The State of AI in Franchising, 2026: Grading January's Predictions

The state of AI franchising 2026, at the midpoint: of five January predictions only one was due now, and it half-landed. Two more are interim halves, one cannot be settled before December, and one turns out to be unfalsifiable from outside. The most useful finding is how many were written badly enough to be ungradeable.

In January I published five predictions and promised to mark them in July, on the grounds that a prediction nobody can settle is marketing. The marking is worse than I would like. The list sits in franchise technology trends 2026, in its original wording, and leaving that wording alone turned out to be the part I had underestimated.

The grading rule, and the flaw I built into it

January's test for a gradeable prediction had three parts: it names a threshold, it names who must act, and a stranger can check it from outside.

Two of the five fail the third test, and I did not notice while writing them. A council reviewing a policy, a headquarters role quietly not backfilled — these happen inside organisations that publish nothing. I wrote thresholds that can only be met in rooms with no window.

One further honesty: only the second prediction carried a July deadline. The rest run to December, so what follows is a verdict on one and an interim mark on four — and a scorecard reporting only its checkable half is how an annual list becomes a ritual.

One measurement problem sits underneath all of it. The most recent thing anyone can cite about AI here is still the franchisor survey in which 75% of franchisors expected to increase capital spending on technology and innovation, and 28% mentioned incorporating AI and increased automation — FRANdata and the IFA's work, fielded in early 2025. Half a year of announcements has produced nothing newer, which tells you about the pace of measurement rather than of adoption. How thin the published record gets past the headline totals is the subject of franchise industry statistics.

The state of AI franchising 2026, prediction by prediction

Prediction one: AI language reaches the franchise relationship. The threshold was a national brand adding AI-use terms to a manual addendum or disclosure document and calling it a standard publicly, plus an advisory council formally reviewing a policy before issue.

Grade: half at best, and the second clause is ungradeable. Manual addenda are not public documents and councils do not publish minutes. What I can report is what I hear in rooms: brands have stopped asking whether to permit AI use and started asking what to say to a franchisee already using it. That is movement in the right direction, it is not evidence, and I will not dress it as evidence.

Prediction two: push beats pull, and the dashboard stops being the demo. The July one, and the only test I could run myself: sit through three vendor demonstrations and count how many open on a grid of tiles.

Grade: half. The opening slide has changed almost everywhere: an alert, a digest, an exception with a location's name on it. Three clicks in, the grid is still the product, and when a demonstration goes off-script the presenter navigates back to it, because that is where the answers live. The language moved a year ahead of the default. I also set a threshold about demos rather than products, which is my error rather than the industry's.

The two that run to December

Prediction three: standard financial definitions become a diligence expectation. The reasoning was about who does the asking. FRANdata puts private-equity ownership or backing at more than 12.4% of active US brands, reported via Franchising.com, and those owners do not review reporting once and go away. The grading question: was a brand asked for standardised location-level data by an outsider in the first half of the year?

Grade: too early. The asking is lumpy — it arrives with a transaction, a refinancing or a board change, so half a year of quiet at one brand proves nothing. The threshold itself needs revising: "were you asked" measures deal flow more than expectation. The better question, which I will grade in January, is whether a brand that cannot produce it now loses time it used to be forgiven — the lever set out in increase franchise valuation, where standardising definitions goes first because everything else consumes its output.

Prediction four: the sub-hundred-unit brand gets tools built for it. The threshold: published pricing or a named package aimed explicitly at brands under a hundred units, from two vendors, one of them shipping something an eleven-person headquarters can turn on without a services engagement.

Grade: half, on the checkable part. Named packages for emerging brands are not scarce; a published price an operator can read without booking a call still is, and "starting from" with no ceiling is not a price. What I got wrong was assuming packaging and pricing move together. Naming a segment costs a vendor nothing. Publishing a number costs their discretion in every negotiation after.

The prediction I made against myself

January's fifth was written against my own interest: no franchisor under five hundred units would remove a headquarters role because of AI this year. Not one.

Nobody has proved me wrong, and that is not a win. Headcount plans are private, "we did not backfill" is not a thing brands announce, and the absence of a counterexample from an unpublished population is the weakest evidence in this post. Grade: standing, for a bad reason.

The claim I would still defend is the shape of the gain rather than its size. Where these tools work, the same people cover more locations and answer faster, and neither appears in a headcount plan — which is why the prediction was easy to make and impossible to settle. If you know of a role genuinely removed and not refilled within a quarter, that data point is worth more than the rest of this scorecard.

What moved that nobody predicted

Two, offered as observation, not measurement.

The first is that the governance conversation has changed owner. A year ago the AI question came from a board or a vendor. This year it comes from franchisees asking to be told what is allowed — an easier conversation to start and a harder one to postpone.

The second is who is doing the asking. Nobody has published a usable series on franchisee age and I would not trust one that appeared; what is observable is the questions candidates ask during development, described in millennial gen z franchise owners. An owner who has never worked without these tools does not ask whether they are permitted.

Grade the list whole and it is a bad half-year for my forecasting and a good one for the method. Three predictions I could not settle is itself the result: the state of AI franchising 2026 is a subject with far more announcements than instruments, and every confident percentage you are shown before January rests on the same year-old survey as everyone else's.

Read the list being graded in its original wording: franchise technology trends 2026.