The ArchIQ Mandate: McDonald's Next and the Question Nobody Asked in Las Vegas

During the first week of June, roughly 15,000 franchisees, suppliers, and corporate staff filled Las Vegas for the McDonald's Worldwide Convention — the biennial gathering where the company sets the system's direction for the next two years. What they got was "McDonald's Next," a growth strategy spanning menu, hospitality, restaurant design, and technology. The centerpiece was ArchIQ, an AI operating system built with Google, including a drive-thru voice assistant nicknamed "Archy" that is currently live in five U.S. test stores.

CEO Chris Kempczinski framed the ambition in a single line: "We can't ask our customers to choose. Hospitality or speed."

The trade press covered the convention as an AI story. The consumer press covered it as a drive-thru story. Both coverage angles miss the only question that matters to the people who actually own the restaurants: who pays for this, who controls it, and what happens to the operator whose unit economics degrade under a system they were required to install?

That question has a fresh, specific, $100 million answer sitting in a Texas courtroom right now. Five weeks before McDonald's took the stage in Las Vegas, a 111-unit Pizza Hut franchisee sued its franchisor over a mandated AI delivery system that allegedly collapsed its delivery metrics. The two stories are not separate. They are the same story at different points on the timeline.

What Was Actually Announced

Strip away the keynote production and McDonald's Next contains four operationally significant components.

First, ArchIQ itself — positioned not as a single tool but as an operating system layer for the restaurant: order taking, kitchen coordination, and the connective tissue between them. The voice assistant is the visible piece. The platform underneath is the consequential one.

Second, the early performance claims. A widely followed franchisee account reported that the system has processed more than one million transactions across the test stores with roughly 90% of orders completed without human escalation, operating in both English and Spanish — and that U.S. restaurants are receiving Google Edge Cloud hardware installations "in anticipation of this rollout." Treat those figures with appropriate caution: they come from franchisee channels, not from McDonald's corporate disclosure, and the company has not published verified accuracy data. But the hardware claim is the one to watch. Servers being physically installed in restaurants ahead of a system-wide deployment is not a pilot. It is a rollout with the press release pending.

Third, a new restaurant redesign — a prototype available to franchisees with an upcoming remodel. In franchising, "available with your remodel" is a phrase with a known life cycle. Optional prototypes have a way of becoming brand standards, and brand standards have a way of becoming renewal conditions. The remodel-at-renewal trap is one of the oldest structural red flags in the franchise agreement, and a new prototype unveiled at a convention is the starting gun for the next remodel cycle.

Fourth, the margin context nobody put on a slide. McDonald's spent the past year co-investing with franchisees to subsidize value-menu pricing — and that subsidy support was slated to wind down at the end of the first quarter of 2026, shifting the cost of the value war back onto operator P&Ls. The system is also rolling out standards that assess whether individual operators' menu prices are too high. Read those two facts together: corporate is stepping back from funding the discounting while reserving the right to police the pricing. Into that margin environment arrives a technology platform with an unpublished cost structure.

The Second Attempt

McDonald's has run this experiment before, and the first run is worth remembering precisely because the company would prefer you didn't.

In June 2024, McDonald's ended its Automated Order Taker test with IBM — a voice AI deployed across more than 100 drive-thrus — instructing restaurants to shut the system off by late July of that year. The technology had become a social media genre of its own: bacon added to ice cream, hundreds of dollars of unwanted nuggets. BTIG analyst Peter Saleh put numbers on the failure at the time, noting accuracy "in the low-to-mid 80% range" and high operating costs.

Two years later, the vendor has changed, the architecture has changed, and — if the franchisee-reported figures hold — the accuracy has materially improved. An escalation-free completion rate around 90% across a million transactions would represent real progress over the IBM run.

But the investor lesson from the IBM episode was never about accuracy. It was about who absorbed the cost of the experiment. Operators who installed, trained on, and worked around a system that was withdrawn two years later did not get those hours or that disruption back. The technology improved. The risk allocation did not. That allocation is written in the same place it has always been written: the franchise agreement and the operations manual it incorporates.

Follow the Hardware

Here is where the Operations analysis has to get specific, because the costs arrive through three separate doors.

Door one: the technology stack itself. Edge computing hardware, integration, maintenance, and the monthly platform fees that have historically accompanied every McDonald's technology initiative. The company's technology fee structure has been a point of open franchisee friction for years. ArchIQ's pricing to operators has not been disclosed, which means every operator modeling 2027 cash flow is modeling around a blank line item. When you evaluate a franchise technology stack, the undisclosed fee is the fee that deserves the most scrutiny — and a platform described as the restaurant's operating system will not be priced like an app.

