| White Castle initial tests, order completion rate (vendor claim) | 90% | SoundHound AI, reported by Restaurant Business 2023 |
| Orders SoundHound says its AI can take without human intervention (vendor claim) | More than 90% | SoundHound AI, reported by CNBC, July 2024 |
| Hi Auto order completion “at scale” (vendor claim) | 93%+ | Hi Auto company website 2026 |
Outlier note: the two Wendy’s rows describe the same pilot under two definitions, roughly 13 points apart. Read every completion figure with its definition attached. The field data is summarized in Intouch Insight’s 2026 study overview.
3. Off-Site Human Agents: What the Presto Filings Disclosed
The clearest look inside a drive-thru AI stack came from a company that has since left the public markets. Presto Automation told investors in its fiscal 2023 annual report that Presto Voice achieved an 85% non-intervention rate on average. The same filing explained that the rate referred to restaurant staff, while off-site human agents would “enter, review, validate and correct orders.”
By March 31, 2024, Presto disclosed that only approximately 46% of orders at its most advanced locations did not require human agents to enter them, and that locations on its then-current technology used human agent intervention “in all instances” (Presto Automation, Form 10-Q, March 2024). The work had not disappeared. It had moved to a remote headset, which makes drive-thru automation partly a story about outsourced voice labor, the same workforce covered in our call center turnover and burnout statistics.
| Metric | Value | Source |
|---|---|---|
| Presto Voice non-intervention rate by restaurant staff (fiscal 2023) | 85% on average; approximately 95% at certain locations | Presto Automation Form 10-K, fiscal 2023 |
| Contractors, mostly off-site human agents (June 30, 2023) | 149, primarily in the Asia-Pacific region | Presto Automation Form 10-K, fiscal 2023 |
| Orders needing no human agent entry at most advanced locations (March 31, 2024) | Approximately 46% | Presto Automation Form 10-Q, March 2024 |
| Presto Voice locations installed, live, and on the most advanced AI | 202 installed, 163 live, 55 most advanced | Presto Automation Form 10-Q, March 2024 |
| Checkers locations powered by Hi Auto’s solution (excluded from the row above) | 347 | Presto Automation Form 10-Q, March 2024 |
| Workforce reductions | 17% on November 15, 2023; 24 corporate roles, or 18%, from March 14, 2024 | Presto Automation Form 10-Q, March 2024 |
| Cash and cash equivalents | $4.2 million at March 31, 2024 vs $15.1 million at June 30, 2023 | Presto Automation Form 10-Q, March 2024 |
| SEC order: period of misleading statements and civil penalty | November 2021 through May 2023; no penalty | SEC Administrative Proceeding File No. 3-22413, January 2025 |
Context: the SEC found that the speech recognition in all Presto Voice units deployed for a period was owned and operated by a third party, and that under Presto’s own later version “the vast majority of drive-thru orders” required human intervention. Presto consented to a cease-and-desist order without admitting or denying the findings (SEC administrative proceeding summary). The fiscal 2023 figures are the most recent available data for that metric. The March 2024 report is the last Form 10-Q Presto filed, and in a Form 8-K filed in September 2024 the company said its shares were delisted from Nasdaq on August 8, 2024 and that it intended to terminate its SEC registration. The 46% disclosure is in the Form 10-Q for March 31, 2024.
4. Order Accuracy and Service Time: AI Lanes vs Human Lanes
Accuracy is where the technology changed most in twelve months. Order accuracy at voice AI locations rose 8.3 points, from 82.5% in 2025 to 90.8% in 2026 (Intouch Insight, 2026 Drive-Thru Study; the difference is 90.8 minus 82.5). McDonald’s reportedly wanted 95%-plus before a broad rollout, and none of the mystery-shop measurements cited here puts an AI lane at that level.
