“Your next customer might not be a person.” That is the uncomfortable premise behind agentic commerce, and the thesis is simple: retail is moving from omnichannel to a world where AI agents surface, compare and buy on a consumer’s behalf — and the first question for any store is no longer whether shoppers can find you, but whether their agent can.

From omnichannel to agentic commerce

For a decade the retail mantra was omnichannel: meet the customer on every screen and in every aisle. The emerging shift, reported by PYMNTS and framed by platform vendors like commercetools, is different in kind. Instead of a human browsing your channels, an AI agent does the surfacing, the comparing and increasingly the buying. The scale of the projection is what makes this more than a demo. Juniper Research projects agentic-commerce transaction value at roughly $8 billion in 2026, growing to about $3.5 trillion by 2031.

That is a jump of several orders of magnitude in five years. Even allowing that long-range forecasts are directional rather than precise, the direction is the point: a channel that is a rounding error today is projected to become a primary way commerce happens.

People want a co-pilot, not an autopilot

Here is the nuance that separates hype from reality. Consumers are enthusiastic about assistance and wary of surrender. Splitit and PYMNTS research found that 61% of consumers would accept AI-assisted “Pay Later” recommendations — a share that rises to 80% among Gen Z. But only 2% want full automation. The overwhelming majority want to review and approve before anything is committed on their behalf.

And there is a hard line they will not cross. Credit-score protection ranks as the top non-negotiable, named by 59% of consumers. The signal for merchants is clear: agents are welcome as advisors and helpers, not as unsupervised spenders. The trust curve is real, and skipping it loses customers rather than winning them.

Why agent-discoverability becomes a sales channel

If shoppers increasingly delegate discovery, the economics of visibility change. Per commercetools, AI-referred traffic reportedly converts at roughly 8 times the rate of social-media traffic, and retailers running their own shopper agents reportedly grew sales about 59% faster during the 2025 holiday season. Those figures come from a platform vendor and should be read as reported claims rather than settled fact — but even discounted, they describe a channel worth being present in.

The pattern that keeps recurring is what commercetools calls progressive delegation: assisted discovery, then guided decisions, then selective automation, and only eventually full autonomy for the narrow tasks a customer explicitly trusts. It maps neatly onto the consumer data above — start as a co-pilot, earn the autopilot.

• Assisted discovery: the agent finds and surfaces relevant options.
• Guided decisions: it compares and recommends, human decides.
• Selective automation: it acts on narrowly scoped, pre-approved tasks.
• Full autonomy: reserved for the 2% of cases customers explicitly hand over.

Being callable is the new being findable

All of this lands on a practical question. If an agent is doing the shopping, is your store something that agent can actually query and buy from? Being agent-callable — exposing a machine-readable, MCP-native surface — starts to look less like a feature and more like a sales channel in its own right.

This is where an honest, proportionate note about tooling belongs. The self-hosted, source-available VBWD platform ships an MCP-native, agent-callable surface alongside a shop and a provider-agnostic payment core. The point is control: a merchant can expose exactly what an agent may query and buy — settling against the same invoice line item as a human checkout — without handing a third-party platform the keys to the customer relationship. The documentation covers the agent surface and payment core, and the code is on GitHub. It is one option among several; the broader lesson stands regardless of stack.

The strategic takeaway is that discoverability is being redefined. For twenty years it meant ranking in search and showing up in feeds. In an agentic market it means being structured, callable and buyable by software. The merchants who treat the agent as a customer — and design a surface it can actually transact with, on terms the merchant sets — are the ones positioned for the channel Juniper says is coming.

Written for Red Robot with AI assistance and human editing. Based on reporting by PYMNTS, commercetools and Juniper Research.