AI Rewards Conviction
ALDO Group CIO Matthieu Houle on three-tier AI bets, why supply chain beat the front end, and the adoption problem nobody talks about enough.
When you become CIO of a major North American retailer, you expect some technology challenges. It’s practically in the job description overview: “As CIO, you’ll unearth (and be responsible for!) some weird and probably hairy technology stuff.”
What you don’t expect to find is a parallel universe built entirely in Excel and email, running quietly alongside running quietly alongside heavily customized ERP systems that didn’t fully meet user needs. But that parallel universe is exactly what Matthieu Houle found when he became CIO of ALDO Group.
Discovering that the official infrastructure existed largely in name while the real work happened in spreadsheets shaped everything that followed for Houle. It told him where the genuine inefficiency lived, why the front end wasn't the right place to concentrate ALDO's biggest AI bet, and what "solving a hard problem" actually means when you're deciding where to put conviction behind it.
For Houle, the question of what will actually move the needle versus what will get commoditized before you've finished building it has driven every significant AI decision his team has made.
Three Bets, Very Different Stakes
To sift through the commoditized choices and surface the needle-moving investments, Houle developed a straightforward framework for thinking about AI investment. Three levels, clear distinctions, and a tangible way to see where real value lives.
The first level is AI for individual productivity, using tools like ChatGPT and Copilot that make day-to-day work faster and easier. Houle is clear-eyed about what this tier represents: table stakes. "It's just about staying relevant and current," he says. The main KPI is employee usage, although there isn’t always a clear financial return in the traditional sense. You invest here not because it differentiates you, but because not investing means falling behind.
The second level is AI that moves beyond the individual to team-level automation. Squads organized around specific business functions, working through a prioritized backlog of use cases with estimated business value attached to each. This tier produces real efficiency gains, but they tend to be temporary; competitors can (and will) replicate them. The advantage compounds only if the organization is building the capability and the muscle memory to keep moving.
The third tier is where Houle gets genuinely animated: the big bets. AI for meaningful org change. These investments tend to be riskier and more capital-intensive, requiring a full business case with the same investment metrics you'd apply to any major capex decision. But they’re also potentially differentiating in ways the first two tiers can't be.
"AI will reward conviction," he says. The organizations that commit to a genuinely hard problem—not just an obvious one—and build the organizational and technical infrastructure to solve it are the ones that will be difficult to catch.
The framework also has a sequencing logic that's easy to miss.
"It'll take you a lot longer to get to the big bets if you don't have level one in place," Houle says. Individual adoption isn't just a nice-to-have precondition, it's the foundation that makes everything above it possible. Which means the organizations racing past individual and team-level programs to get to the exciting stuff are, in his view, skipping steps they'll eventually have to go back and build.
The Adoption Problem Nobody Talks About Enough
Arguably the most practically useful thread in Houle's MACH X session wasn't about technology or investment frameworks. It was about getting people to actually use the tools, and why that problem is harder and more important than most AI roadmaps acknowledge.
Executives have read about AI, attended briefings, approved budgets. But reading about AI and using AI in context are genuinely different experiences, and the gap between them is where a lot of programs stall. "The trick is to get them to use it," he says. Houle tracks usage the way a consumer product team tracks engagement—daily, weekly, and monthly active users—because if the tools aren't being used, nothing else matters.
The approach that's worked best at ALDO is direct exposure. Leadership team sessions where the people who've used the tools most become the de facto instructors. Running structured proofs-of-concept with business stakeholders and franchisees, and showcasing the results. Getting the founder and executive team hands-on early, because their visible engagement changes the organizational signal. ALDO’s approach is part of a pattern that emerged at MACH X. From the cross-functional hackathons at Groupe Dynamite to the Copilot rollout at Holt Renfrew, the act of using AI is itself an adoption strategy. And the familiarity that engagement breeds is a prerequisite for the kind of critical engagement that makes AI programs actually work.
"I'm trying to get my team to think more about the solutions using AI, not AI as a solution itself," Houle says.
The distinction is subtle, but crucial. Organizations that treat AI as the answer to a question they haven't fully asked yet are the ones producing the pilots and POCs that never make it to production. The ones that start with a hard problem, like a specific inefficiency, a measurable gap, or a decision that needs to be made faster, and reach for AI as the toolkit are the ones building something that compounds.
Why Supply Chain Beat the Front End
Houle came to the CIO role from the digital and composable commerce side of the business, which gives him an unusually clear view of why front-end investment has become a less compelling place to concentrate AI effort.
"With an all-in-one solution, you get the feature set out of the box," he says. "You find yourself in a place where you're trying to keep up with the features." The front end has been commoditized. Whatever advantage you build there, a competitor can replicate it quickly, often by switching platforms.
The back office is a different story. ERP systems, supply chain planning, demand forecasting— these are areas where the tooling is expensive, the processes are manual in ways that would surprise outsiders, and the financial impact of even marginal improvements is significant.
"When fashion was two seasons a year, manually handling two high-volume weekends was workable," Houle says. "Now, with the pressures of a global market, you need to make those decisions way more often. It’s constant, and you can’t manage it using the old ways."
Agentic systems, he argues, are purpose-built for exactly this kind of high-frequency, data-intensive decision-making.
What struck him most when he started building in supply chain wasn't the technology, it was the gap between the systems nominally in place and what people were actually doing.
"A lot of the technology they were using was not helping them, it was a barrier," he says.
The ERP existed, the data existed, and the workflows had been rebuilt in spreadsheets because the official system was too rigid to use well. AI didn't introduce a new problem so much as it made an existing one addressable.
The build-versus-buy logic followed directly from this. Anything that could be commoditized—basic content generation, standard productivity features likely to be absorbed into existing platforms—was a buy. Supply chain was a build, specifically because it can't be installed and activated. "It has to be built with the retailer and the brand, and accepted by the brand," Houle says. That specificity is precisely what makes it defensible as a long-term investment.
The parallel universe in Excel that Houle found is, in some ways, the most honest map of where AI investment should go. Not toward the things that look impressive in a demo, but toward the places where the work has already been done manually for years because the official systems couldn't keep up. Those are the problems worth owning. And for organizations willing to commit to them—fully, with a real business case and the patience to build rather than buy—the returns compound in ways that table-stakes productivity tools simply can't match.
That's what Houle means when he says AI rewards conviction. It's not a motivational claim; it's an architectural one.
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.