Getting started with artificial intelligence at your convenience store chain doesn’t have to be complicated.
Just ask Babir Sultan, founder and CEO of FavTrip, with three c-stores in Missouri, who has found success implementing several artificial intelligence (AI) agents to handle tasks at his stores.
His advice to other retailers is to think in terms of “problem first, not tool focused.”
In other words, begin by identifying the problem to be solved. At FavTrip, the chain’s team members are instrumental in identifying the best problems to address for two reasons. One, they know best which tasks are most time consuming, and how they could better spend their time. Secondly, it’s important that team members are onboard with the ways AI is being implemented.
“You have to have a buy-in with your staff,” Sultan said. Sultan starts by speaking with various employees, from those stocking coolers to those placing orders, and asks them to list their current headaches and rate them on a scale of one to 10. He also discusses what else they would be able to focus on instead if a particular pain point was removed. Then he looks at which of those problems AI can realistically solve, and then he determines which problem he’s going to address first.
“Keep your expectations low because I know a lot of people go in guns blazing where they say, ‘I’m going to solve all the problems with AI, and they come out disappointed,” he said.
Sultan has quickly progressed with AI by methodically mastering one level of AI ability before adding more complicated tasks.
He recently used a three-step workflow to solve issues around grocery ordering at the chain, which he shared in a recent CStoreTalks podcast for other retailers to consider. When describing the problem he needed to solve, Sultan shared a story many retailers can relate to.
“There is an employee, Jeff, that always did the grocery order,” he said. “Jeff leaves. He took the data in his head and left. Jeff was the guy that could look at the order and tell you exactly how many Snicker bars we needed. He was just that good, and it’s amazing to have a Jeff for everybody, right? …. But when he leaves or takes a vacation … everybody’s scrambling. … For us, that was a problem, right?”
FavTrip took the grocery ordering tasks that Jeff handled, passed them to an AI agent and broke the tasks down into three progressive workflows.
Workflow One: Analyze Data
As part of Workflow One, Sultan selected an AI agent — in this case Claude — and put it to work analyzing data pertaining to grocery ordering on a regular basis. Soon, the AI agent could figure out things like how many Snickers bars, for example, it needed to order each week.
Workflow Two: Prepare the Order
Once the AI agent had mastered Workflow One, meaning it had downloaded the data, analyzed it and was able to identify how many of X item the store needed, Sultan was ready to progress to Workflow Two. For the second workflow, Sultan set the AI agent up to prepare the actual order for him.
Sultan tells the AI agent, “Now tell our vendor how many Snicker bars I need to order this week, so I make sure I’m never out.”
“We thought only the big players of 2,000 stores could do this. Me, as a small operator, we did this last week,” he said.
Workflow Three: Send the Order
Once Sultan was sure the AI agent was properly calculating needs and preparing the order appropriately, it was time to take the training wheels off and move to Workflow Three: letting the AI agent actually place the order.
“We’re still to the point where we want the human to review (the order),” Sultan explained. “This was the first week on Sunday (Aug. 23) that it actually made a $18,000 order for us, and each manager of the store reviewed it and said, ‘This is perfection.’”
“Small operators, we have our limitations, but I never let that stop me from being curious,” Sultan said, adding that he’s not going to wait for his vendors to develop these tools, when he can begin finding solutions for his biggest pain points now.
“This has been a game changer, honestly, for us. And we’re not going to stop here,” he said.
So far, FavTrip has been focused on using this three-step workflow specifically for grocery ordering, but that’s about to change.
“We’re going to take it a step further into drinks now. Drinks per brand,” he said.
Shifting Employee Duties
Importantly, when FavTrip’s team members were introduced to the three-step workflow, they were impressed, he said.
“They have a buy-in on it. They have to buy into this and say, ‘This is good for us vs. not good for us,’” he said. That’s why transparency with team members and including them in the process is so important. He noted that if you tell team members, “’Hey guys, we’re going to take this away from you. You don’t have to worry about it,’ then they feel like you’re taking away their job,” he cautioned. Instead, it’s crucial to come in with a plan for how you can shift their responsibilities. “We never came up short of more work to do in the store,” he said.
For example, FavTrip gets about 280 job applications per month and, as a small chain, one employee was messaging applicants for open interviews, but replying back to everyone was challenging due to the volume of applications. “She told us, ‘Hey, if you could take this away from me, I’ll work on the shoplifting videos, scripting, voiceovers and design for the thumbnails,” Sultan said, referencing the chain’s popular YouTube shoplifting videos that have gone viral.
“We freed up her time to do this,” he explained. The employee was able to switch to a task she enjoyed more, and an AI agent took control of messaging job applicants.
“It’s not complicated at all,” he added. “I encourage a lot of operators, and that’s the reason we’re talking here is to encourage operators to leverage AI to solve their problems.”
For more examples, advice and the inside scoop on how one of Sultan’s AI agents caught a $25,000 error, listen to the CStoreTalks podcast with Sultan HERE.