Speed is a wash on average and a win only when the handoff never happens. AI locations finished in 5:14 against 5:15 for the same brands with people at the speaker. Orders the AI carried start to finish took 4:19 against 5:25 at non-AI drive-thrus, while encounters where the AI only greeted the customer took 5:16 (QSR Magazine, 2026 QSR Drive-Thru Report). The time savings live in the 40% of encounters described in section 2, and every handoff dilutes them.
| Metric | Value | Source |
|---|---|---|
| Order accuracy at voice AI locations (2026 vs 2025) | 90.8% vs 82.5% | Intouch Insight, 2026 Drive-Thru Study |
| Total time, AI locations vs same brands in main study | 5:14 vs 5:15 | Intouch Insight, 2026 Drive-Thru Study |
| Total time when AI handled the full order vs non-AI drive-thrus | 4:19 vs 5:25 (66 seconds faster) | QSR Magazine, 2026 QSR Drive-Thru Report |
| Voice AI vs traditional drive-thru (2025) | About 12 seconds faster on average | Intouch Insight, Emerging Experiences 2025 |
| Friendliness rating, voice AI vs employee order taking (2025) | 83% vs 79% | Intouch Insight, Emerging Experiences 2025 |
| Wendy’s FreshAI pilot site vs Columbus market average | 22 seconds faster | Wendy’s, reported by Restaurant Dive, December 2023 |
| Human order accuracy and lane speed-up cited by SoundHound (vendor claims) | 80% to 85% typical human accuracy; lanes roughly 10% faster with AI | SoundHound AI, reported by CNBC, July 2024 |
| McDonald’s accuracy target vs reported IBM test accuracy | 95%-plus target vs low-80% range (BTIG analyst report 2022) | Nation’s Restaurant News, June 2024 |
Context: the low-80% figure for McDonald’s is a third-party analyst estimate from 2022, not a company number. McDonald’s own statement to Nation’s Restaurant News said only that, after a thoughtful review, it had decided to end its global partnership with IBM on automated order taking, and that it would decide on a future voice ordering solution by the end of 2024. CNBC reported in July 2024, citing two sources, that the technology struggled with accents and dialects, the failure mode measured in our speech recognition accent bias statistics. The upsell number explains the commercial interest better than the stopwatch does: suggestive selling happened on 79.2% of visits at AI locations, about 20 points above the same-brand comparison (Intouch Insight, 2026 Drive-Thru Study), and an AI that offers the add-on on nearly four of five orders does not get tired at 9 p.m.
5. The Human Baseline: Drive-Thru Benchmarks in 2026
Any AI claim needs a denominator, and the human lane just had its best year in a while. The average drive-thru visit took 5:24 in 2026, 11.1 seconds faster than 2025 and the fastest since 2018 (Intouch Insight, 2026 Drive-Thru Study, based on 2,145 mystery shops at 13 chains in June and July 2026).
The average hides a long tail. Half of visits finished in under 4:25 and the slowest 10% took 9:56 or longer, with Tim Hortons fastest at 4:20 and Chick-fil-A slowest at 7:06 (Intouch Insight, 2026 Drive-Thru Study). Satisfaction was 97.8% on visits with zero issues and 57.9% with two or more (QSR Magazine, 2026 QSR Drive-Thru Report).
Two numbers from the study matter more for automation than total time. Customized orders drove 43% of all errors, which is exactly the kind of request that trips speech systems. And friendly service was associated with 98.6% satisfaction against 26.8% without it, a 71.8-point swing that no speed gain can offset if the AI lane makes the exchange colder.
