Set as preferred source
Insights based on Rewiring retail in Europe: The AI imperative.
Retail has absorbed several technology shifts without changing its core economics. E-commerce moved transactions online. Mobile shortened the path from discovery to purchase. Marketplaces aggregated choice. AI is different because it changes how demand is formed, how decisions are made, and who controls the commercial interface.
The economic opportunity is substantial. End-to-end AI transformation could unlock €240 billion to €320 billion in value across European retail over the next five years, equal to a 4–10 percentage-point improvement in EBITDA. But this value will not come from isolated automation initiatives. It will come from integrating commercial decision-making, operating workflows, product data, governance, and machine-to-machine commerce infrastructure.
This is why AI in retail should not be treated as another productivity layer. It is becoming part of the operating system for the sector.
The margin problem is becoming a decision problem
Retail has always operated with thin margins, volatile demand, labor-intensive operations, and complex supply chains. AI matters because managing these pressures requires a large number of decisions. Pricing, promotions, replenishment, assortment, markdowns, returns, labor scheduling, supplier negotiations, and last-mile fulfillment all depend on thousands of repeated decisions across stores, categories, channels, and markets.
The greatest AI value lies in the commercial core. Commercial merchandising can contribute a 2–4 percentage-point improvement in EBITDA, while buying can add another 1–2 percentage points. Marketing, supply chain, omnichannel operations, and support functions also matter, but the core lever is clear: AI creates the most value where better decisions directly affect margins, availability, conversion, and working capital.
This makes current investment patterns revealing. Only 15% of AI investment among surveyed European retailers is concentrated in the commercial domain, despite its outsized value potential. More investment is flowing into marketing and support functions, where use cases are easier to deploy and produce more visible results. This points to a familiar technology trap: funding what is easiest to implement before funding what can change the underlying economics.
The consequence is a delayed response to margin pressure. Retailers that use AI mainly for content, reporting, and copilots may improve efficiency. Retailers that apply it to pricing, promotions, assortment, supplier terms, replenishment, and inventory allocation can change their underlying margin structure.
The AI paradox is integration, not adoption
Most retailers are already experimenting with AI. The harder question is whether those experiments are changing financial performance. Eight in ten surveyed participants say it is too early to determine AI’s impact on EBITDA. That is the practical AI paradox: activity is rising faster than measurable value.
The report applies a 70% realization factor to its projected EBITDA impact. This adjustment accounts for overlap among initiatives, dependencies across functions, process inefficiencies, and execution constraints. It matters because theoretical value potential does not automatically translate into captured value. Workflows collide. Data is incomplete. Incentives conflict. Adoption varies across functions.
The 70% factor turns AI valuation into an execution question. It avoids the unrealistic assumption that the value of every use case will accumulate without friction. It also clarifies why pilots fail to scale: the bottleneck is rarely the model alone. It is the organization’s ability to embed AI into the daily work of buyers, planners, store teams, marketers, supply chain teams, and finance functions.
AI does not scale through more pilots. It scales when workflow ownership, data quality, talent, governance, and technology advance together.
Softline retail shows where complexity becomes leverage
AI value varies by retail model. Grocery may see a 4–6 percentage-point improvement in EBITDA. Hardline retail may see a 6–8 percentage-point improvement. Softline categories, including fashion, beauty, and personal care, may see an 8–10 percentage-point improvement.
The reason is structural. Softline retail has volatile demand, broad assortments, greater personalization potential, markdown exposure, return risk, and faster trend cycles. These are exactly the conditions in which AI can improve commercial outcomes. Better demand sensing reduces overstock. Better personalization improves conversion. Better sizing tools and product-recommendation models can reduce returns. Better content-generation tools shorten campaign cycles.
Complexity becomes leverage when decision systems improve. But the reverse is also true. Weak product data, fragmented inventory visibility, inconsistent customer records, and disconnected supplier information can quickly reduce AI’s value. Data quality is no longer merely an item on the technical backlog. It is a commercial constraint.
Agentic commerce changes the front door to retail
The most significant strategic shift may be the rise of agentic commerce. AI systems are beginning to influence product discovery, comparison, basket building, and eventually transaction execution. The report estimates that the global opportunity could reach $3 trillion to $5 trillion by 2030. According to the report, 84% of European consumers surveyed use AI tools in their daily lives, while 38% use AI to research products or inform purchasing decisions.
This changes the customer journey at its most sensitive point: discovery. Retailers are accustomed to optimizing for search engines, marketplaces, apps, social platforms, and owned channels. Agentic commerce adds a new intermediary. AI agents may increasingly determine which options are visible, comparable, trusted, and actionable.
The first wave involves generative engine optimization. This is more than traditional SEO under a new label. It involves making product information, claims, reviews, availability, specifications, and brand signals understandable to AI systems that generate direct answers rather than lists of links.
The second wave is AI-orchestrated commerce, in which assistants on retailer websites, apps, or platforms help customers search, choose, configure, and buy. The third wave is autonomous commerce, in which agents execute purchases within customer-defined boundaries.
That progression changes the idea of the customer. In a growing number of categories, the first “visitor” to a digital storefront may not be a person. It may be an AI agent acting on a person’s behalf. Retail environments designed only for human browsing will become incomplete.
The next infrastructure bottleneck is agent readiness
Agentic commerce creates three interaction models: agent-to-site, in which an AI agent navigates a traditional website; agent-to-agent, in which a buyer’s agent communicates with a seller’s agent; and agent-to-broker-to-agent, in which an intermediary platform manages the interaction.
