Key takeaways

• The first-order effect is higher productivity across existing retail infrastructure, translating into more efficient demand capture and margin accretion.

• Stores will become more experience-led, leaning into discovery, immediacy and the social value of in-person shopping.

• Agentic commerce introduces future upside risk to online sales penetration, but within a more mature omnichannel network that will limit disruption to physical retail.

• Stronger margins support retailers’ ability to sustain rents in competitive prime locations, without relying on higher sales volumes or channel reallocation.

As AI rewires the way European consumers search, evaluate, and transact, we assess how improvements in sales capture efficiency and margin performance will interact with the growth of e‑commerce, and what this means for the future role of the physical store.

The story so far

Since late 2023, generative AI has trickled into Europe’s consumer decision journey: 38% of consumers across France, Germany, and the UK now use AI to inform shopping choices (McKinsey). Yet, despite widespread use on the demand side, this has not materially altered channel choice. The share of European retail sales conducted online has risen only marginally, from 13.4% to 14.1% over the past two years; slower growth than the trajectory of the 2010s, while remaining below the pandemic peak of 15.2% (GlobalData).

Rather than drawing spend away from physical into digital channels, the first‑order impact has been to raise the productivity and efficiency of existing infrastructure: lifting conversion, reducing returns, tightening labour and inventory efficiency, and improving fulfilment, without necessarily creating new demand. From a consumer standpoint, AI is influencing product choice far more than channel choice at this stage.

If AI is not yet influencing where people shop, where is it showing up instead?

Efficiency in operations and online product discovery

AI is enabling retailers to generate the same turnover with fewer inputs. Footwear retailer FLO, for instance, cut lost sales by 12% and increased revenue by 5% through AI‑driven allocation and replenishment, achieving higher availability without raising inventory levels (invent.ai).

Superior AI-led inventory mapping and demand prediction also enhances in‑store click‑and‑collect, reducing failed picks and substitutions while sharpening the store’s function as a fulfilment node.

At the same time, we’re seeing product discovery shift upstream into AI‑mediated environments, recasting retailer websites from primary entry points into execution layers. AI‑led search doubled from 4% to 8% of organic visits between 2024 and 2025 and is on track to overtake traditional search by 2028 (Boston Consulting Group). Fewer visits can therefore sustain the same turnover, lowering acquisition effort and digital infrastructure intensity while increasing the value extracted per visit.

Addressing conversion, cart abandonment and returns leakage

While traditional e‑commerce converts just over 3% of visits, according to RepAI, AI‑driven conversational commerce can convert over 12% through guided decision‑making and real‑time intervention at moments of hesitation, translating existing demand into realised revenue. This matters given the scale of unmet demand: UK retailers lost £34.4 billion to online basket abandonment last year (Retail Economics).

Virtual try‑on technology offers another lever, especially given that poor fit accounts for 60–70% of apparel returns (McKinsey). Following the rollout of virtual fitting, Zalando lowered jeans returns by 40% across 14 markets. This translates into fewer failed transactions, healthier margins, reduced fulfilment pressure and lower space intensity across operations.

So, who captures the surplus?

AI investment is centralised, scalable, and, in many cases, cheaper than legacy marketing or labour – although arguably priced to encourage rapid adoption at this stage. While consumers will see some benefit, the more consequential channel is tenant affordability; stronger margins allow healthier rent cover and lower default risk. For landlords, these margin improvements may matter more than topline sales growth, helping explain why occupier health can increase even as the share of online sales rises. That link between AI-driven productivity and tenant affordability is central to the story.

Competitive retailer dynamics

How these cost efficiencies are competed over is equally important, with implications for pricing and deflation across retail operations. By lowering customer acquisition costs, reducing return rates and optimising inventory, AI affords retailers the choice to either protect margins or contest more aggressively on price. In highly competitive markets, the latter is more likely, as efficiency gains are competed away, reinforcing downward pressure on prices and supporting demand.

Meanwhile, digital scale may become a less decisive source of competitive advantage. Historically, digital retail has favoured larger players with superior data, marketing and logistics competencies. AI has the potential to disrupt this dynamic by democratising access to personalisation, demand forecasting and customer engagement capabilities. Smaller retailers can increasingly plug into off-the-shelf tools, lowering barriers to entry and encouraging a more diverse retail landscape. In doing so, it promotes a broader tenant mix, limits concentration risk and helps sustain the vitality of prime high streets and retail centres.

Implications for asset managers

From dynamic tenant-mix curation to real-time footfall and dwell-time analytics, AI encourages a more data-driven approach to asset management. Across Savills-managed shopping centres, early deployments are focused on forecasting tenant performance, identifying emerging covenant risk, analysing leases and optimising portfolios through tighter cost control and value extraction. In time, these tools should strengthen income resilience, support more disciplined capital allocation and help sustain asset values, especially in complex, multi-tenant environments.

The next chapter

Agentic commerce is considered the next step change in retail, referring to AI systems that can search, compare and transact on a consumer’s behalf. Early examples include ChatGPT-enabled shopping in the US, and in Europe, Frasers Group integrating in assistant discovery and checkout through platforms such as Gemini and Perplexity. In the near term, the greatest impact is likely to be on routine household spending, as automated basket creation and replenishment gain traction.

European adoption, however, appears constrained by trust, data privacy and cultural factors. Given that 84% of consumers are uneasy with AI completing transactions without explicit approval (Retail Economics), uptake should build gradually as a younger, more digitally native cohort accrues spending power. For now, agentic commerce is perhaps best understood as an additional layer rather than a substitute for established behaviours, much as mobile and social commerce initially were – less a new channel than a shift in how demand is expressed.