| Metric | Value | Source |
|---|---|---|
| Average total time (2026) | 5:24, 11.1 seconds faster than 2025 | Intouch Insight, 2026 Drive-Thru Study |
| Order accuracy, all lanes | 89.3%, up 2.4 points from 2025 | Intouch Insight, 2026 Drive-Thru Study |
| Overall satisfaction | 93.8%, up from 91.2% in 2025 | Intouch Insight, 2026 Drive-Thru Study |
| Accuracy with an order confirmation board vs without | 93% vs 88% | QSR Magazine, 2026 QSR Drive-Thru Report |
| McDonald’s drive-thru footprint | Nearly 29,000 locations globally; over 95% of about 13,700 U.S. restaurants | McDonald’s Form 10-K 2025 |
| Customers shown an upsell offer, by channel (2025) | Drive-thru 64%; in-store 75%; mobile pickup 57% | Intouch Insight, Emerging Experiences 2025 |
| Mobile order-ahead vs a traditional drive-thru order (2025) | Nearly 2 minutes less time | Intouch Insight, Emerging Experiences 2025 |
| Mobile order-ahead accuracy (2026, 449 mystery shops at 9 brands) | 92% | Intouch Insight, Emerging QSR Experiences Study 2026 |
Context: the drive-thru speaker now competes with the phone in the customer’s hand. Mobile order-ahead already beats the lane on time and matches or exceeds it on accuracy, and it needs no speech recognition at all. Only 44.4% of visits had an order confirmation board (Intouch Insight, 2026 Drive-Thru Study), even though the board is worth 5 points of accuracy in the QSR cut. A screen that shows the customer what the system heard is cheaper than a better speech model, and it helps hard-of-hearing and non-native speakers too, a theme that runs through our kiosk speech accessibility statistics. The full report is published by QSR Magazine.
6. Vendor Revenue and Deployments: Who Is Getting Paid
No vendor discloses drive-thru revenue as a line item, so the money has to be read sideways. SoundHound AI reported $168.9 million in 2025 revenue, up 99% from $84.7 million (SoundHound AI Form 10-K 2025), but that covers automotive, IoT, customer service and restaurants together, and its filings give no count of live drive-thru lanes.
The only clean drive-thru revenue number in any SEC filing is tiny. Presto Voice produced $0.8 million in the quarter ended March 31, 2024, with 163 live locations. That is the size of the business that was paying for the off-site human agents described in section 3, while reporting substantial doubt about its ability to continue as a going concern.
| Metric | Value | Source |
|---|---|---|
| SoundHound AI revenue, full year 2025 | $168.9 million, up 99% from $84.7 million | SoundHound AI Form 10-K 2025 |
| SoundHound AI revenue, Q2 2026 | $61.9 million, up 45% | SoundHound AI Q2 2026 earnings release |
| SoundHound AI full-year 2026 revenue outlook | $230 million to $260 million | SoundHound AI Q2 2026 earnings release |
| SoundHound AI GAAP net loss, Q2 2026 | $42.8 million | SoundHound AI Q2 2026 earnings release |
| SYNQ3 restaurant voice AI acquisition, total purchase consideration | $15.8 million (January 3, 2024) | SoundHound AI Form 10-K 2025 |
| Share of top 20 QSR brands working with SoundHound (February 2025) | More than 30% | SoundHound AI Q4 2024 earnings release |
| Presto Voice revenue, quarter ended March 31, 2024 | $0.8 million, 17% of total revenue | Presto Automation Form 10-Q, March 2024 |
| Hi Auto stores live (vendor claim) | About 1,000 | Hi Auto company website 2026 |
Context: SoundHound’s Q2 2026 release lists a large seafood QSR adopting its drive-thru ordering product and a pizza brand with its technology live in more than 75% of locations, without naming either or giving location counts. Its Q4 2024 release said its phone ordering technology had surpassed 100 million customer interactions, a reminder that a large part of its restaurant volume arrives by phone line, not by drive-thru lane. Figures are in the SoundHound AI Form 10-K for 2025 and the Q2 2026 earnings release.
7. Customer Acceptance and Friction
Customers are getting more experience with AI lanes and fewer of them are neutral about it. The share of shoppers rating the AI experience somewhat difficult or difficult and frustrating rose from 7% in 2025 to 19% in 2026 (QSR Magazine, 2026 QSR Drive-Thru Report), in the same year accuracy improved by more than 8 points.
