Dairy Queen deploys Presto AI voice chatbots in drive-thrus

The story Dairy Queen and Presto want to tell this week is a clean one: successful pilots, promising early data, an ice-cream-season rollout across 25+ franchisees and, eventually, substantially all of Dairy Queen's roughly 3,000 U.S. and Canadian drive-thrus. The story they are not telling is the one from December 2023, when reporting based on Presto's own SEC filings revealed that off-site agents — often in places like the Philippines — assisted Presto Voice chatbots in more than 70% of customer interactions. Presto itself said at the time, "Our human agents enter, review, validate and correct orders. Human agents will always play a role in ensuring order accuracy." The U.S. SEC opened an investigation into claims about the company's AI technology in July 2023, and there is no public indication that inquiry has been resolved.

What makes that history load-bearing for today's news is scale. A pilot with human backstops is a defensible staffing model. An announcement that targets ~3,000 drive-thrus without clarifying how much of each order is actually handled by the model — versus a remote agent reviewing, correcting, or leading the interaction — turns a product decision into a disclosure question. Presto's own marketing page describes operational modes including "Supervised AI" with "constant oversight from agents" and an "Agent-led" mode where "remote agents lead order management using Voice AI infrastructure," which means the company's own product taxonomy contemplates deployments where the AI is not the primary decision-maker. Neither Dairy Queen nor Presto has said which mode franchisees will be running, and that silence is the loudest thing about this announcement.

Dairy Queen's menu is the reason this deployment is technically interesting rather than routine. EVP of IT Kevin Baartman said the single biggest evaluation criterion was menu ingestion — "One of the key things we looked at was its ability to ingest our menu. It's done a good job with it" — because Dairy Queen's catalog spans over 1 million order combinations that sit across two fundamentally different product families: frozen treats and hot food. Most QSR voice AI deployments covered to date (Carl's Jr., Hardee's, Taco John's, Wienerschnitzel) are hot-food-dominant chains with comparatively linear combo logic. A Blizzard with custom mix-ins ordered alongside a chicken basket and a kid's meal is a different kind of parse problem.

That complexity also explains the timing. The rollout is "phased to begin over the next few weeks to capitalize on ice cream season," which is the window in which Dairy Queen's frozen-side economics are strongest and in which cross-selling hot food to treat-focused customers has the highest incremental value. Presto CEO Krishna Gupta framed that cross-sell as the real prize: "It was very important that someone could handle that complexity in real time. And then it opened up a world of possibility that you can expose guests to the food in addition to the treats and vice-versa." Baartman's claim that Presto handled Free Cone Day — the brand's annual free-soft-serve promotion that creates traffic spikes most systems would buckle under — with the simple verdict "It was able to handle everything" is the operational proof point Dairy Queen is leaning on.

Strip away the customer-service framing and this is, commercially, an upsell story. Presto's own voice-ai product page markets "up to 6% incremental revenue increase in drive-thru operations" and advertises that its system makes "4x more context-specific upsell attempts" than human staff, while claiming "up to 90% non-intervention rates." Those numbers matter because they describe a business case that can pencil out even in a world where the AI is imperfect: a modestly accurate bot that never forgets to ask about a Blizzard upgrade can outperform a tired teenager on a Saturday rush. That is why the economics survive the transparency problem — the ROI is in the upsell frequency, not the linguistic miracle.

There is, however, a second-order finding that complicates the pitch. Academic research out of Penn State found that customers are more likely to choose indulgent items when ordering via voice AI than when ordering from a human employee. Layered onto Presto's 4x upsell-attempt claim, that suggests the revenue lift isn't just about asking more often — it's about asking a customer who is psychologically more willing to say yes to a larger Blizzard or to add fries they wouldn't have ordered from a person. For a brand whose identity is already anchored in indulgence, that is a powerful flywheel. For regulators, franchisee operators balancing complaint volume, and public-health observers, it's a set of incentives pointed in a very specific direction, and one worth watching as the deployment scales past pilots into thousands of locations.

Dairy Queen is stepping into a category that just had its highest-profile failure. In June 2024, McDonald's announced it was ending its AI voice drive-thru test with IBM, shutting the technology off at test restaurants by July 26, 2024 after persistent issues with accents, misorders, and lane mixing. That shutdown is the cautionary mirror to Gupta's confident claim that "there's always gonna be some people who are anti-AI. But I expect that as a macro trend, that will shift pretty quickly in our favor." McDonald's had the scale, budget, and vendor heft to brute-force the problem, and still walked away. Presto's bet with Dairy Queen is that a more menu-aware, more upsell-tuned, and (critically) human-assisted architecture can land where IBM's could not.

The consumer data is not on Gupta's side — yet. A January 2025 YouGov survey cited in coverage of the rollout found 55% of Americans prefer human order-takers while only 4% prefer AI chatbots, and among people who have actually tried an AI drive-thru chatbot, only 14% said they'd pick it over a human. That preference gap is showing up raw on social: Reddit communities are treating AI drive-thrus as "a glorified automated phone call," venting about broken customizations ("no tomato add bacon"), aggressive upsells ("For just a dollar more" being met with "SHUT UP ROBOT"), and folk workarounds like ordering 55 of everything to force a human to take over. A minority of Wienerschnitzel customers have specifically praised Presto, which suggests the tech can win favor in the right context — but the sentiment gradient between journalists covering the partnership neutrally on X and customers venting on Reddit is the clearest near-term risk to a 3,000-store rollout. Dairy Queen is betting that Gupta's macro-trend thesis arrives before enough customers post the wrong viral video.