This makes agent readiness an infrastructure requirement. Retailers need structured product catalogs, reliable APIs, real-time inventory visibility, access to current pricing, authentication, authorization, payment rules, and machine-readable policies. Returns, delivery options, substitutions, loyalty programs, and sustainability claims also need to be interpretable by machines.
The bottleneck is not only technical. It is also about commercial control. If retailers are not agent-ready, third-party platforms may become intermediaries between retailers and customer demand. These platforms may compare products, negotiate baskets, and route transactions without preserving the retailer’s ability to shape the experience. In commoditized categories, this could push competition toward price and logistics. In differentiated categories, the challenge is making that differentiation clear enough for AI systems to explain and carry forward.
Retail positioning must therefore expand from human persuasion to machine-readable explainability. Human shoppers respond to design, emotion, stories, and familiarity. AI agents compare evidence.
From KYC to KYA
As agents begin interacting with commerce systems, trust infrastructure may need to expand from “know your customer” to “know your agent.”
KYC was built around human identity and transaction legitimacy. KYA extends the problem to machine-to-machine commerce. Retailers need to know whether an agent is authorized, what it is allowed to do, whether it is acting within customer-defined boundaries, and whether its behavior indicates fraud, scraping, manipulation, or malicious intent.
This creates a new risk layer. A poor recommendation is one problem. A malicious or poorly governed agent with access to pricing, inventory, checkout, payment, or returns systems presents a much broader risk. Security, consent, auditability, and accountability therefore become integral to revenue operations.
Europe’s regulatory context makes this even more important. Privacy, consumer protection, AI governance, product claims, and sustainability disclosure requirements are converging. Governance can become either a scaling mechanism or a bottleneck, depending on how early it is embedded in AI systems.
From functions to flows
AI’s organizational impact extends beyond automation. Retail organizations may move from rigid functional hierarchies toward flatter, outcome-driven teams that own end-to-end workflows and use networks of AI agents. The report estimates potential efficiency gains of approximately 35% across role levels and up to 50% in advanced transformations.
These gains would largely come from reducing coordination overhead. Traditional retail relies on handoffs among merchandising, marketing, supply chain, store operations, finance, IT, and HR. Each handoff slows decisions. Each function optimizes its own part of the system.
AI exposes the cost of this structure. A pricing recommendation has limited value if promotion planning, supplier funding, inventory management, and store execution remain disconnected. A demand forecast has limited value if replenishment and allocation cannot adjust. A customer assistant improves the experience only if product data, inventory, delivery, returns, and service policies are aligned.
The operating model shifts from adding AI tools within individual departments to organizing work around end-to-end flows: commercial planning, demand creation, replenishment, fulfillment, service recovery, and workforce planning. AI agents can execute parts of those flows. Human judgment remains central where trade-offs, exceptions, trust, and accountability matter.
Sustainability becomes operational, not separate
AI’s value extends beyond revenue, margins, and productivity. It also affects waste, logistics intensity, supply chain traceability, employee experience, and collaboration with suppliers.
Better forecasting reduces waste. Better replenishment lowers excess inventory. Better routing reduces logistics intensity. Better inventory placement can reduce split shipments and returns. Better supply chain traceability improves supplier oversight.
This matters because sustainability in retail is often treated as a reporting layer. AI connects it to operating performance. In grocery, waste reduction affects both margins and sustainability. In fashion, better demand sensing can reduce overproduction. In omnichannel models, fulfillment optimization affects both costs and emissions.
The strongest sustainability gains will come when environmental and social considerations are incorporated into everyday decisions rather than managed separately after the fact.
The competitive divide will widen through execution
AI tools are becoming widely available. Access alone will not create an advantage. Deployment quality will.
The retailers best positioned to benefit will have reliable data foundations, clear value maps, agent-ready infrastructure, mature governance, and the discipline to concentrate investment in high-impact domains. Large omnichannel players can benefit from integrated supply chains and fulfillment networks. Softline players can benefit from faster demand sensing and personalization. Platform-native players can benefit from data scale and search infrastructure.
Smaller retailers face a sharper constraint. Agent readiness requires structured data, APIs, authentication, governance, and technical investment that may exceed their internal capacity. The main risk is not that AI will eliminate niche retail. It is that third-party agents may take control of product discovery before smaller brands make their offerings machine-readable.
The new retail divide will not be only between online and offline retailers. It will be between those that are visible to AI agents and those that are not.
Key risks
Commercial underinvestment remains the first risk. AI budgets concentrated in content, copilots, and support functions can create visible activity while leaving the largest margin opportunities untouched.
Agent-driven disintermediation is the second. If AI platforms control discovery and comparison, retailers in commoditized categories may lose direct influence over customers and compete mainly on price, availability, and delivery.
Workflow fragmentation is the third. Scaling dozens of disconnected use cases can increase complexity without changing enterprise economics.
Data invisibility is the fourth. Weak product taxonomy, incomplete attributes, inconsistent claims, and poor structured data reduce AI discoverability and distort how products are represented.
The lack of a “know your agent” framework is the fifth. Machine-to-machine commerce creates risks related to authorization, fraud, malicious agents, and unclear accountability when automated transactions go wrong.
Strategic implication
Retail AI is moving from broad experimentation toward a clearer divide between companies that capture value and those that do not. The next source of competitive advantage will depend on how effectively organizations use AI to improve margins, demand visibility, workflow speed, trusted machine-to-machine commerce, and operating resilience.
Product data becomes demand infrastructure. APIs become channel infrastructure. Governance becomes scaling infrastructure. Workforce capability becomes a source of operating leverage. Sustainability becomes part of inventory and logistics intelligence.
The stronger position will belong to retailers that treat AI less as a technology program and more as a redesign of how value is created, captured, and defended.
