Agentic commerce is where shopping shifts from clicking around to delegating outcomes: a customer (or employee) states intent, and an AI agent plans, compares, and executes across systems.
In [McKinsey’s Oct 2025 report on agentic commerce](https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-agentic-commerce-opportunity-how-ai-agents-are-ushering-in-a-new-era-for-consumers-and-merchants#/), they argue this isn’t just “better personalisation” it’s a structural change in how discovery and transactions happen, with big upside but real hurdles around infrastructure, monetisation, and trust.
Here’s the practical version: what’s exciting, what’s hard, and what to do next.
**Where the upside actually is**
1) Conversion moves upstream
In an agentic flow, the “storefront” isn’t only your PDP, it’s your data + policies + reliability. Agents reward merchants who are easy to understand and safe to transact with.
What matters more:
- clean product attributes and variants
- accurate stock + delivery promises
- clear returns/warranty rules
2) Concierge-level experiences become normal
A good agent experience feels like: “I said what I wanted, and it happened.”
That can lift repeat purchases (replenishment), AOV (bundles), and loyalty (preferences remembered).
3) Ops improves beyond internal automation
Agents can coordinate actions between tools: support → OMS → carrier → payments. This is where “exception handling” gets cheaper.
**The three big challenge zones**
1) Capabilities, go-to-market, and brand
First-mover advantage = integration quality.
Not flashy demos. Boring reliability: stable APIs, structured feeds, predictable fulfilment.
Brand changes too: you’ll need a “machine-facing brand” (consistency, trust signals, policy clarity) alongside the human-facing one (story, community).
2) Monetisation (when agents compress browsing)
If agents reduce browsing, some classic tactics weaken (especially impulse-driven funnels). Winning models tend to be:
- Paid convenience: premium shipping windows, setup, concierge support
- Bundles as outcomes: “move-in kit”, “travel essentials”, “starter routine”
- Subscription + replenishment: agents are naturally good at recurring optimisation
3) Trust and risk (the make-or-break layer)
The key question: if an agent makes a bad purchase, who’s accountable?
To protect trust, mature setups lean on transparency, human override, audit trails, and strong governance—often anchored to established risk frameworks like NIST AI RMF and regulatory expectations like the EU’s emphasis on transparency/oversight.
**A simple “agent-ready” checklist**
Agent-friendly interfaces: catalog, pricing, inventory, shipping quotes, cart/checkout
Identity + consent: delegated authorisation, limits, approvals
Payments: secure, auditable, low-friction
Knowledge: structured product truth + consistent policy logic
Guardrails: monitoring, escalation, kill switch, rollback
**In summary**
This is where agencies like Lmo7 typically help: making stacks “agent-ready” without overbuilding, and putting governance in place before automation becomes customer-experience debt.