The sequencing most retailers get backwards
Morrisons has partnered with Duvo to find where AI can add the most value across its existing operations, and the detail that matters is the order of operations. Duvo's engagement starts by mapping how work actually flows through Morrisons' people, teams, and systems today, before any decision gets made about what to automate. That sounds obvious until you consider how most retail AI programs actually run: a vendor demo creates enthusiasm, a pilot gets funded against a narrow use case, and only after the pilot stalls does anyone ask whether the underlying process was understood well enough to automate in the first place.
Duvo was co-founded by Tomas Cupr, who also built Rohlik Group, one of Europe's larger online grocery operations, which gives the company an unusual vantage point. It has operated the kind of complex, thin-margin, high-SKU-count business it is now advising other grocers on, rather than arriving purely as a consulting or software vendor with no operating scar tissue of its own. That distinction matters to a Group Productivity Director deciding which vendor to trust with a map of the company's internal weaknesses, since a founder who has run a grocery operation is less likely to recommend automation that looks good in a slide deck but ignores the exception handling that actually consumes staff time on the floor.
What Morrisons is actually buying
Gordon MacPherson, Morrisons' Group Productivity Director, framed the engagement in terms every CFO will recognize: the company's processes have developed over decades, and Duvo helps it understand them in enough detail to base automation decisions on evidence rather than assumptions. That is a quietly significant admission from a major grocer. It concedes that Morrisons does not currently have a reliable internal map of how its own operations run, which is a common condition at large retailers but rarely stated this directly in a vendor announcement.
Cupr's framing of what Duvo delivers is equally pointed at the pilot-fatigue problem plaguing retail AI budgets. He argued that AI should carry out work inside the systems teams already use, with human oversight and controls, rather than bolt on more disconnected pilots that each require their own change management, training, and support burden. For a retail CTO sitting on a backlog of a dozen unconnected AI proofs of concept, that is the pitch worth testing against Morrisons' own experience over the next year.
The governance signal: a Chief AI Officer first, then a vendor
Morrisons recently appointed Dr Mohsen Ghasempour as its first Chief AI Officer, and the Duvo partnership follows that appointment rather than preceding it. That ordering matters for any retail organization still debating whether to centralize AI governance before or after committing to specific automation vendors. Morrisons appears to have decided that accountability and decision rights needed to exist first, so that a mapping exercise this broad has a single executive owner who can arbitrate which departments get prioritized and who signs off on what gets automated.
That sequencing is a defensible template for other retailers still running AI initiatives out of whichever department happens to have budget. Without a named executive accountable for the overall AI portfolio, a process-mapping exercise like Duvo's risks becoming another input that different teams interpret differently, producing the same fragmented pilot landscape the engagement was meant to prevent. The Chief AI Officer role gives Morrisons a forcing function that most peer grocers evaluating similar vendors do not yet have in place.
Why this is a CPG and grocery-specific problem
Grocery retail carries process complexity that general merchandise retailers often underestimate: perishability windows, supplier-specific ordering cadences, promotional pricing that cuts across systems built at different points in a company's history, and store-level labor scheduling constraints that vary by format. A generic automation platform applied without first mapping how those processes actually interact can automate the wrong step efficiently, which is often worse than not automating at all, because it hides the dysfunction instead of fixing it.
This is also why a founder with direct grocery operating experience, rather than a pure enterprise software background, is a meaningful credibility signal for this specific engagement. Rohlik's operating history gives Duvo's team pattern recognition for where grocery process maps typically break down, around supplier data quality, store exception handling, and seasonal demand swings, that a generalist automation consultancy would need to learn on Morrisons' dime instead of bringing in on day one.
What CIOs should borrow from this approach
The transferable lesson for any retail CIO is to resist funding automation pilots until there is an honest internal map of the process being automated, including where exceptions happen and who actually makes judgment calls that a flowchart would miss. That mapping work is unglamorous and does not produce a demo-able output the way a new AI tool does, which is exactly why it gets skipped under budget and timeline pressure, and exactly why skipping it produces the graveyard of disconnected pilots Cupr is explicitly selling against.
It is also worth noting what this approach does not promise: fast, visible wins in the first quarter. Morrisons' own language, framed around evidence-based prioritization rather than specific automation targets or cost savings figures, suggests this is a multi-quarter diagnostic engagement before major automation spend follows. CIOs under pressure to show immediate AI ROI should expect that same timeline if they adopt a similar sequencing, and should set expectations with their boards accordingly rather than promising savings before the map exists.
The open questions this partnership leaves unanswered
Neither Morrisons nor Duvo disclosed commercial terms, a timeline for completing the mapping phase, or which business functions get prioritized first, which makes this announcement a statement of intent more than a measurable commitment. Retail leaders watching for signal should look for Morrisons' next disclosure, likely tied to specific automated workflows or productivity metrics, as the real test of whether evidence-based sequencing produces better outcomes than the pilot-first approach it is implicitly critiquing.
The broader test is competitive. If Morrisons can point to measurable productivity gains from this approach within a year, expect rival UK grocers to follow with similar process-mapping engagements rather than more point-solution pilots. If the mapping phase drags without visible output, it will reinforce the opposite lesson: that pilots, however fragmented, at least produce something boards can see, which is its own kind of organizational pressure that evidence-based sequencing has to overcome.
