Agentic Commerce – Consumers are moving faster than merchants think

Everyone is talking about AI agents buying things on your behalf. But here i would like to talk about what's actually blocking it.

Here's the real market picture - from the data, not the hype. I will use all the information wich i receive or colelct. In case you have some interesting reports or link would aprpaciate if you can share them.

The gap nobody mentions

AI agents are involved in roughly 3% of online transactions today. Yet 85% of merchants expect agent-led shopping to account for at least 10% of their online volume within the next two years.

And it tells you something important: the direction is clear, the destination is agreed on, but the road is not built yet. All data is collected frompublick reports of Checkout.com, VISA, Shopify, Mastercard.

Consumers are moving faster than merchants think

On the consumer side, the signal is real.

• 13% already use AI agents for at least 1 in 10 of their purchases
• 33% expect to reach that point within 12 months
• 71% of 18–24 year olds are already familiar with AI shopping agents

Compare that to just 19% of people aged 55+. This is a generational adoption curve - which means the window for businesses to prepare is shorter than it looks.

Geography matters too. UAE leads at 77% consumer familiarity, followed by China (67%) and Brazil (61%). Western markets are behind. But the pattern is consistent: the markets that know agentic commerce best are adopting it fastest.

The ceiling is real though. 24% of consumers say they would never delegate purchases to an AI agent. This isn't going to be universal or smooth. But the direction of travel is set.

The trust and liability problem is the actual bottleneck

Here's where it gets interesting - and unresolved.

It's not the technology. AI agents can already search, compare, decide, and execute. The infrastructure to connect them to merchants is being built. The protocols exist or are emerging.

The problem is: nobody has agreed on who's responsible when something goes wrong.

The Checkout.com report asked merchants and consumers the same question: when an AI agent makes an unauthorized or incorrect purchase, who should fix it?

Merchants said: payment providers (35%). Consumers said: the agent platform (38%).

Nobody pointed at themselves. And everyone pointed at someone different.

That's not a technical failure. That's a governance vacuum. And until liability frameworks are defined - consent standards, dispute handling, accountability chains - the whole ecosystem is building on uncertain ground.

Consumers are clear about what they need before they trust agents with their money: spending caps, instant permission revocation, easy cancellation windows, approval thresholds. Basic controls. None of them standardized yet.

The infrastructure race - four protocols, no winner

Meanwhile, the biggest players in payments and AI are racing to establish the rails.

Visa launched its Intelligent Commerce initiative and Trusted Agent Protocol - and in June 2026 announced a direct partnership with OpenAI to enable Visa payments inside agentic experiences. Tokenized credentials, real-time fraud monitoring, spending controls baked in.

Mastercard took a different approach. In March 2026, together with Banco Santander, they completed Europe's first live payment executed by an AI agent - using Mastercard Agent Pay, inside a controlled environment.

Google announced the Universal Commerce Protocol (UCP) - designed to make merchants discoverable and transactable across Search AI Mode and Gemini.

OpenAI is building its own Agent Commerce Protocol (ACP).

Four frameworks. Competing approaches. No interoperability standard yet.

For merchants, this creates a difficult situation: you can't bet on a single protocol. You need to stay infrastructure-agnostic while the standards war plays out - which means investing in the fundamentals that work across all of them: structured product data, clean APIs, tokenized payments, clear agent permissions.

What this actually means for businesses right now

The Checkout.com report surveyed 400 Heads of Payments and landed on three things merchants need to do - not eventually, but now:

1. Make your catalog agent-readable. Both ACP and UCP depend on structured, machine-readable product data. If an AI agent can't understand your inventory, pricing, and availability in milliseconds - you don't exist for it.
2. Make sure your payments can handle delegation. Agent-initiated transactions are different from human-initiated ones. They require tokenization, credential storage, and flows designed for automated decision-making - not human checkout sessions.
3. Define what agents are allowed to do with your product. Permissions, spend limits, category restrictions, dispute rules, consent frameworks. If you haven't defined these, agents can't safely operate within your environment.

The merchants getting ahead aren't waiting for a winner to emerge. They're building the foundations that will matter regardless of which protocol wins.

One thing that hasn't been said enough

57% of consumers say they would let AI switch brands for better value.

That's not a small number. It means brand loyalty - as we've known it - gets disrupted when agents are making the call. The criteria shift from emotional to functional: price, reviews, availability, returns policy, machine-readability.

In an agent-led world, the best product might not win. The most agent-accessible product wins.

Where Portalio fits in

At Portalio, we build the technical layer that makes this possible for real businesses - API architecture, agent-ready integrations, payment flows, and the logic that connects intent to execution. If you're a merchant, a marketplace, or a platform wondering how to get ready for agentic commerce, that's exactly the kind of problem we work on.

Sources: Checkout.com Agentic Commerce 2026 Report (12,000 consumers, 400 Heads of Payments globally); Visa × OpenAI Partnership Announcement, June 2026; Visa/IFTF Commerce of Tomorrow, Today - CEMEA Report.