AI Adoption in Franchising: What the 28% Actually Counted

AI adoption in franchising is usually summarised with a single survey line, and the line is misread. What the FRANdata and IFA franchisor survey counted was mentions of a plan, not systems in production, which makes the familiar 28% both an overstatement and an undercount depending on which question you meant to ask.

I have now seen that figure on three slides in three rooms, relabelled each time as the share of franchisors who have adopted AI. It is a useful number. That is not what it says.

What the survey counted

Here is the line, in the survey's own words, from FRANdata's franchisor technology research.

Three in four franchisors — 75% — expect to increase capital spending on technology and innovation. Separately, 28% mentioned incorporating AI and increased automation.

The second figure counts mentions. Somebody at a franchisor, asked about plans, named AI and automation among them — a statement of intent recorded in a survey response, and a long way from a system a franchisee touches on a Tuesday.

So as a measure of deployment, 28% is generous. Read it instead as a measure of how much AI is touching brand work and it is far too low, for a reason unrelated to survey design.

The survey asked headquarters about capital spending. The AI in use across most networks generates none. It sits in personal accounts, on free tiers, at the unit — the operator drafting a schedule, the manager rewriting a training sheet, the owner working out what to do about a supplier who shorted a delivery. None of that reaches a franchisor's technology plan, because the franchisor did not buy it and often does not know about it.

Both readings hold at once, which makes the number interesting rather than damning:

• The share of franchisors running governed, brand-sanctioned AI is well below 28%.
• The share of networks with AI somewhere inside them, working on brand material, is far above it.
• The distance between those two is a governance deficit, not a stage on an adoption curve, and it widens quietly.

What survives all that is a narrow description of the 28%, and it is still useful: by early 2025, that share of franchisor leadership teams had the subject far enough along to say the word out loud in a planning document.

Why AI adoption in franchising stalls, and it is not scepticism

Nobody I meet thinks this is a fad. The delay is structural, and three forces do most of it.

The payoff and the liability sit in different pockets. A good answer helps a location close on time. A bad answer — a wrong refund threshold, a confidently invented employment rule — arrives at the brand, because the brand's name is on the sign and on the demand letter. A franchisor weighing a modest gain at the unit against a tail risk at the top of the system is reading their own incentives correctly, not being timid, which is why the fix has to change the risk shape rather than the enthusiasm level.

The obvious response looks like control. A system that answers a franchisee's staff on scheduling, discipline or hours is a brand giving instructions inside an independent business. Most franchisors' instinct is to check that with counsel, which is sensible, and the check carries no deadline, which is fatal. It goes on a list. The list has no owner. Eighteen months later the tool is unbuilt and the network's usage has doubled.

There is no category to buy. A franchisor's default move before committing to a platform is to ask which comparable brand already runs it and how that went. Here, in most segments, there is nobody to ask — no reference call, no analyst grid, no peer two states over who can say what broke. That constraint is social rather than technical, and it resolves abruptly: two visible brands in a segment go first, and the rest follow inside a year.

Why the gap closes anyway

Three reasons, none of which depend on vendors getting more persuasive.

The demand is already inside the network and it is not waiting for anyone's roadmap. Your operators have a capable tool, they found it themselves, and their reasons are practical rather than rebellious — why operators found the tool themselves.

A mention rate is a leading indicator. What franchisors name in a planning survey becomes a budget line within a cycle or two, because naming it creates the internal expectation that somebody will act. The 28% is less a report on the present than a forecast of the next two planning seasons.

And the funding conversation is already structured for it. Roughly six in ten franchisors disclose a technology fee in FDD Item 6, on IFA's analysis of FDD data — so the mechanism, the disclosure and the franchisee-council conversation about network technology all exist. Nothing has to be invented for this to get paid for. It only has to be defensible in that meeting.

What being early buys, and what it costs

Not a technology lead. The models are rented from the same short list by everybody, so nobody builds a moat out of intelligence.

What an early start buys is calendar time on the parts you cannot compress later. Your document set has to be current, deduplicated and machine-readable before any of this answers reliably — the unglamorous prerequisite in grounded AI for franchise operations — and that takes months of somebody's attention whichever product you choose. It also buys a network that has used the thing on low-stakes questions for a year before you need it for a contentious one: the difference between a tool operators trust and a tool they were sent a memo about.

The cost of being early deserves saying plainly. You will own the first wrong answer, and be the brand explaining it on a panel. Budget for that rather than pretending the risk sits elsewhere.

And late is survivable. This is not a winner-take-all market; no brand disappears for deploying governed AI in 2028 rather than this year. Waiting costs two more years of usage you cannot see, and a document set nobody cleaned.

The number I would rather have

Industry adoption rates make good slides and poor decisions. No version of the 28% tells you anything about your own brand, and a founder quoting it in a board meeting has borrowed somebody else's uncertainty.

One question does the work instead. How many of your operators asked a machine something about your business last month, and what did they ask? No analyst can supply that. It takes a survey nobody has to sign, sent by operations rather than legal, and the free-text answers matter more than the percentage at the top — they are a list of what your own material failed to answer, in your operators' words.

The 28% does not measure how many franchisors have adopted AI. It measures how many had a sentence ready about it when somebody asked. Your operators did not wait for the sentence.

Surviving a network is a higher bar than launching in one: governed AI, built rather than announced.