Column: Wingstop Built Its AI. Most Brands Can't.

Wingstop's CEO calls his company a technology company that happens to serve chicken wings. After reading the numbers, I'm not going to argue.

Digital sales now make up more than 71% of systemwide sales across more than 3,200 restaurants, Nation's Restaurant News reported Oct. 19. Since the AI-enabled Smart Kitchen system rolled out in 2025, average wait times have dropped about 50%, according to the company, and its customer database tops 60 million people. “Smart Kitchen has [created] an operational transformation I’m not aware of another restaurant brand achieving at this scale,” CEO Michael Skipworth said.

Days earlier, Pokeworks showed a smaller version of the same move. The 86-unit poke chain built an agent called Pokebot on top of a central database, and co-founder Wen Wei told Restaurant Business the company does not plan to add headquarters staff as it grows. “Our goal is to maintain headcount with our future growth,” Wei said.

The tempting lesson is that the winners build their own. I think that's the wrong lesson, and an expensive one for the average franchisor.

Look at who built. Wingstop is a public company that went digital-first in 2021 and has kept investing ever since. Pokeworks has two co-founders with advanced computer science degrees who wrote their own software at an earlier company. That isn't a template. It's an exception. Plenty of franchisors run lean support centers with no engineers on staff, and they shouldn't read these stories as a mandate to hire a dev team.

Now look at Hand & Stone, which we covered when its AI recruiting cut time to hire in half. It didn't write software. It layered conversational AI onto CareerPlug and Paradox, recruiting tools its franchisees already used. Locations using the automation converted 69% of applicants to interviews, versus 16% with manual outreach, the company said in an awards interview.

So what do these three have in common? Not who wrote the code. Each picked a specific job, connected the data that job depends on, and kept people on the decisions that carry weight. Pokeworks consolidated its data in 2025, before it built a single agent. Hand & Stone let each owner set the screening criteria and kept final interviews human. Wingstop uses what it learns from its database to decide what to test next. That's the argument from my last column: the agent race is really a data race.

I'll push back on myself here. Buying has a cost too. If a vendor owns your workflow and your data, you're renting your advantage, and Pokeworks says building let it drop or consolidate other software. Franchisors should ask hard questions about where their data lives and who can take it with them.

My bet is that most systems land in the middle, on tools a franchisor configures rather than codes. Revscale, which publishes this site, works that way: its AI agents are set up in plain language for whatever workflow a business can describe, from answering inbound leads within minutes to taking calls to re-engaging prospects who went quiet, and the activity can flow into the CRM and tools a business already runs. It's one route among several, and Wingstop shows what a full build can do when the money and talent are there.

For a 60-unit brand, the question isn't whether to build its own Pokebot. It's which three jobs franchisees most want off their plate, and whether the data is clean enough for any tool to do them. Answer that first and build versus buy gets a lot easier.

What I don't know: whether AI stays this cheap to run. Wei acknowledged the big labs are subsidizing token costs today. If that changes, the build math changes with it.

Revscale Media is published by Revscale.