Column: Franchisees Aren't Waiting for Corporate's AI

Ted Taylor spent 25 years in semiconductors before he bought into restaurants. Now he runs three co-branded Schlotzsky’s and Cinnabon units in New Mexico, and he reprices his menu in an hour or two.

It used to take five days. That figure, and most of what follows, comes from a case study OpenAI published Oct. 20, so read it as a vendor’s showcase. Still, the workflow is specific enough to take seriously. Taylor gives ChatGPT Work recipe costs, competitor prices and the different economics of in-store and delivery orders, and gets back item-level price recommendations. He gave it read-only access to sales and labor data so it can flag overtime and coverage gaps. Monthly cash reconciliation across three stores fell from three hours to 5 to 10 minutes. One store manager now gets a draft weekly schedule that she checks before entering it.

“ChatGPT Work’s recommendations are so good that it’s much faster to start there and then make adjustments,” Taylor said in the case study.

Here is my take. This is the most interesting AI story in franchising this month, and not because of the hours saved. It matters because nobody at the franchisor built it.

Most of the AI news we cover runs top-down. Bojangles is taking its drive-thru order taker systemwide. Wingstop built its own stack, which I argued most brands can’t copy. Pokeworks built an internal agent and stopped hiring at headquarters, and its franchisees are not using it directly yet. Those are franchisor projects, handed down the system.

Top-down has a failure mode, and it is ugly. Chaac Pizza Northeast, a Pizza Hut franchisee with 111 restaurants, sued the brand for $100 million this year, alleging that a mandated AI delivery dispatch rollout pushed rack times from under five minutes to as much as 20, Franchise Times reported. Pizza Hut denied the allegations. Whoever is right, the lesson holds: when corporate picks the tool, corporate owns the outcome.

Taylor’s version flips that. He picks the task, compares the output with how he would do the work himself, and widens the scope only when he trusts the results. The first time he used it for pricing, he ran the analysis by hand in parallel. That is not a technology strategy. It is an operator’s strategy, and franchisors should be studying it.

Where I would push back: the back office is the shallow end of the pool. A bad schedule draft gets caught by a manager. A bad answer to a customer on the phone does not get caught by anyone. That is where franchisee-built AI gets riskier, and where brand standards legitimately matter.

Full disclosure on that front: Revscale, which publishes this site, builds AI agents for the customer-facing side, such as answering calls, qualifying inquiries and following up with leads that went quiet. They are configured in plain language, so an operator can describe a workflow instead of coding one. I would hold any of those tools, ours included, to Taylor’s standard: run it next to the human version first, and keep a person on the handoff.

Franchisors have a choice. They can treat local AI experiments as a compliance problem, or as free research and development. The brands that learn to bless what their best operators are already doing, with guardrails on anything that touches the customer, will get further than the ones still designing the perfect systemwide rollout.

What I don’t know is how many Ted Taylors are out there. One case study is one operator. My guess, and it is a guess, is that the number is bigger than most franchise development teams assume, and it grows every time a payroll run gets four hours shorter.

Revscale Media is published by Revscale.