That is not a contradiction. As the AI takes more complete orders, more customers reach the hard part of the conversation instead of being handed to a person after the greeting. Nearly three in ten shoppers had to repeat part of their order. The age split in stated preference is wide enough that a chain’s customer mix will decide the outcome more than the model will.
| Metric | Value | Source |
|---|---|---|
| Consumers who dislike the idea of voice AI drive-thru ordering | 45%, down from 47% a year earlier | Intouch Insight consumer survey 2024 (n over 1,480) |
| Dislike by age | 33% of ages 18 to 44 vs 54% of ages 45 and older | Intouch Insight consumer survey 2024 |
| Consumers who had experienced voice AI ordering | 19%, of whom 61% reported a positive experience | Intouch Insight consumer survey 2024 |
| Millennial and Gen Z adults who would order from an AI bot | Roughly 6 in 10 | National Restaurant Association, State of the Restaurant Industry 2026, reported by Restaurant Dive |
| Restaurants using AI for customer orders | 6% (26% of operators use any AI-related tool) | National Restaurant Association, State of the Restaurant Industry 2026, reported by Restaurant Dive |
| AI interaction rated easy vs difficult (2026 mystery shops) | 63% vs 21.7% | Intouch Insight, 2026 Drive-Thru Study |
| AI experience rated difficult (2025 vs 2026) | 7% vs 19% | QSR Magazine, 2026 QSR Drive-Thru Report |
| Customers comfortable with automated drive-thru voice assistants (vendor survey) | About 60%; 57% named long lines as their biggest pain point | SoundHound AI study, reported by Restaurant Business 2023 |
Context: overall satisfaction at AI locations was still 94.2%, and speaker clarity was rated 97.4% (Intouch Insight, 2026 Drive-Thru Study). The friction is concentrated, not general. The under-reported cost is at the window: staff said thank you 23.9 percentage points less often when AI took the order (QSR Magazine, 2026 QSR Drive-Thru Report). When the machine takes the order, the crew says less, and friendliness is the strongest satisfaction driver in the baseline data in section 5. The consumer survey figures come from Intouch Insight’s article in Modern Restaurant Management, and the operator figures from Restaurant Dive’s coverage of the NRA report.
Summary: Drive-Thru Voice AI Ordering by the Numbers
| Metric | Value | Source |
|---|---|---|
| Encounters where AI handled the full order (2026 vs 2025) | 40% vs 20% | QSR Magazine, 2026 QSR Drive-Thru Report |
| Total time when AI handled the full order vs non-AI | 4:19 vs 5:25 | QSR Magazine, 2026 QSR Drive-Thru Report |
| Order accuracy at voice AI locations (2026 vs 2025) | 90.8% vs 82.5% | Intouch Insight, 2026 Drive-Thru Study |
| Total time, AI locations vs same brands | 5:14 vs 5:15 | Intouch Insight, 2026 Drive-Thru Study |
| AI-led orders that still required an employee (2025) | 22% | Intouch Insight, Emerging Experiences 2025 |
| Taco Bell U.S. restaurants with voice AI | 890+ in 38 states | Omilia press release, July 2026 |
| Taco Bell voice AI orders processed (Q3 2024) | Over 2 million, in 300+ U.S. stores | Yum! Brands Q3 2024 earnings call, reported by Sixteen:Nine |
| Wendy’s FreshAI restaurants (May 2025) | 160+, target 500+ by year-end 2025 | Wendy’s Q1 2025 earnings call, reported by Restaurant Dive |
| Wendy’s FreshAI pilot orders without crew intervention | 86% | Wendy’s blog, December 2023 |
| McDonald’s IBM test size | About 100 restaurants | McDonald’s statement to Nation’s Restaurant News, June 2024 |
| McDonald’s drive-thru locations globally | Nearly 29,000 | McDonald’s Form 10-K 2025 |
| Presto orders needing no human agent entry (most advanced sites) | Approximately 46% | Presto Automation Form 10-Q, March 2024 |
| Presto Voice revenue, quarter ended March 31, 2024 | $0.8 million | Presto Automation Form 10-Q, March 2024 |
| SEC order against Presto: period of misleading statements | November 2021 through May 2023 | SEC Administrative Proceeding File No. 3-22413, January 2025 |
| SoundHound AI revenue, full year 2025 | $168.9 million, up 99% | SoundHound AI Form 10-K 2025 |
| SoundHound AI 2026 revenue outlook | $230 million to $260 million | SoundHound AI Q2 2026 earnings release |
| Restaurants using AI for customer orders | 6% | National Restaurant Association, State of the Restaurant Industry 2026 |
| Consumers who dislike voice AI drive-thru ordering | 45%; 54% of ages 45 and older | Intouch Insight consumer survey 2024 |
Methodology and Sources
We ran targeted searches on October 11, 2026, then read the primary documents directly: SEC filings on EDGAR, company releases, and the two publications of the 2026 drive-thru mystery-shop fieldwork. Company numbers come from filings, earnings releases or company statements; where a trade publication carried the statement, both are named. Vendor performance claims are labeled as claims. Every time is written as minutes:seconds, as the source reports it.
