Your next customer might be an AI agent

Two competing commerce protocols launched within weeks of each other earlier this year. Shopify and Google released the Universal Commerce Protocol. OpenAI and Stripe shipped the Agentic Commerce Protocol. Both do roughly the same thing: let AI agents discover products, negotiate deals, and complete transactions on behalf of real people.

This isn’t a prediction. It’s already happening. Amazon, Google, OpenAI, and Meta have all launched AI shopping tools in 2026. Shopify announced that select brands are now selling directly inside Google’s AI Mode and the Gemini app, powered by UCP. According to McKinsey, the U.S. B2C retail market alone could see up to $1 trillion in agentic commerce revenue by 2030.

So what does this actually mean for a mid-market retailer or distributor?

Personalization just changed who the customer is

Personalization engines have been about showing the right product to the right person at the right time. That’s still valuable. The personalized recommendations segment holds 33% of the AI retail market right now, and Gartner just released their 2026 Magic Quadrant for personalization engines — it’s a mature, competitive space.

But agentic commerce shifts the question. It’s no longer just “how do I personalize the experience for a human buyer?” It’s also “how do I make my catalog readable, trustworthy, and transactable for an AI agent?”

That’s a different problem. And most retailers aren’t ready for it.

What agents actually need from you

An AI shopping agent doesn’t browse your category pages. It doesn’t get pulled in by a hero banner or a lifestyle shot. It reads structured data — product attributes, pricing, availability, specifications, shipping terms — and makes decisions based on that data, plus whatever rules its human owner gave it.

This means a few things become table stakes fast:

Structured, complete product data. If your PIM is messy, agents will skip you. They can’t fill in gaps the way a human can by looking at a photo or reading a description. Attributes need to be consistent, normalized, and complete.

Transparent pricing and terms. Agents compare. That’s their job. If your pricing requires a login, a phone call, or a quote form, the agent moves on to the next result. I’m not saying everything needs to be publicly priced — but the buy path needs to be machine-readable.

API-first infrastructure. Both UCP and ACP are protocol standards. They assume your commerce platform can speak to them. If your stack is held together with custom middleware and cron jobs, this is going to hurt.

The B2B angle is where this gets interesting

B2B purchases are already quasi-automated. Procurement teams use rules, approved vendor lists, and reorder thresholds. AI agents are a natural extension of that behavior — they just do it faster and with more inputs.

I’ve seen distributors who still take orders over email and phone. That works because the relationship is strong and the buyer knows who to call. But when a procurement agent starts evaluating vendors based on API response times and real-time inventory feeds, relationship alone won’t keep you on the short list.

The distributors who win here will be the ones who treat their product data and integration surfaces as a first-class product, not a cost center.

Don’t rebuild everything

I’m not suggesting you drop your current roadmap and build an agentic commerce platform. That would be a mistake. Here’s what I’d do instead:

1. Audit your product data quality. Run a gap analysis on your PIM. Missing attributes, inconsistent units, stale descriptions — fix those first. This pays off whether agents buy from you or not.

2. Make your catalog available via API. If you’re on a modern commerce platform, this is probably already possible. If you’re on something custom, start planning the investment. It doesn’t have to be fancy — a read-only product and pricing API is enough to start.

3. Watch the protocol standards. UCP and ACP are early. They’ll evolve. Don’t build to a specific version yet, but start understanding the data model they expect so you’re not starting from zero when the time comes.

4. Talk to your buyers about their AI plans. This sounds obvious, but I’ve been surprised how few distributors have asked their customers whether they’re using or planning to use AI-assisted procurement. The answer will tell you how urgent this is for your business specifically.

The bottom line

Personalization has always been about removing friction between intent and purchase. Agentic commerce removes the human from part of that equation. The retailers who treat this as a data and infrastructure problem — not a marketing problem — will be the ones agents actually buy from.

The question isn’t whether AI agents will place orders on your site. The question is whether your site will be ready when they show up.

What are you seeing on your end? Are your customers talking about AI-assisted procurement yet?