‍

Part of Bikky's AI Summer School — a series where restaurant operators share how they're using AI day-to-day, what tools they're using, and what it's done for their brand.

‍

In this AI Summer School session, Bikky CEO Abhinav Kapur sat down with Kyle Gordon, founder and CEO of Dilla's Quesadillas, a six-location QSR brand based in Dallas-Fort Worth, TX. Kyle walked through how he built PEX, the Primo Execution Engine, an in-house AI tool that turns raw store scorecard data into specific, prioritized coaching direction for his area leaders and GMs.

Scaling consistent coaching

Kyle has said that he believes the restaurant business is fundamentally about consistency: the same high-quality product and experience delivered every time, the same connective tissue he sees in brands like In-N-Out, Raising Cane's, and Chick-fil-A. Coaching his team toward consistency has always been a priority at Dilla's.

To do this, Kyle built a scorecard where every location's KPIs were logged each week.

‍

But the scorecard alone couldn't show whether a location was trending up or down, and by the time something needed attention, it was often a week or two old. Coaching defaulted to whatever seemed most visible that week, applied the same way across every location.

‍

Kyle didn't want AI to replace Dilla’s established coaching practices, but he needed a way for his operators to catch what mattered sooner and spend more of their time actually working to improve it.

Creating Dilla’s Primo Execution Engine

To address the coaching challenge at scale, Kyle used AI to create the Primo Execution Engine, which he calls  PEX. PEX reads through each location's weekly scorecard data and generates a coaching brief telling area leaders exactly what to focus on.

‍

Dilla's had spent close to a decade building out detailed SOPs and a framework Kyle calls MOM: Manpower, Culture, Operations, and Marketing, in that priority order. Any time something looked off at a location, Kyle's instinct was to check manpower first, then culture, then operations, in that sequence, before touching marketing. That framework became the diagnostic filter PEX runs on: when a location's numbers dip, PEX works down the same sequence Kyle would use himself.

‍

The years of careful documentation Dilla’s kept meant PEX didn't have to be built from square one. Kyle built the first version in a single weekend, then spent another 10 to 15 hours refining the technical setup and connecting PEX to the right data sources.

The three pieces of PEX

PEX is built from three components. The first is a document Kyle calls the brain doc, roughly 16 pages, written in plain language, containing the coaching philosophy, the MOM framework, and the specific voice PEX should coach in.

‍

Getting that voice right took editing. Kyle found that AI models default to soft, hedging language, telling a GM something like "it'd be really nice if maybe you could drop your labor a little," when the way he'd actually coach someone in person is closer to "labor is high because of improper cuts, fix it."

‍

The second component is the script. It’s code that pulls each location's weekly scorecard data from a Google Sheet, checks it against the brain doc, and generates a written brief. 

‍

The third is the output itself, a short email brief per location, plus a separate rolled-up summary for leadership so they don’t have to read six pages per unit to get the full picture. PEX runs once a week and costs Dilla's around seven dollars per run, on top of roughly $300 to build.

PEX in action

The clearest example of PEX's impact came from one of Dilla's area leaders. Before PEX, he visited his locations on the same fixed rotation, regardless of which ones needed attention that week. Now his visit schedule is built around which locations get flagged. A location with no issues might not get a visit at all that week, while one with three or four flagged metrics might get visited multiple times.

‍

In his own words: "It changed how I build my week. I stopped visiting stores in an arbitrary order and started visiting them based on priority. I can now quickly identify trends and stop them before it negatively affects a period."

‍

PEX also surfaces wins, not just problems. If a location has had perfect retention for six straight weeks, that shows up in the brief too, so a coaching conversation doesn't always start with what's wrong.

‍

Since rolling PEX out, Dilla's has seen sales and traffic trend up and drive-through times come down. Kyle was careful to note he can't fully credit PEX alone for that, but he pointed to something harder to measure and just as real to him: the restaurants feel calmer and more consistent day to day, something he says customers can feel too.

Conclusion

When it comes to getting started with AI, Kyle's advice is to find a process in your business you've never documented because it always felt too hard or too big a lift, and build there first.

‍

PEX took a single weekend to build because the hard part, years of SOPs and the MOM framework Kyle used to run Dilla's, was already written down. Restaurants tend to be documentation-heavy by nature, largely because of how much staff turnover the industry deals with, which means most operators already have more of what AI needs to work well than they realize.

‍