Agentic commerce is a model where AI shopping agents help consumers discover products, compare options and complete purchases—reducing friction across the buying journey. McKinsey describes an “automation curve” where shoppers delegate more tasks to agents over time. Retailers that make product data, policies and offers machine-readable will be easier for agents to find and transact with.

Retail is entering a new phase of digital change: not just personalisation, but delegation. In agentic commerce, the customer is still human—but AI shopping agents increasingly mediate discovery, decisions and transactions. McKinsey argues this won’t happen as a single jump to full autonomy. Instead, it will unfold along an “automation curve”, shaped by how much of the commerce journey people are willing to hand over to machines.

What is agentic commerce?

Agentic commerce describes shopping and purchasing experiences where an AI agent can search, compare and potentially complete a transaction on a customer’s behalf—often with a lightweight approval step.

Think of it as moving from:

• browsing websites and filtering manually, to

• stating an intent (“I need a carry-on suitcase under £120 that fits easyJet rules”), and letting an agent do the legwork.

What’s new: McKinsey’s “automation curve” for agentic commerce

McKinsey’s latest piece introduces the agentic commerce automation curve, which maps how shopping changes at different levels of delegation—starting with rules-based convenience and progressing toward more autonomous, multi-step, multiagent coordination.

The key takeaway for retailers is blunt: agents will increasingly act upstream, triggered by intents like calendar events, reminders, or low-supplies signals. If your offer can’t be understood by machines, you risk becoming invisible in agent-mediated shopping journeys.

How AI shopping agents work

At a practical level, shopping agents typically:

1. interpret user intent (constraints, preferences, budgets, timing),

2. gather options across merchants,

3. evaluate trade-offs,

4. present a shortlist (or make a recommendation), and

5. complete checkout after approval—sometimes without the user ever seeing a traditional product page.

Why machine-readable retail data becomes the competitive edge

In an agentic world, your “shop window” isn’t only your website—it’s your data.

Retailers that win tend to make key elements structured and machine-readable, including:

• product attributes and availability,

• shipping and returns policies,

• pricing rules and promotions,

• brand promises (e.g., sustainability claims, warranties),

• and the problems products solve (not just specs).

That’s because agents need clean, reliable inputs to compare offers and complete transactions confidently.

Practical steps for retailers (what to do now)

Even if fully autonomous shopping is not yet mainstream, you can prepare without betting the business on hype:

1) Audit your product data for “agent-readiness”

Check completeness, consistency, and structure across core categories and high-margin lines. Prioritise standard fields and reduce ambiguity.

2) Make policies easy for machines to parse

Returns windows, delivery thresholds, warranties, and exclusions should be explicit and structured—not buried in prose.

3) Treat AI discovery as a channel

Optimise so agents can reliably find and interpret your offers—this is the next evolution of SEO: not only ranking for humans, but being usable by machines.

4) Prepare for agent trust, identity and payments

As more transactions happen through agents, identity, fraud controls and secure checkout flows become central. Industry efforts are emerging to make autonomous commerce viable at scale.

Summary and next steps

Agentic commerce is the shift toward AI-mediated shopping, where agents reduce friction and increasingly handle steps from discovery to checkout. McKinsey’s “automation curve” suggests retailers should plan for gradual delegation—not a single leap. The businesses that move first on machine-readable retail data and policy clarity will be easiest for agents to find, compare and transact with.

Next steps:

• Identify 1–2 categories to make “agent-ready” first (data + policies + offers).

• Build measurement around conversion, returns, and customer satisfaction for those categories.

• Treat agent discovery and checkout as an emerging channel—then iterate.

FAQs

Q1: What is agentic commerce?
Agentic commerce is shopping where an AI agent can search, compare options and potentially complete a purchase on the customer’s behalf, often with a simple approval step.

Q2: How do AI shopping agents work?
They translate user intent into constraints, gather options, evaluate trade-offs, and either recommend a shortlist or complete checkout after approval.

Q3: What benefits does agentic AI offer retailers and customers?
It can reduce purchase friction, improve matching quality, and accelerate decisions—while pushing retailers to improve data quality and policy clarity so agents can transact reliably.

Q4: What does McKinsey mean by the “automation curve”?
It’s a framework describing levels of delegation, from basic convenience automation to more autonomous agent-driven shopping journeys.

Q5: What should retailers do first?
Start by standardising product and policy data so it’s machine-readable, then treat AI/agent discovery as a channel you actively optimise for.