There are many ways that artificial intelligence (AI) can apply to convenience stores. Some are using AI tools to boost productivity at the corporate level. Some are using external partners to innovate and learn.
At CSP’s C-StoreTEC event, which took place in October in Plano, Texas, NexChapter partner Matt Riezman asked three c-store leaders about how they are using and thinking about AI.
Here is what Chris Edwards, director of retail platforms at Circle K; Jesse Wolcott, assistant director of IT at Royal Farms; and Scott Smith, vice president of IT at Parker's Kitchen, shared.
- Laval, Quebec-based Circle K owner Alimentation Couche-Tard is No. 2 on CSP’s 2025 Top 202 U.S. c-store chains by store count. Baltimore-based Royal Farms is No. 30 and Savannah, Georgia-based Parker’s Kitchen is No. 73.
These answers have been edited for length and clarity.
What has been your entry point into AI? What did you choose as first step?
Fyxer.ai. It showed up, the notetaker showed up, and all of my emails started with, “Hi first name,” and my IT manager said, “What did you do? You don't say hi in your emails.”
And then it sort of snowballed into these productivity tools with Perplexity sort of taking off and even the basic version of Copilot really spiraling out of control into something that everybody realized was already on their computer. So, the productivity platforms in the corporate space have really been the first real foothold, and it’s breaking through the barrier of, “this is more than chat GPT.” —Wolcott
You mentioned working with external partners at Parker’s. How are you thinking about the balance when you’re bringing AI in and building in-house versus going out to an external partner?
Historically, Parker’s, when I first got there, we had built everything internally. I can say it was a fire when I got there—having all of that, supporting it, having a staff of five developers who were constantly running around trying to improve things and really nothing was improved. So we're taking a new approach now where it's hybrid.
We feel that we can still innovate very quickly by using external partners. Let them do all the work outside, we'll own it internally at some point, and then we'll kind of innovate upon it once we get it right. But we can still keep our internal teams focused on our stores, on our corporate office, on anything that we're doing project wise for the company and just let those third parties really run.
And sometimes using Georgia Tech, or maybe some other universities out there, is a lot faster because they have a lot of really, really smart people doing the work. And so let them learn, we can learn with them and then ultimately, we can bring it back in house and kind of keep it going instead of trying to build it. —Smith
Operating at global scale, how do you think about prioritizing where to start with AI and what opportunities or challenges does a global perspective create for you?
The prioritization probably isn't any different from a small company really to a large company. Everybody's trying to solve the same problems.
I think the challenges are just, at scale, the amount of licenses that you need just to do something simple, right? If you want to put Copilots on everybody's computer in an organization with thousands and thousands of people, that can be pretty costly.
Somebody was talking before about regulation. So, if you think about all the regulation there is in the U.S., and then you think about all the regulation there is in Europe, and you're operating, you know, globally, it can make it very difficult to make sure that you're transparent and ethical and you understand all the models. You know, it's really easy if you build something internal. You build it internally, you know what you did, you know why you did it. When you work with a vendor, a lot of times we're working with vendors that have kind of a predefined solution.
Getting into that black box and understanding exactly what they built, you know, it's one thing to go through the infosec checklist and you saw compliant and you're this that. It's another thing to really understand, how was the model trained? What does it consist of? How can we get them kind of the smallest chunk of data to suffice and do whatever we need to do? —Edwards
Members help make our journalism possible. Become a CSP member today and unlock exclusive benefits, including unlimited access to all of our content. Sign up here.