The payback worksheet: deciding whether a mandated AI upgrade clears the hurdle at one outlet

Why this is usually impossible

When a brand mandates a technology upgrade, franchisees are rarely given the numbers needed to evaluate it. The benefit arrives as a claim about efficiency, and the cost arrives as an invoice. That makes a real decision impossible, so most operators either comply or resist on instinct.

This week produced an unusually complete public example. Use its figures as a template for the method, and replace every one of them with your own.

The published template

At its investor day on 23 September 2026, McDonald s announced $8.5 billion of NEXT partnering support through 2036, delivered to franchisees as rent relief and capital support, targeting about 250 basis points of gross restaurant-level efficiency, per 24/7 Wall St. Food Chain Magazine reported on 28 September 2026 that about $5 billion is expected by 2030; that per-restaurant investment runs about $800,000 for a US drive-thru and $650,000 to $700,000 internationally; that the average US restaurant is expected to gain about $100,000 of annual cash flow; and that this implies a payback of roughly four years. The programme bundles redesign, kitchen equipment, revised procedures and ArchIQ, a generative-AI-enabled platform.

So the template is: cost per outlet, annual cash-flow benefit per outlet, payback period, and a system-level efficiency target. Those four lines are what any franchisor should be able to supply, and what any franchisee should request before funding an upgrade.

Rebuilding it for one outlet

**Line 1: your cost, not the average cost.** Start from the quoted per-outlet figure, then add what the average hides: site-specific construction, permits, any landlord consent, and equipment differences. Ask explicitly whether the quoted figure is before or after the brand contribution.

**Line 2: split the benefit by source.** A bundled programme reports a blended gain. Ask the brand to attribute it across labour hours saved, throughput gained, waste reduced, and any price or mix effect. This matters because the AI layer and the building work have very different costs and can sometimes be bought separately.

**Line 3: scale the benefit to your volume.** A gain quoted for an average restaurant is a function of that restaurant volume. If your outlet runs at 70% of system average revenue, a labour-efficiency or throughput gain scales down roughly in proportion. Applying the system average to a below-average outlet is the single most common way a payback calculation flatters itself.

**Line 4: subtract the disruption.** Weeks of closure or reduced capacity during a remodel are a real cash cost that almost never appears in a published payback. Estimate lost contribution over the works period and treat it as part of the investment.

**Line 5: model the support correctly.** Rent relief lowers ongoing cost month by month. A capital contribution lowers the upfront outlay once. They are not interchangeable: an operator close to a financing limit needs the second, while one with thin monthly cash flow needs the first. Put them on the schedule they actually arrive on.

**Line 6: add the cost of the money.** If the outlay is financed, interest belongs in the calculation. A four-year payback before financing cost is a longer payback after it.

**Line 7: test the downside.** Recompute at 50% and 70% of the promised benefit. If the outlet still survives its loan schedule at 50%, the decision is robust. If it only works at 100%, you are relying on a company estimate that nobody has yet verified in practice.

Reading the hurdle honestly

A payback period is not a return. It says when you get your money back, not what you earn afterwards. Two upgrades with identical four-year paybacks differ enormously if one keeps delivering for ten years and the other needs replacing in five. Ask about the expected useful life of the equipment and the software contract term, because a technology benefit that depends on an ongoing licence is a subscription, not an asset.

Also check what happens to the benefit if the brand changes the system later. Efficiency gains that depend on a central platform are only as durable as the brand commitment to that platform.

A note on scepticism

Markets were not uniformly convinced by the announcement. On 23 September McDonald s shares fell 4% to $239.56, while Wendy s slipped 1% and Yum Brands eased 0.4%. The bear argument reported alongside was that spending begins now while benefits arrive over a decade. That is the same question a single franchisee faces, at a smaller scale and with less ability to wait.

It is also worth separating efficiency from demand. The same reports note second-quarter 2026 comparable sales rose 1.3% globally and 0.8% in the US. An efficiency programme improves what you keep from each sale; it does not by itself bring more customers.

What we do not know

• Every figure in the template is a company estimate published at an investor day, not a verified outcome. No evidence yet demonstrates the $100,000 annual gain or the four-year payback in practice.
• The split of the $8.5 billion between rent relief and capital contribution is not broken out, so the cash-flow timing for an individual operator cannot be modelled from public information.
• No attribution of the 250 basis points across ArchIQ, equipment, redesign and procedures has been published, so the AI portion alone cannot be isolated.
• Equipment useful life, software contract terms and any ongoing licence cost were not disclosed in these reports.
• Whether participation is mandatory, and the consequences of declining, is not addressed.

This article is educational analysis of publicly available information. It is not investment, legal or accounting advice, and it promises no level of return.