Door two: the remodel prototype. McDonald's 2024 FDD put the estimated initial investment for a traditional restaurant at $1,470,500 to $2,728,000, with a $45,000 franchise fee, a 40% cash down payment requirement on new restaurants, and a royalty raised to 5% for new franchise agreements as of January 2024. Those are the entry numbers. The reinvestment numbers — the remodel cycles, the equipment refreshes, the technology retrofits — are what determine whether a McDonald's operator builds wealth or services a treadmill. A new prototype announced in June 2026 tells you what the early-2030s remodel mandate will look like. The operators who will absorb it cheaply are the ones who start modeling it now, the same way a disciplined break-even model prices a known future capital call instead of discovering it.

Door three: labor reallocation, not labor savings. The pitch for drive-thru AI is always labor efficiency. The operational reality, documented across every voice-AI deployment in the category, is reallocation — the order-taker becomes an expeditor, a runner, a kitchen position. That can genuinely improve throughput. It rarely reduces headcount at the rate the investor deck implies. Operators who underwrite ArchIQ as a labor-line reduction are repeating the oldest modeling error in franchise labor analysis: pricing the promise instead of the deployment.

The Operations Manual Is the Delivery Vehicle

None of this arrives through contract renegotiation. It arrives the way every franchisor mandate arrives — through the operations manual amendment clause that nearly every franchise agreement contains and almost no franchisee prices.

The Chaac v. Pizza Hut complaint, filed May 6, is the live demonstration of how that mechanism performs under stress. Pizza Hut mandated Dragontail, an AI delivery platform owned by its own parent company. Chaac alleges the mandated rollout took its on-time delivery rate from over 90% to roughly 50%, tripled the time finished pizzas sat waiting for drivers, and converted a top-performing franchisee into a below-system-average one — and it is now suing for no less than $100 million in lost business and enterprise value.

The ArchIQ situation differs in one structurally important way: Google is a third-party vendor, not a captive affiliate. McDonald's is not collecting vendor margin on both sides of the mandate the way Yum does with Dragontail. That removes one conflict of interest. It removes none of the operational risk. A mandated system that degrades drive-thru times hurts the operator identically whether the franchisor profits from the vendor relationship or not — and a third-party vendor relationship adds its own exposure, because the system's roadmap, pricing, and support now sit with a counterparty the franchisee has no contract with at all. This is the same captive-customer dynamic that surfaced in the Pizza Hut sale process, where franchisees face the prospect of depending on a technology platform their franchisor may no longer even own.

The Pre-Rollout Baseline

The single most actionable lesson from the Chaac litigation is that the franchisee is litigating retrospectively — reconstructing pre-rollout performance to prove post-rollout damage. The architects in the McDonald's system, and in every system with an AI mandate on the horizon, will build that record prospectively. The window to do it is now, before installation, not after.

The baseline is not complicated. It is the discipline of a weekly KPI dashboard pointed at the metrics the new system will touch: drive-thru total experience time, order accuracy and remake rate, cars-per-hour at peak, average check, labor hours by daypart, and customer satisfaction scores. Twelve months of clean weekly data, captured before the hardware goes live, converts a future dispute from anecdote into evidence. It also does something more valuable than litigation prep — it tells you within a single rollout cycle whether the system is helping or hurting, while there is still time to escalate through the field organization instead of through counsel.

Pair the baseline with three questions to the franchisor, in writing: What is the per-unit cost of the ArchIQ hardware, installation, and ongoing platform fee? What were the verified accuracy and throughput results from the five test stores, by daypart? And what is the remediation process if a unit's drive-thru metrics degrade post-installation? The answers matter. The willingness to answer matters more. Then validate with operators in the test markets — the five stores running Archy today contain more truth about this system than any convention keynote.

This Is Not a McDonald's Story

Zoom out and the convention announcement is one data point in a category-wide pattern. Yum's partnership with NVIDIA targets a 500-restaurant AI rollout. Taco Bell runs voice AI in more than 500 drive-thrus. Burger King is piloting an OpenAI-based assistant across 500 restaurants. Wendy's FreshAI operates in hundreds of locations. Every major QSR franchisor is now in the business of mandating intelligent systems into franchisee-owned restaurants — which means every major QSR franchisee is now in the business of absorbing technology risk they did not select, from vendors they did not choose, at prices they did not negotiate.

The operator reacts to each mandate as it arrives. The architect recognizes the era: the operations manual has become the primary channel through which franchisors deploy capital obligations onto franchisee balance sheets, and AI platforms are simply the largest object yet to move through that channel. The true cost of the franchise was never the fee on page one. In 2026, it is increasingly the technology line — recurring, mandated, and repriced at the franchisor's discretion.

McDonald's may execute this rollout well. The test-store numbers, if they hold up under disclosure, suggest a genuinely better system than the one that failed in 2024. But "the franchisor executed well" is a hope, not a position. The position is a baseline, a cost model, and a paper trail — built before the blades are bolted into your back office.