UiPath and The Very Group Sign Three-Year Deal to Price 200,000+ Products with Agentic AI
UiPath has signed a three-year partnership with UK digital retailer The Very Group to run agentic AI pricing across a range of more than 200,000 products. Here is the Peak acquisition behind it, why explainability leads the pitch, and what regulators are watching.
Key Takeaways
- On 15 July 2026, UiPath announced a three-year partnership with UK digital retailer The Very Group to run pricing and merchandising decisions across a range of more than 200,000 products using agentic AI. Contract value and go-live timing are undisclosed.
- The pitch leads on AI explainability rather than speed. Pricing sits under audit and regulatory scrutiny, and an agent whose reasoning cannot be traced has no place on a retail floor.
- Buy-side AI comparing prices and sell-side AI setting them are converging. Merchants need to draw the line between what an agent executes and what a human approves before the tooling arrives, not after.
What is actually in the three-year deal
UiPath has announced a three-year partnership with The Very Group to revamp its pricing strategy using agentic AI. The pair will work together to implement an AI powered, agentic pricing solution that aims to bring faster and more transparent decision-making across the latter's retail brands.
retailtechinnovationhub.comAutomation vendor UiPath announced on 15 July 2026 that it had signed a three-year partnership with The Very Group, the UK online retailer behind Very and Littlewoods. Per the announcement carried by The Retail Bulletin, UiPath won a competitive pitch and will implement an AI-powered agentic pricing solution across the group's retail brands.
The construction is worth reading closely. UiPath layers clear AI explainability onto the data-led pricing strategy The Very Group already runs, then adds campaign simulation and optimisation plus scenario planning on top. Manual steps come out so that staff can spend time on more technical, analytical work. This is framed as a replacement for a decision process, not as price automation.
We have a range of over 200,000 products and pricing is one of the most powerful levers in retail.
The stated benefits are optimised gross margin and stock management, plus pricing agility in a competitive market. The commercial terms, though, are almost entirely absent. Contract value, go-live date, the exact brand scope and any target KPIs are all undisclosed. As far as I could verify, UiPath's own newsroom carries no standalone release for this deal; distribution ran through UK retail trade press.
What 200,000 SKUs means, and why pricing is the strongest lever
Calling pricing the strongest lever in retail sounds like a platitude from outside the industry. The Very Group's own numbers make it concrete.
In its full-year results for the 52 weeks to 28 June 2025, group revenue fell 1.8% to £2.09bn. Adjusted EBITDA nonetheless rose 15.9% to £307.1m, delivering a 14.7% EBITDA margin, the highest in the group's history. Gross margin improved 1.0 point to 36.6%. This was a year of growing profit without growing sales, which the company describes as prioritising profitability over volume.
For a business shaped that way, pricing is close to the only source of incremental profit. Once you decline to chase volume, gross margin and markdown precision become the P&L. A takeover by Carlyle was announced in November 2025, which does not reduce the pressure on margin.
Then 200,000 products introduces a hard physical limit. Even at three seconds per item, a single pass across the range runs to more than 160 hours of work, and real decisions require competitor prices, cover days, seasonality and promotional overlap. In practice, most SKUs simply keep last cycle's price. What agentic AI is attacking here is throughput before it is accuracy.
UiPath's own July 2026 survey of 500 UK retail leaders found 69% only respond once an issue has already hit commercial performance, with only 19% able to react to pricing or inventory disruption in real time. Vendor-commissioned research deserves discounting, but the finding that 47% have deployed AI and are still waiting on measurable commercial impact is telling.
The Peak acquisition that made this possible
Why is an RPA company setting retail prices? The answer sits a year upstream.
UiPath acquired Peak, an AI-native company out of Manchester, in March 2025. The acquisition release states plainly that Peak's inventory and pricing optimisation platform would form the backbone of new Pricing and Inventory Agents for UiPath customers. Deal terms were not disclosed.
Peak had spent more than a decade applying AI to commercial decisions with retailers including Nike, Debenhams Group and The Body Shop, covering pricing, promotions, inventory and markdowns. UiPath characterises the combination as the shift from AI that advises to AI that acts, connecting prediction, decision and execution into one continuous loop.
Read that way, The Very Group is not a one-off partnership. It is the proof point for a retail AI business UiPath has been assembling since the acquisition, won roughly sixteen months after closing.
How this differs from a conventional pricing engine
Price optimisation software is not a new category. Revionics, Competera, Pricefx, Blue Yonder, PROS and Eversight have competed here for years, and most already embed machine learning. "AI-optimised pricing" on its own is not a differentiator.
