AI agents are beginning to mediate discovery, comparison and execution. Retailers must become legible to both people and machines.
ANGED’s July analysis captures a shift retail leaders can no longer treat as a distant interface experiment: AI is moving from helping consumers search to participating in comparison, negotiation, complaint handling and—within limits—purchase execution.
The immediate temptation is to build a shopping agent. The deeper implication is that retailers must make their products, policies and operations understandable to other machines.
Consumers are changing the delegation boundary
Accenture’s research surveyed 25,590 people across 16 countries. It reports that 74% would delegate routine tasks such as negotiation or complaint resolution to an agent when it acts under instruction. Thirty-two per cent would allow an agent to decide what to buy while retaining payment control, while only 9% were open to fully autonomous purchases. In another measure, payment decisions remained a much stronger boundary.
These figures are survey responses, not observed purchasing behaviour. They should not be read as a forecast that autonomous commerce will replace human choice next quarter.
They do reveal something useful: delegation is not binary. The same consumer may automate replenishment, ask for help comparing insurance and retain complete control over a meaningful gift. Trust depends on consequence, identity and reversibility.
Retail experiences will need a configurable “delegation dial”.
Brands must become machine-legible
Search optimisation has traditionally focused on pages that people and search engines can find. Agent-mediated commerce adds another requirement: products and policies must be structured enough for an agent to compare and verify.
That includes:
- accurate product attributes and availability;
- explicit price, delivery and return conditions;
- machine-readable guarantees and exclusions;
- reliable inventory and fulfilment APIs;
- clear identity and provenance for reviews and claims;
- stable product identifiers across channels;
- evidence for sustainability or quality assertions.
A brand proposition that depends entirely on visual mood or unstructured campaign language may lose information when an agent reduces alternatives to comparable attributes.
This does not make brand irrelevant. It changes where brand value must be proven.
The internal transformation is larger than the storefront
ANGED also points to retailers moving beyond experiments and applying AI to demand forecasting, inventory, supply chain and operations. KPMG describes a similar transition from pilots to scaled operations, while emphasising the operating, governance and platform choices required.
The customer-facing agent is therefore only the visible edge. The economic value may sit in:
- assortment and merchandising support;
- demand and replenishment decisions;
- supplier document flows;
- product-content operations;
- store and field-service workflows;
- returns and claims;
- contact-centre resolution;
- workforce knowledge;
- fraud and anomaly investigation.
Spain’s ecommerce market exceeded €114.8 billion in 2025 according to CNMC. Continued volume growth increases both the opportunity and the cost of fragmented operational processes.
Retail agents need commercial boundaries
A shopping agent that negotiates, applies promotions or makes substitutions can affect margin, fairness and customer trust. The architecture must distinguish:
- recommendation from commitment;
- personalisation from unlawful discrimination;
- dynamic pricing from opaque treatment;
- helpful substitution from manipulation;
- delegated payment from inferred consent.
Agents should expose the basis for consequential decisions in a form people can understand. Retailers also need to know which agent acted, under whose authority, against which catalogue and policy version.
Three priorities for large retailers
1. Build an authoritative product and policy layer
Before adding conversational interfaces, repair the data contracts on which agents will depend. Product data, inventory, eligibility, delivery and returns must agree across systems.
2. Choose workflows with measurable completion
Start where an agent can move from request to verified outcome: delivery changes, order-status exceptions, return eligibility or supplier-document triage. Avoid open autonomy around price, credit or vulnerable consumers until governance is proven.
3. Design for a two-sided agent market
Retailers will operate their own agents while receiving traffic and requests from consumer agents. They need interfaces, policies and observability for both directions.
The Lexdatum view
Agentic commerce is not simply another sales channel. It is a change in who—or what—performs the work of discovery, comparison and execution.
The retailers that benefit will not be those that add the most chat interfaces. They will be those that make their enterprise truth accessible, their capabilities callable and their commercial boundaries explicit.
The next customer journey may contain fewer clicks. Behind it, the operating architecture will be substantially more demanding.
Sources
- ANGED — TrendHunter: el retail entra en la era de las decisiones algorítmicas (2026-07-28)
- Accenture — Talk to my AI agent: The new rules of brand value (2026-06-03)
- KPMG — KPMG Global tech report 2026: Consumer & Retail (2026)
- CNMC — El comercio electrónico superó en España los 114.800 millones de euros en el año 2025 (2026-07-03)
Research note: This analysis is based on public information available on 31 August 2026. Product capabilities, prices and regulatory interpretations can change. Vendor-reported metrics are identified as such. This is not legal, security or investment advice.