Agentic Commerce 2026: How AI Agents Are Changing Ecommerce

Ecommerce has always put the customer in the driver's seat — searching for products, reading reviews, comparing options, and completing checkout themselves. That model served a decade of online retail, but it assumes a human is doing the browsing. AI is now changing how consumers discover products and complete purchases, shifting some of that work from the shopper to software. A customer can describe what they want in plain language and let an assistant narrow the field, check availability, and move toward a purchase.

Agentic commerce represents the next stage of ecommerce, where AI agents can help users discover products, compare options, and interact with businesses. Rather than replacing the customer, the agent acts as an intelligent layer between the shopper's intent and the merchant — and that has direct consequences for how businesses present products and how payments are processed.

Quick Summary

Agentic commerce is an emerging ecommerce model where AI agents assist customers with product discovery, decision-making, and transactions. As AI becomes more involved in commerce, businesses will need new approaches to customer interaction, payment, and transaction infrastructure. The shift moves the buying journey from a manual, search-led process toward an assisted, conversational, and increasingly automated one — which means product data, checkout, and settlement all have to become machine-readable and programmable.

What Is Agentic Commerce?

Agentic Commerce Definition

Traditional ecommerce follows a path the customer walks end to end:

Customer → Search → Compare → Checkout

The shopper types queries, scans results, weighs alternatives, and pays. Agentic commerce inserts an intelligent assistant into that path:

Customer → AI Agent → Product Discovery → Transaction

The key difference is autonomy and intent. AI agents act as intelligent assistants that understand user intent and help complete shopping tasks — they can interpret a request like "find a durable carry-on under £150 that fits Ryanair's cabin limits," filter against real constraints, and surface a shortlist the customer would otherwise have built by hand. At higher levels of autonomy, the agent can also place the order within boundaries the user sets.

It is worth separating two shades of the term. In its lighter form, agentic commerce is assisted discovery: the agent recommends and the human still clicks buy. In its stronger form, the agent is authorised to transact on the user's behalf within a defined mandate. Both are "agentic," and most of today's deployments sit somewhere in between.

How Agentic Commerce Will Change Ecommerce

From Search-Based Shopping to AI-Assisted Shopping

Traditional ecommerce puts the work on the buyer:

Users search for products.
Users compare options.
Users complete purchases.

Agentic commerce redistributes that work toward the agent:

AI agents understand customer preferences from natural-language requests and history.
AI agents recommend relevant products instead of returning a raw list of keyword matches.
AI agents assist purchasing decisions by applying the customer's stated constraints (budget, specs, delivery window).

For merchants, this changes which shelf gets seen. When a human types "running shoes" into a search box, ranking is driven by SEO and ads. When an agent selects "lightweight neutral trainers under £90 with free returns," ranking is driven by structured product data the agent can parse — attributes, stock, price, shipping, and policies expressed in a machine-readable way. Discovery stops being a page of blue links and becomes a query answered from a feed.

New Opportunities for Businesses

Businesses that want to be found and chosen by agents may need to optimise for a different discovery layer. That includes:

AI-driven product discovery — making sure products are exposed through feeds, APIs, and structured data agents can read, not only through a traditional web storefront.
Better product data — clean, consistent attributes (size, material, compatibility, lead time) so agents can match intent to the right SKU.
Faster customer interactions — conversational and API-based touchpoints that let agents retrieve real-time availability and pricing.
Automated commerce experiences — the ability for an agent to move from selection to transaction without a human re-entering everything at checkout.

The opportunity is reach: agents route demand toward the merchants they can understand and trust. The risk is disintermediation — if the agent becomes the customer-facing brand, individual merchants may lose the direct relationship. How that balance settles will shape the next few years of retail.