The difference shows up in the shape of the output and in where accountability lands.
| Dimension | Conventional price optimisation | Agentic pricing |
|---|---|---|
| Output | Surfaces recommended prices | Runs decision through to execution |
| Review cycle | Weekly or campaign-based cycles | Continuous review as demand shifts |
| Human role | Receives recommendations, applies them manually | Shifts to policy setting and exception approval |
| Central question | How accurate is the model | Can each pricing decision be traced afterwards |
| Scope | Tends to stay inside the pricing module | Runs on shared goals with stock, promotions and planning |
Rows three and four carry the weight. In the conventional pattern a merchandiser receives a recommendation and decides whether to accept it, so accountability naturally stays with a person. Once execution is automated, nobody can explain a price unless the product itself retains a traceable record of why that price was set. Explainability leading this announcement is not marketing decoration; it follows from the structure.
Explainability is not only an internal requirement
If explainability were only about internal audit and sign-off, this would be simple. Regulators are looking at the same territory.
The UK's Competition and Markets Authority published its view on algorithms, AI and collusion in March 2026. Alongside explicit agreements, it sets out hub-and-spoke coordination through a shared algorithm or data hub, "predictable agents" that soften competition by reacting foreseeably to rivals, and the possibility that advanced AI told to maximise profit learns its way to coordinated outcomes with no human intent to collude. The CMA is explicit that using an algorithm is not a shield from liability. In late February 2026 it opened an investigation into information exchange between hotel chains via a data analytics tool.
US attention runs on a different track, focused on surveillance pricing built from personal data. New York's Algorithmic Pricing Disclosure Act now requires an all-caps notice that a price was set by an algorithm using the shopper's personal data. California's Attorney General has begun issuing inquiry letters to online retailers, and the FTC published findings from its surveillance pricing study in January 2025.
In fairness, The Very Group's project is described as SKU-level optimisation, not per-shopper pricing, so those specific rules do not map directly onto it. Even so, the faster, broader and more frequent automated pricing becomes, the higher the risk of unintended coordination and of opacity as consumers experience it. The vendor's phrase "more transparent decision-making" is a promise still to be demonstrated, not a state already achieved.
Buy-side agents meeting sell-side agents
For readers of this site, this is where the deal matters most.
Agentic commerce discussion has skewed heavily toward the demand side: a consumer's AI finds, compares and buys. In doing so it collates prices across merchants at machine speed. The friction that used to protect merchants, the shopper who could not be bothered to look further, disappears at that point.
A merchandising team working weekly cycles cannot answer that pressure. So the supply side agentifies too. Placed in that context, The Very Group's decision reads as a response rather than an experiment: demand-side agents are pulling supply-side agents into existence.
Nobody has observed a market with both sides agentified at scale. At the level of theory and simulation, the research the CMA cites points toward coordinated equilibria. The opposite case is also coherent: with comparison friction gone, competition intensifies and prices fall. Which way it breaks likely depends on how many participants there are and how diverse their algorithms remain.
What merchants should prepare now
The scale gap is not a reason to dismiss this. Thinly staffed merchants are, if anything, further behind on pricing throughput than a group with 200,000 SKUs and a dedicated commercial function.
Three things come before tool selection. First, write the pricing policy down. Floor margin, whether to follow competitors, maximum discount depth, maximum repricing frequency. Constraints you cannot state in plain language cannot be handed to an agent either. Second, design the approval boundary: which price bands and which magnitudes of change execute automatically, and where a human signs off. Note that the announcement describes staff moving to more analytical work rather than disappearing. People are being redeployed to designing judgement and handling exceptions, not removed from it.
Third, keep records. When, on which inputs, and for what reason a price moved. Without that history there is no internal explanation, no supplier explanation, and no future regulatory answer. Explainability is a requirement to specify up front, not a feature to bolt on later.
One step beyond that, it is worth checking how your prices look to a buy-side agent. A store without structured product and price data loses before the comparison even starts.
Conclusion
Pricing is the retail decision that translates most directly into profit, which is why UiPath bought Peak and why The Very Group committed three years. With neither deal size nor go-live disclosed, the results remain to be verified.
The thing to watch over the next year is whether concrete numbers emerge: gross margin uplift, stock turn, and above all what share of agent-set prices humans overrode. That last metric tells the honest story of whether agentic decision-making actually took hold on the floor.
