Debenhams Tests Agentic AI Commerce in PayPal App
Debenhams pilots agentic AI commerce within PayPal app, testing autonomous shopping agents to reduce mobile checkout abandonment through conversational interfaces.
A major UK retailer is testing whether agentic AI can solve mobile checkout abandonment by keeping the entire purchase flow inside a payment app. Debenhams Group deployed an autonomous shopping agent within PayPal's ecosystem, making it the first UK retailer to test this approach.
The pilot addresses a fundamental commerce problem: revenue leakage from mobile friction. Rather than driving traffic to their own properties, Debenhams is placing inventory where transaction liquidity already exists.
Natural Language Shopping Without Redirects
The agentic AI system operates entirely through conversational prompts within the PayPal app. Users can search for items across Debenhams Group's portfolio using natural language instead of keyword matching.
The agent scans user profiles to align recommendations with budget constraints and purchase history. It asks contextual follow-up questions to narrow product selection and locate available inventory.
Key technical capabilities include:
- Profile-aware recommendations — Budget and preference alignment
- Conversational refinement — Follow-up questions to narrow options
- Real-time inventory — Live stock and pricing visibility
- Embedded checkout — Transaction completion within chat interface
Once a customer selects a product, the entire transaction occurs within the chat window. The backend automatically applies saved account credentials for delivery and payment, eliminating mobile site redirects.
Data Infrastructure Requirements
Effective agentic commerce demands real-time inventory and pricing accuracy to prevent hallucination errors. Debenhams recently partnered with Peak AI to improve forecasting across stock, sales, and pricing systems.
This data lineage work appears deliberate — autonomous agents need clean, current information to function reliably at scale. The company also launched an internal AI Skills Academy to ensure teams can manage these automated workflows.
The infrastructure requirements span several areas:
- Real-time inventory — Live stock levels across all brands
- Dynamic pricing — Current promotional and markdown data
- User profile data — Purchase history and preference modeling
- Payment processing — Seamless credential handling
Integration Architecture
The system co-developed by Debenhams and PayPal currently focuses on select US customers. A wider release across both US and UK markets is planned for later this year.
In the US deployment, the agent integrates with external tools including Perplexity and Microsoft Copilot. This suggests a plugin-style architecture that could expand to additional AI services.
Strategic Transaction Flow Compression
The business rationale centers on transaction volume optimization. Debenhams Group processes 16 percent of its total sales through PayPal.
By embedding inventory discovery in a channel where customers already complete transactions, the retailer compresses the traditional sales funnel. This eliminates multiple friction points:
- App switching — No redirects to mobile sites
- Credential re-entry — Payment details already stored
- Search complexity — Natural language vs keyword matching
- Decision paralysis — Agent-guided product selection
The approach tests whether third-party platforms can capture high-intent traffic more effectively than proprietary storefronts. Instead of forcing traffic to owned properties, Debenhams positions inventory where purchase intent already exists.
Multi-Brand Scope
The pilot spans Debenhams Group's full brand portfolio, including boohoo, boohooMAN, Karen Millen, and PrettyLittleThing. This cross-brand inventory access within a single conversational interface could drive discovery across the group's properties.
Users can find items from multiple brands through unified natural language queries, potentially increasing basket size and brand cross-pollination.
Implementation Challenges
Success depends heavily on data accuracy and the agent's ability to interpret queries without hallucination. Agentic commerce systems must handle ambiguous requests, inventory constraints, and pricing changes in real-time.
The technical complexity includes:
- Query interpretation — Understanding vague product descriptions
- Inventory synchronization — Preventing overselling across channels
- Pricing consistency — Real-time promotional updates
- Error handling — Graceful degradation when systems fail
Unlike traditional search interfaces, conversational agents create expectations for contextual understanding and memory across interaction sessions.
Bottom Line
This deployment represents a significant test case for agentic AI in e-commerce applications. Rather than building proprietary chat interfaces, Debenhams leverages existing payment platform distribution.
The approach could influence how retailers think about customer acquisition — optimizing for where transaction intent exists rather than driving traffic to owned properties. Success metrics will likely focus on conversion rates, average order values, and customer acquisition costs compared to traditional mobile commerce flows.