Retail is entering the agentic commerce era: AI agents that not only recommend but plan, compare, negotiate and complete purchases on a shopper’s behalf. Early deployments—from Amazon Rufus to auto-purchasing via Alexa+—hint at a near future where agents become the default shopping interface. Here’s what’s changing and how to prepare.

Key points / benefits

• Hyper-personalisation at scale: Agents maintain a memory of preferences, constraints and context to curate shortlists and bundles.

• Autonomous transactions: From price-watch rules to one-click fulfilment, agents can execute routine buys end-to-end with user-set guardrails.

• Merchant efficiency: Better demand signals, smarter pricing, lower service load (e.g., AI assistants resolving the majority of service chats).

How it works

McKinsey describes agentic commerce as AI agents acting across the full journey—needs discovery → evaluation → purchase → support—in retailer apps or third-party platforms (ChatGPT, Gemini, Perplexity). Tooling is maturing fast: storefront assistants (e.g., Rufus), shopping briefs in Google Shopping, and open protocols/payment rails enabling in-chat checkout.

Two deployment models

1. Owned-channel agents (your site/app): maximum control of catalogue, pricing, policies and data.

2. Ecosystem agents (chat platforms, marketplaces): reach and acquisition, but you must expose structured endpoints (catalogue, stock, delivery, returns) so external agents can transact reliably.

Practical steps (playbook retailers can use now)

1. Data readiness. Fix product metadata: titles, attributes, images, availability, structured data (Product/Offer/Review) so agents can reason and compare.
   
   

2. Agent endpoints. Publish dependable APIs for search, pricing, stock, baskets, payments and returns to let third-party agents build carts and check out safely (see Stripe’s guide).
   
   

3. Payments + trust. Support tokenised payments and clear spend limits / approval rules for autonomous buys (e.g., price-drop auto-purchase). Start with opt-in, event-scoped autonomy.
   
   

4. Service automation. Use an AI assistant to resolve common issues (delivery ETAs, returns, refunds) and measure CSAT/deflection/conversion—Klarna’s early results show what’s possible.
   
   

5. Merch & pricing. Let agents assemble bundles and apply dynamic offers; give them constraints (margin floors, inventory thresholds) so automation never breaks your trading rules.
   
   

6. Governance & compliance. Map obligations under the EU AI Act (transparency, risk management) and UK guidance before scaling autonomous features.

Examples in the wild

• Amazon Rufus (UK): a generative shopping expert embedded in Amazon’s app/site to help evaluate options.

• Alexa+ auto-purchasing: price-watch rules that can place orders automatically to your saved address and payment details.

• Google Shopping AI briefs: AI-generated “what to consider” guides + product picks accelerate research.

• Klarna AI assistant: handled two-thirds of customer-service chats in its first month, illustrating post-purchase automation potential.

FAQs

Q1: What is “agentic commerce”?
Shopping mediated by AI agents that can interpret goals, source products, compare options, negotiate, and execute transactions—either in your channels or in third-party platforms. McKinsey & Company

Q2: How do agents personalise shopping?
They build a persistent profile of preferences, constraints (budget, size, ethical choices), and real-time context, then generate tailored shortlists and bundles. McKinsey & Company

Q3: What benefits do merchants see?
Higher conversion, lower service load, smarter inventory/price decisions—especially when assistants resolve routine queries and surface intent signals pre-purchase. Klarna

Q4: What about risk and regulation?
Start with opt-in autonomy, clear spend caps and audit trails. Align with the EU AI Act (transparency, risk controls) and UK guidance as rules phase in. Digital Strategy+1

Summary

Agentic commerce is moving from concept to reality. The winners will treat agents as a new channel: clean product data, reliable commerce APIs, safe payments, and measurable automation. Build your owned agent for differentiation—and be ready to serve ecosystem agents where customers already are.