- Intouch Insight, 2026 Drive-Thru Study: Key Drive-Thru Trends (26th annual study; 165 mystery shops per brand at 13 U.S. chains plus 120 voice AI orders at Burger King, Taco Bell and Wendy’s; fieldwork June to July 2026).
- QSR Magazine, The 2026 QSR Drive-Thru Report (published October 1, 2026; same fieldwork, with 174 total AI engagements including 79 incidental encounters).
- Intouch Insight, Emerging Experiences 2025 press release (May 2025) and the Emerging QSR Experiences Study 2026 (449 mystery shops of mobile order-ahead at 9 brands).
- CNBC, AI drive-thru ordering is on the rise, republished by NBC News (July 3, 2024; SoundHound vendor claims and the reporting on the McDonald’s test).
- Intouch Insight consumer survey 2024, published in Modern Restaurant Management (more than 1,480 North American consumers).
- Omilia press release on Taco Bell’s voice AI expansion (July 7, 2026) and Nation’s Restaurant News coverage (July 8, 2026).
- Restaurant Dive on the Yum! Brands press release of July 31, 2024 and Sixteen:Nine on the Yum! Brands Q3 2024 earnings call.
- NVIDIA blog, Yum! Brands and NVIDIA partnership (March 18, 2025).
- Wendy’s blog, Drive-Thru Innovation with Wendy’s FreshAI (December 11, 2023), Restaurant Dive on the pilot results (December 2023) and Restaurant Dive on the Wendy’s Q1 2025 earnings call (May 2, 2025).
- Nation’s Restaurant News, McDonald’s is ending its AI drive-thru test with IBM (June 17, 2024) and the McDonald’s Form 10-K for 2025.
- Restaurant Business, White Castle is expanding its use of drive-thru AI (August 2, 2023).
- Presto Automation Form 10-K for fiscal 2023, Form 10-Q for March 31, 2024, Form 8-K filed September 2024 and the SEC administrative proceeding summary, File No. 3-22413 (January 14, 2025).
- SoundHound AI Form 10-K for 2025, Q2 2026 earnings release and Q4 2024 earnings release.
- National Restaurant Association, State of the Restaurant Industry 2026 (paid report; operator and consumer figures taken from Restaurant Dive’s summary, February 18, 2026).
- Hi Auto company website, 2026 (vendor marketing claims on store count and completion rate; not independently verified).
- Data watch: Intouch Insight and QSR Magazine publish different cuts of the same 2026 fieldwork, and they do not always match. Intouch reports AI-location accuracy of 90.8% (82.5% in 2025), while QSR reports accuracy in AI encounters of 89% (84% in 2025). Intouch says the AI took some or all of the order on 92 of 120 targeted visits, while QSR counts 95 AI interactions at the same 120 shops. Chick-fil-A’s total time is 7:06 in one and 7:05 in the other. We cite each figure to the publication that printed it. Completion rates are not comparable across vendors: Presto’s 85% counted interventions by restaurant staff only, Wendy’s 86% counted orders with no crew help, and the mystery-shop 40% counts whole encounters. Hi Auto’s site gives two different annual order volumes, so we used neither. We could not verify a primary source for drive-thru’s share of quick-service sales, a current Wendy’s FreshAI restaurant count for 2026, or Bojangles’ deployment figures, and left all three out. The Presto fiscal 2023 and White Castle 2023 figures are the most recent available data and are now about three years old. Newer data is expected from SoundHound AI’s Q3 2026 filings and from Wendy’s and Yum! Brands third-quarter earnings calls.
Last updated: October 11, 2026. We update this roundup quarterly, and the next refresh is expected when SoundHound AI files its Q3 2026 Form 10-Q and earnings release.