I’ve been hosting retail technology conversations for a long time now, and every so often someone says something in the first two minutes that completely reframes how I think about a topic. That’s exactly what happened when I sat down with Tomáš Čupr, founder and Group CEO of Rohlik Group, the pan-European e-grocer doing roughly $1.5 billion in revenue, and now also the founder and CEO of a startup called Duvo.
I introduced the episode by saying that AI can not only help improve retail operations, but also help determine where those operations stand to benefit from AI in the first place. Tom chuckled. And then he said, “I think you said it perfectly.” That chuckle told me everything. Because what sounds like a clever reframe is actually one of the most important and overlooked ideas in retail AI right now, and Tom is building a company around it.
Let’s Start By Getting Something Off Our Chests
Here’s the uncomfortable truth Tom laid out early in our conversation: even as the CEO of a major retailer, he didn’t actually know his own processes.
Not the real ones.
“I realized that not only do I not know the real processes,” he told me, “but people below me do not know the real processes. And the only people that know the real process, but only part of it, are the specialists doing the process.”
That hit me hard, because I’ve seen it firsthand at multiple organizations. There’s a famous distinction in business school between the explicit and implicit parts of any process, i.e. the things that are written down and the things people just know from doing the job every day. The specialists on the floor hold that implicit knowledge. But the people with the authority and budget to invest in automation? They’re often working from a mental model of how things run that is either years out of date or based on their own personal assumptions about how processes should work.
So, ultimately, you end up with a maddening disconnect. Leadership wants to automate, but doesn’t know what to automate. Consultants are called in to close that gap, at great expense, over many weeks, and then that knowledge often walks out the door when the engagement ends.
Tom’s answer to that is what he calls Duvo Clarity, and it’s a genuinely novel approach. One also that I think I must have mentioned to over 20 people at Shoptalk in casual conversation last week, too.
Democratizing What Consultants Used to Own
The way Duvo Clarity works is almost deceptively simple. Instead of a consulting team interviewing a curated set of stakeholders, AI agents, with voice capability, conduct structured interviews with every person in the organization. They probe into the exceptions, the handoffs, the systems, the risks. All the stuff that normally gets lost or filtered when it travels up the org chart.
“Agents are cheap,” Tom told me. “You can now interview every single person in your organization.”
I love that because It takes the bias out of who’s doing the explaining.
In my experience, process mapping exercises tend to reflect the loudest voices in the room, or the most charismatic person who got the most time with the consultants. Duvo interviews everyone and lets patterns emerge from the data. In less than a week, leadership gets a map of what the organization actually does, not what it thinks it does.
Tom shared a real example from Rohlik that made this concrete for me. His team believed that when a supplier missed a delivery window, someone was following up. When Duvo mapped the actual process, it turned out nobody was making that call. The SOP said it should happen, but it wasn’t happening.
Now, a Duvo agent makes that follow-up call automatically, within a minute of a missed delivery window, at 100% SLA. The problem wasn’t technology. It was visibility into even knowing there was a problem to begin with.
The Examples Are Endless
The other application Tom described that genuinely blew my mind was what he called should-cost analysis. This is the practice of analyzing whether a supplier’s proposed price increase is actually justified, breaking down commodity prices, labor rates, logistics costs, packaging trends, etc., and then negotiating from a position of real data rather than just accepting or pushing back based on gut feel.
The funny thing about this example is that I didn’t know what should-cost analysis was before this conversation. Tom admitted he had no idea what is was the first-time he heard it either.
And he ran a $1.5B retailer for 12 years!
That’s the point. Most retailers aren’t doing this analysis at all, and even those that are can only cover their biggest suppliers because that’s their bandwidth.
The long tail?
Tom put a number on it, and he estimates that for 95% of retailers, should-cost analysis simply isn’t happening.
With Duvo, that changes. Every time a supplier sends a price list, an agent can analyze whether the proposed increase is justified against objective market data and surface a recommendation, for every SKU, every time. Retailers already have people who could act on that information. What they’ve been missing is the process.
The Future of Retail Leadership
My conversation with Tom also got a little mind-bending toward the end, in the best possible way, I might add. Tom and I got into what all this means for the next generation of retail leaders, and I walked away thinking about the future in ways I never had before (and hence why I couldn’t stop blathering on at Shoptalk).
The future Tom describes isn’t one where people disappear from retail operations, but rather one where the nature of what people manage fundamentally changes. In his words, “everybody becomes a coach or an agent operator, as opposed to an individual contributor trying to make sense of an Excel file.” And the “blast radius” (his term, which I thought was perfect) of what each leader is responsible for gets dramatically wider.
That’s a big ask. It means leaders will need to understand the work deeply, not just manage the people doing it. Because giving agents clear, effective instructions requires actually knowing the process. The leaders who will thrive in this environment are generalists with real operational depth as opposed to the narrow specialists whose slice of the workflow gets automated away first.
Stop Waiting on Clean Data. Start With Process.
If you’re a retail executive who’s been told you need to get your data house in order before you can really leverage AI, Tom has a direct message for you. Tom thinks that’s becoming less and less true, and waiting on it is a trap.
“I’ve seen organizations cleaning data for the last decade and still being six months away from being data clean,” he said. “And in six months they will still be six months from data clean.”
AI today can make sense of messy, imperfect data far better than previous technology. The real starting point isn’t data, in Tom’s mind. t’s process. Understand what you’re actually doing, identify where the gaps and inconsistencies are, and then your automation options open up. Every vendor, every platform, every technology partner is going to ask you the same first question anyway: what’s the process?
The Bottom Line
The most important AI conversation you can have as a retail leader right now is about whether you actually know how your organization works, not which fancy new tech solution you should buy. Tom Čupr is a needle in a stack of needles, a founder with real operator credibility who has lived the exact pain he’s now solving. And his core message is one I think every retail executive needs to hear.
Let AI show you where you stand before you decide where to go.
To hear the my full conversation with Tom, watch or listen to this episode of the Omni Talk Retail Technology Spotlight Series wherever you get your podcasts:
Apple Podcasts | Spotify | SoundCloud | Amazon Music | YouTube
Be careful out there,
– Chris, Anne, and the Omni Talk team
P.S. See our past 8 years of wonderful Spotlight Series podcast guests, featuring roughly 200 movers and shakers in retail, by clicking here
Music by hooksounds.com
*Sponsored Content*
Omni Talk® is the retail blog for retailers, written by retailers. Chris Walton founded Omni Talk® in 2017 and have quickly turned it into one of the fastest growing blogs in retail.