It’s worth noting that measurement may lag reality. As AI blurs the boundary between channels, traditional distinctions between “online” and “offline” sales become less meaningful. Transactions initiated by AI but fulfilled through click‑and‑collect, ship‑from‑store or in‑store assisted selling may not appear in headline online penetration figures. As such, official data may understate the true digital influence on retail behaviour, even as physical stores continue to play a central role in fulfilment and service.

Still, the pipeline is real: around half of global retailers are evaluating agentic tools, and 20% have begun deploying agents somewhere along the value chain, according to Nvidia, suggesting its impact will grow more meaningful through the decade.

What future impact will AI have on sales channels?

Agentic AI will matter, alongside indirect and halo effects from wider AI adoption, and we do not wish to downplay their potential impact on retail sales channels.

PwC estimates that agentic commerce could drive 8-15% of European online revenues by 2030, with adoption gathering pace faster than previous e-commerce waves. Given that the technology is still emerging, these effects may not be fully embedded in existing bottom-up forecasts, which largely extrapolate current conversion dynamics and channel behaviour. We therefore treat PwC’s estimate as an indicator of upside risk rather than a central case, implying scope for online sales to exceed baseline projections as agentic and related AI effects mature.

Last year, Europe’s online share stood at 14.1%, with GlobalData projecting a rise to 15.5% by 2030 under current expectations (+1.4pp). Layering in agentic commerce alongside adjacent AI effects - stronger fulfilment confidence, wider adoption of virtual try-on and more efficient product discovery - points to a sharper growth trajectory than existing forecasts. As such, the top end of PwC’s estimate can be taken as a plausible uplift scenario.

On this basis, online penetration could reach up to 17.9% by 2030 (+3.8pp versus 2025). At the upper bound, this implies a 2.3 percentage point increase in online share, equivalent to as much as €97 billion in incremental online spending attributable to AI.

This is, of course, a high‑level view. We anticipate adoption will be uneven, with slower uptake in more experience‑driven Southern European markets. Importantly, we do not expect a repeat of earlier e‑commerce displacement effects, reflecting the stronger position of physical retail today - a theme explored in the following section.

Have we seen an online share uplift of this scale before?

Yes, and from a pan-European perspective, the sector weathered the storm, albeit with regional variation. Between 2015 and 2019, Europe’s online share of retail sales rose by around 3.1 percentage points over a comparable period. Similarly, the combined baseline growth and AI-related uplift outlined in Chart 1 could add up to 3.8 percentage points; a convex and potentially larger increase, but still within a range the market has previously managed.

That earlier adjustment created pressure, but many markets absorbed it. Average shopping centre vacancy rose only modestly to 8.5% (+1.4pp), while combined shopping centre and retail warehouse stock grew by around 7% – adding 10 million sq m of floorspace – and rental growth remained largely positive, averaging 1.2% per year across both formats (Chart 2). Polarisation did deepen and weaker stores came under strain, but the sector adapted and evolved, giving credence to the idea that a sustained increase in online share need not be destructive for physical retail.

A more resilient footing for physical retail

The conditions facing physical stores today appear brighter than a decade ago. Much of the adjustment to e-commerce has already occurred: retailer networks have been optimised, and total retail space (shopping centre and retail warehouse) has somewhat stabilised, growing by just 1.2% per year since 2020 (PMA LLP). Yet, stores now play a broader role in anchoring experiential retail, brand engagement and omnichannel fulfilment, from click-and-collect to ship-from-store and returns. This integration lowers competitive tension between channels and, in turn, the sensitivity of physical retail to further increases in online penetration. Early signs of a more harmonious relationship came into view last year, as the average European shopping centre vacancy rate diverged from its long-standing correlation with online sales growth (Chart 3). It’s a small sample, but nonetheless indicative of a more positive coexistence.

Outlook

Looking ahead, AI should continue to raise the productivity of both online and offline retail. As forecasting, inventory mapping, discovery and fulfilment capture sales more efficiently, retailers may require fewer physical sites to generate the same or higher levels of revenue. However, existing stores will become more experience-driven, leaning into the immediacy and social value of in-person shopping.

Agentic commerce may lift online penetration beyond current forecasts. For real estate, this next phase of growth will unfold within a more mature omnichannel ecosystem, allowing stores to absorb greater digital volumes without repeating the disruption seen in earlier e-commerce phases.

The winners will be prime, digitally enabled locations, irrespective of format. Stores that combine strong footfall, data infrastructure and omnichannel capability are best placed to capture productivity gains and margin uplift from AI, particularly where they support discovery and fulfilment. This, in turn, favours prime locations across formats – key city flagships, retail warehousing, food stores and value-led propositions – while legacy department stores and weaker secondary centres face growing pressure to adapt. Even so, rising productivity may temper the need for store rationalisation at the margin, leaving retailers less exposed to the threat of network contraction.

AI’s primary impact will be on margins rather than volumes. Stronger online conversion, fewer returns and more efficient operations should strengthen retailers’ ability to sustain rents in competitive prime locations. More broadly, these gains may improve occupier affordability and reduce default risk among more operationally exposed tenants. AI is therefore more margin-accretive than demand-destructive for bricks-and-mortar retail, supporting a more robust, higher-quality tenant base.

Risks remain. AI adoption is uneven and capital-intensive, potentially widening performance gaps between best-in-class retailers and the long tail. Meanwhile, regulatory constraints, data privacy considerations and consumer trust are likely to limit the speed and breadth of adoption across European markets.

Investors should view data infrastructure and analytics capability as integral to asset quality and long-term value, not a discretionary add-on. For owners and asset managers, this will fuel heightened selectivity. Assets lacking scale, digital readiness or a clearly defined role within omnichannel networks will find it increasingly difficult to compete.