The Role of Payments in Agentic Commerce

Why Payment Infrastructure Needs to Evolve

AI-driven commerce creates new payment requirements that traditional, human-centric checkout was not built for. Cards were designed around a person with a billing address and a device; agents don't fit that mould. Four requirements stand out:

Secure transaction authorization — agents need scoped, revocable credentials rather than a static card number. Emerging approaches use tokenised, mandate-based authorization so an agent can act within limits the user set, with a verifiable record of what was approved.
Automated payment workflows — checkout has to be callable by software through an API, not just clicked by a human. That means payment initiation, confirmation, and receipt handling that an agent can complete without a manual screen.
Global payment capabilities — agents routinely shop across borders and currencies, so the underlying rails need to settle in the supplier's local currency without forcing the customer through a poor-rate conversion.
Flexible settlement options — businesses receiving agent-initiated orders benefit from programmable settlement, including multi-currency balances and, for some workflows, stablecoin settlement alongside fiat.

The industry is moving here quickly. Standards and pilots announced across 2025 and 2026 include Google's Agent Payments Protocol and Universal Commerce Protocol, Mastercard's Agent Pay and Agent Suite, Visa's Intelligent Commerce and Trusted Agent Protocol, and Stripe's Agentic Commerce Suite with shared payment tokens. In the UK, Visa's agent-payment testing has extended to major UK banks, and the Payment Systems Regulator is reviewing how existing authorization rules apply to agent-initiated transactions. For UK merchants, the practical takeaway is to watch FCA and PSR guidance and to choose payment partners whose infrastructure is built for software-initiated flows.

Payment Operating System Built for Agentic Commerce

The diagram above is simple, but each arrow hides real engineering. The "Payment Infrastructure" layer has to accept a software-initiated request, authorize it against a scoped mandate, route it to the right rail, and settle reliably — often across currencies. Platforms built as programmable payment operating systems, rather than as single-purpose checkout pages, are better positioned for this shift.

PhotonPay is an FCA-authorised payment operating system in the UK that provides the building blocks agentic and automated commerce depend on. It is presented here as one example of the kind of infrastructure this transition requires, not as a replacement for a merchant's existing stack.

API-driven payments

Accept customer payments through a checkout layer that software can call — supporting cards (Visa, Mastercard, Discover, JCB), e-wallets, and USDC— so an agent or integration can complete a purchase without a manual checkout.

Multi-currency settlement

Convert and settle in the supplier's local currency and payout globally, so cross-border agent transactions land in the right currency without unnecessary FX.

Multi-asset wallet

A multi-asset wallet gives businesses programmable balances they can hold, convert, and deploy — useful when an automated workflow needs to fund or receive payments repeatedly.

Stablecoin settlement

Businesses can fund their PhotonPay account with stablecoins (USDC or USDT) and use that balance for automated settlement, adding a digital-asset rail alongside fiat for eligible workflows.

Built for ecommerce

Ecommerce businesses exploring AI-assisted and agent-driven selling can treat PhotonPay as the programmable payment layer beneath their storefront, marketplace, or agent integration.

FAQs About Agentic Commerce

What is agentic commerce?

Agentic commerce is an ecommerce model where AI agents assist customers with discovering products, comparing options, and completing transactions. The agent interprets the shopper's intent and helps move the journey forward — from assisted recommendations to, in some cases, authorised purchases made on the customer's behalf within set limits.

How will agentic commerce change ecommerce?

It shifts discovery from keyword search to intent-based selection. Agents read structured product data and choose matches against the customer's constraints, so merchants must optimise feeds and APIs rather than only web storefronts. It also pushes checkout and payment toward software-initiated, automated flows rather than manual human clicks.

What do businesses need to prepare for agentic commerce?

Businesses should make their product data machine-readable (clean attributes, real-time stock and pricing), expose discovery through feeds or APIs, and adopt payment infrastructure that supports API-based, authorised transactions. They should also watch how regulators such as the FCA and PSR treat agent-initiated payments in the UK.

What role do payments play in agentic commerce?

Payments are the part that has to change most. Traditional checkout assumes a human with a card; agents need scoped, revocable authorization, automated payment workflows, global multi-currency settlement, and programmable rails. The payment layer is what turns an agent's selection into a completed, settled transaction.