Getting Your Agentic Brief Straight
Groupe Dynamite CTO David Stevens on four years of data work, RFID-enabled inventory intelligence, and why AI strategy only works when it starts with the customer.
David Stevens, CTO of Groupe Dynamite, has a useful filter for deciding whether an AI initiative is worth pursuing. It has nothing to do with cost reduction, operational efficiency, or competitive benchmarking. It comes directly from the founder's desk: does this idea improve the customer experience tenfold?
Groupe Dynamite’s "10x the CX" mandate has shaped every technology decision at the fashion retailer since he took the CTO role. And it's done something more useful than a typical strategic framework: it's made prioritization straightforward in an environment where the volume of possible AI applications is genuinely overwhelming. When everything has to earn its place against a customer experience outcome, a lot of ideas fall away quickly, and the ones that remain have a clear reason to exist.
Before any of those ideas could be acted on, though, Stevens spent four years doing something less glamorous than deploying AI. Over the past four years, while many retailers were racing to deploy AI capabilities, Stevens and his team were doing something less glamorous: getting their data right.
"Bad AI is the result of bad data," he says. Groupe Dynamite built an in-house AI and data team, centralized and governed its data infrastructure, and held off on building customer-facing applications until the foundation was genuinely ready. The CX mandate and the data foundation are, in Stevens' telling, the same investment approached from different directions: one sets the destination, the other makes the journey possible.
Two Brands, Two Completely Different Shoppers
Groupe Dynamite operates two brands, Garage and Dynamite, across Canada, the US, and a recently launched UK market. The clearest early evidence came from the personalization engine Stevens' team built after the data foundation was in place. The results were strong across the board—double-digit increases in add-to-bag rates for one brand—but what was more revealing was what the data showed about the difference between Groupe Dynamite's two customer bases.
Garage shoppers, typically younger, search for specific products. They know what they want—a particular style, a specific color, a singular item they saw an influencer wear on social media—and they come to the site to find it.
Dynamite shoppers search differently. They're looking for solutions: a dress for a beach wedding, an outfit for a first day at work.
The distinction sounds simple, but it has significant implications for how search, recommendations, and content are built and optimized for each brand.
"That insight came directly from AI analyzing behavior patterns," Stevens says. Semantic search has driven significant add-to-bag improvements for Dynamite, where the solution-oriented search behavior is a natural fit. For Garage, the priority is different: trend-setting, product-by-name discovery, specificity. What it means behind the scenes is they can use the same underlying technology and calibrate it differently for each to optimize outcomes. The data made what was once a hidden behavioral difference visible, while the CX mandate created the pressure to look for it in the first place.
The Store Floor as a Data Layer
The same filter that shaped the personalization work led Stevens' team to inventory management, where the customer experience implications of getting it wrong are immediate and measurable.
In the majority of Groupe Dynamite's stores, RFID readers are installed in the ceilings and every garment has an RFID tag, creating a continuous real-time picture of what's on the floor. Which items are getting attention, which are being picked up and put back, which are moving toward purchase. When stock drops below a threshold, staff are alerted to replenish. (After all, if an item isn’t on the shelves, a customer can’t pick it up and purchase it.) This one small shift in how Groupe Dynamite leverages data has driven an estimated 35% decrease in lost sales, according to Stevens.
But the ambition for RFID data extends beyond replenishment, driven by the CX imperative. "We want to know how we can use the visual to tell us more, not just for loss prevention or to boost sales, but to understand the logic for decision making," Stevens says.
Camera data is being integrated to understand customer behavior in more granular terms, with the intent of feeding those insights back into product design. The store floor, in other words, is becoming a research environment as much as a retail one, and the CX mandate is what keeps that research pointed in a useful direction.
AI From the Top Down
When Stevens describes the organizational conditions that have made Groupe Dynamite's AI program work, he's unambiguous about where the momentum originates.
"We have a very forward-thinking board, and a forward-thinking CEO," he says. "The push for AI is coming from the top and it changes the game. You're not fighting against an upstream current that's anxious. You're getting someone who understands the value of AI and has said: I want the business to run on AI."
A clear mandate and a strong data foundation create the conditions for good AI initiatives. Getting an organization to actually generate and sustain them is a different problem, and Stevens has been deliberate about solving it.
Cross-functional hackathons have been one of the primary mechanisms. Teams of five or six people, drawn from call centers, HR, in-store operations, and other parts of the business, work alongside in-house data and AI scientists on ambitious goals— moonshot briefs designed as much to get people working directly with AI tools as to produce viable ideas.
The approach works on both counts: Over 95% of participants said they would continue using AI tools in their daily work after the hackathons, and the ideas that emerged were filtered through the CX lens and prioritized as a group, with the people who generated them kept involved in the process regardless of whether the initiative falls within their line of business.
That last part matters more than it might seem.
"Keeping the people who generated the ideas in the process has massive cultural benefits for keeping things moving," Stevens says. It also surfaces something that purely top-down programs tend to miss: the perspective of people closest to the customer, including, as Stevens notes, super shoppers within the organization's own demographic who don't have design titles, but have exactly the right instincts.
Composability as a Measurement Enabler
All of this—the mandate, the data foundation, the organizational mechanisms—would be considerably harder to act on without an architecture that lets the team isolate and test individual interventions. Composable architecture is what provides that.
"Without a monolith, we can leverage and adapt just the one component we're looking at," Stevens says. AI can be inserted into a specific service or piece of the customer journey without disturbing anything adjacent to it, and the impact of that insertion can be measured discretely. The UK site launched composable from the start, and the North American sites are being migrated to modular structures. The direction is clear, and the measurement benefits are already visible in how precisely the team can attribute outcomes and iterate toward better ones.
Attribution remains complex, and omnichannel customers don't move in straight lines. But the combination of a strong data foundation, a clear customer experience mandate, an architecture that supports precise intervention, and an organization structured to generate and sustain good ideas has given Groupe Dynamite something most enterprises are still working toward: genuine confidence in what their AI is actually doing, and a clear sense of where to take it next.
Leigh Bryant
Editorial Director, Composable.com
Leigh Bryant is a seasoned content and brand strategist with over a decade of experience in digital storytelling. Starting in retail before shifting to the technology space, she has spent the past ten years crafting compelling narratives as a writer, editor, and strategist.