Berkshire Hathaway-owned Dairy Queen CEO Troy Bader disclosed at the company's annual shareholder weekend that the chain is holding its Middle East expansion, citing war in Iran and shipping restrictions through the Strait of Hormuz that require finding alternative supply routes for products that cannot be sourced locally, while simultaneously confirming a test deployment of Presto voice AI ordering in approximately 50 drive-thrus where initial trials showed 90% order accuracy and a target of exceeding 99%.
The geographic detail matters for understanding the decision. Dairy Queen currently operates in Bahrain, Kuwait, Oman, Qatar, and the United Arab Emirates, and had been actively pursuing Saudi Arabia as its next Middle East market. The pause is not a retreat from existing operations. It is a suspension of new country entry, driven specifically by franchise economics in an environment where supply chain disruption adds cost and uncertainty to the early-stage scaling that Bader explicitly flagged as the hardest part of building a new brand in a new country. His phrasing is precise and worth noting: franchisees in the region are saying they want to wait and see. That is not corporate headquarters pulling back from a strategic bet. It is the franchise network, the entities that carry the actual capital risk of new location openings, exercising caution in response to supply and demand conditions that corporate messaging cannot fully control. When the people bearing the risk say wait, franchise systems stop, because the alternative is opening with under-committed operators who lack the conviction to push through the first difficult months of a new market.
The Presto AI deployment happening simultaneously is the more consequential operational decision, and it is explicitly positioned by Bader as a labor reallocation rather than a labor reduction. His description of the system's purpose is worth reading carefully: without AI at the ordering window, the employee at the window is doing order-taking as 100% of their job. With AI monitoring and processing orders, the same employee can monitor product quality, provide hospitality in other ways, and handle exceptions without being anchored to the speaker. That framing is standard restaurant industry AI positioning, and it is not entirely dishonest. The initial 90% accuracy figure is the number that requires examination. Ten percent of drive-thru orders going wrong before human intervention is a high error rate by any normal customer experience standard, and Bader's acknowledgment that employees monitor AI-generated orders is an admission that the system is not autonomous at current accuracy. McDonald's terminated its IBM AI drive-thru program last year partly for similar reasons. Taco Bell has expressed caution about its own AI drive-thru deployments after customer complaints. Burger King is testing at fewer than 100 locations. The restaurant industry is running distributed experiments with AI ordering at a scale where the failure modes are visible to customers in real time, and the 90% accuracy floor is a known liability that every vendor in this space is working to close.
Presto's position in this deployment is worth understanding for AI infrastructure founders watching the restaurant vertical. The company has faced significant operational difficulties, delisting concerns, and revenue challenges since its SPAC listing in 2022, and the Dairy Queen partnership is one of its largest deployments. CEO Krishna Gupta's public statement about the "human-AI waltz that just works" is aspirational language for a system that Dairy Queen's own CEO describes as currently needing human monitoring on every order. Presto's model, like most restaurant AI voice vendors, charges on a per-location basis, which means the unit economics of the business depend on persuading large franchise networks to deploy broadly rather than in pilot configurations. Dairy Queen's 50-location test, targeting eventual deployment across substantially all of its roughly 3,000 US and Canadian drive-thrus, is the commercial milestone that Presto's business model requires to demonstrate viability. The technology works well enough to expand the test. It does not yet work well enough to remove human supervision from the ordering process.
The franchise adoption dynamic is the structural observation that matters most for AI vendors selling into restaurant chains. Franchisors like Dairy Queen, McDonald's, and Yum Brands negotiate technology partnerships at the corporate level but implementation risk sits with individual franchisees who pay for hardware installation, staff retraining, system integration, and the customer experience consequences of AI errors. A franchisor can recommend or encourage a technology but cannot easily mandate adoption across a network of independently owned operators who have their own views on return on investment. The AI ordering vendors that are gaining franchise network traction are those that can demonstrate payback in under twelve months on hardware and integration costs, that provide human monitoring dashboards that give franchisee operators visibility into what the AI is doing without requiring them to be technology specialists, and that handle system failure gracefully by routing to a human operator rather than failing the order entirely. Dairy Queen's choice of Presto includes a monitoring layer precisely because the franchise network would not accept a fully autonomous system without operator visibility. That requirement is not a deficiency in the technology. It is the minimum viable governance structure for a franchise-based deployment where the brand risk of a failed order is borne by an owner-operator who did not choose the AI vendor.
The broader AI deployment pattern in fast food reveals something specific about where automation investment survives macro and geopolitical pressure. Dairy Queen is pausing Middle East expansion because franchisees are bearing capital risk in an uncertain environment. It is not pausing the Presto deployment because that investment is being made by corporate or absorbed into the technology fee structure in ways that reduce individual franchisee capital exposure. When geopolitical stress or demand softness creates pressure, franchise networks contract the investments with the highest perceived risk and the longest payback period, which is new market entry, and maintain or accelerate the investments that are positioned as cost reduction tools with near-term payback, which is what AI voice ordering vendors pitch. That dynamic is not specific to Dairy Queen. It describes the operating logic under which restaurant AI adoption will expand during the next period of macro uncertainty, precisely because the pitch that AI frees employees for higher-value tasks has more credibility in an environment where labor costs are elevated than in one where they are not.
","excerpt":"Dairy Queen CEO Troy Bader disclosed at Berkshire's shareholder weekend that the chain is pausing Middle East expansion due to war in Iran and Strait of Hormuz shipping restrictions that have made franchise operators cautious, while simultaneously confirming a Presto AI voice ordering test at 50 drive-thrus where initial accuracy is 90% with a 99% target.
Also read: Developers Are Asking If Codex Is the Best Coding Agent Right Now and the Answer Depends Entirely on What You Mean by Best • FastDMS Claims 6.4x KV Cache Compression While Running Faster Than vLLM and the Benchmark Numbers Are Credible Enough to Take Seriously • The Senate Just Voted 22-0 to Ban AI Companions for Minors and Every Founder Building Emotionally Engaging Consumer AI Needs to Read the Bill Carefully
