Three Actions That Turn Agentic Commerce from Ambition into a Shipped Experience

Key takeaways 

• Agentic commerce standards are still consolidating. Google's Universal Commerce Protocol (UCP) drew broad commercial backing in 2026, yet competing protocols aren't going away. Retailers shouldn't wait for a single winner to get started with agentic commerce.
• Readiness comes down to three actions: a use case that removes real purchase friction, data an agent can read and act on, and governance for autonomous payments.
• The back-end data layer now demands the same attention as any customer-facing surface. When an agent shops for a customer, your catalog, product information and APIs are the experience. What isn't encoded in the data doesn't exist to the agent.
• In WWT's work with retail clients, most organizations rank low on a data maturity and governance scale. That gap is why broad rollouts stall and a narrow starting point wins.

The standards won't settle soon, but that's no reason to wait 

Agentic commerce is the shift from consumers searching, comparing and buying products to AI agents doing that work on their behalf. Most retailers understand the importance of agentic commerce to their strategies. According to NVIDIA's State of AI in Retail and CPG survey, nine in 10 retailers plan to increase their AI budgets in 2026, with agentic AI among the top priorities. The harder question is operational, and it lands with the teams who own the architecture: What does preparing for agentic commerce actually require?  

For much of 2025, retailers were reluctant to invest heavily in agentic commerce infrastructure because the standards landscape remained unsettled. OpenAI and Stripe promoted Agentic Commerce Protocol (ACP), while Google advanced Universal Commerce Protocol (UCP) as a broader framework spanning discovery, checkout and post-purchase interactions. That uncertainty is starting to ease, though it is far from settled. Although ACP continues to support commerce experiences within the OpenAI ecosystem, much of the broader retail market has rallied around UCP. Google's launch with Shopify at NRF 2026 was followed by Amazon, Meta, Microsoft, Salesforce and Stripe joining the UCP Tech Council, a strong signal of where momentum is gathering. However, we predict the market will likely stay multi-standard for the foreseeable future. 

Waiting for a single standard to win isn't a strategy. The readiness work now sits squarely with retailers' own architecture, and it comes down to three actions.  

Action 1: Identify a use case with real friction to remove 

Not every experience needs to be agentic. Retailers should start by identifying whether a specific purchase journey carries enough friction that automating it delivers real business value and a noticeably better experience. The best candidates remove effort that customers actively dislike and return something measurable: a larger basket, a faster path to purchase, higher conversion on a considered buy. 

The right fit is driven by how people shop. Three behavioral signals point to where agentic commerce pays off: 

• Research friction: Purchases that demand comparison (e.g., furniture, electronics, appliances and travel) reward an agent that can search, filter and shortlist across many options.
• Repetition and planning load: Recurring purchases with real cognitive overhead, like weekly grocery and meal planning, reward an agent that remembers preferences and plans ahead.
• Delegation comfort: Shoppers readily hand off routine, lower-risk decisions but keep control of high-value, high-emotion or luxury purchases. The more personal or expensive a purchasing decision is, the less likely a customer will trust an agent to make it.

A strong use case has all three signals working together. Take a shopper configuring a complex purchase like a home theater setup or a gaming PC. The shopper provides a budget, the features they care about, and their space or compatibility constraints. The agent assembles a working build, confirms real-time availability, prices it and then returns a cart to the shopper they can adjust. The shopper has saved hours of cross-referencing specs, resulting in a larger, more confident purchase. That's the profile worth building around. 

Where to start

Score candidate journeys against the three signals: research friction, repetition and delegation comfort. Choose the one where removing friction creates clear business value, then confirm the behavior rewards memory and personalization before you scope it. 

Action 2: Structure your data so an agent can read and act on it 

Data readiness is where most of the 2026 work sits. An agent parses data, not pages. That means structuring product information as machine-readable fields with consistent SKUs and complete attributes and context, plus having agent-ready APIs that expose real-time inventory, pricing and fulfillment. Miss those and the agent either fails or misrepresents the brand. 

The shift most retail IT teams haven't made is that the back-end data layer must now be treated as a customer experience surface. A clean front-end interface can hide a messy data model from a human shopper. An agent removes that cover, surfacing the messiness straight to the customer. This is why some retailers are, for the first time, running experience design against their API libraries. You are now designing for two users at once: the human browsing the app and the agent crawling the catalog. Build this layer well and you are protocol-agnostic by design, because the same clean data serves your owned experiences and any third-party agent that comes calling.  

Where to start 

Run an agent-readiness diagnostic first: Can a bot reach and transact against your site today? Score catalog structure, API accessibility and data completeness, then close those gaps before building any experience. 

Action 3: Design governance for autonomous payments 

The moment an agent can spend money on a customer's behalf, a new class of risk opens. The agent needs its own identity, spending limits and authentication that a payment network will honor, so a legitimate agent-initiated charge isn't treated as fraud. This trust layer is forming quickly, with Mastercard's Agent Pay for Machines and agent-identity work from Stripe and Cloudflare.  

Governance reaches beyond payments. As an agent builds context about a shopper, it can pull in far more personal information than a keyword search ever did, up to health-adjacent details a retailer arguably should not hold. Add the agent attack surface, from prompt injection to unauthorized purchase, and governance becomes a design input rather than a post-launch checkbox. 

Where to start 

Before you deploy, define what an agent may do, what data it retains and how that context is wiped, along with your attack-surface guardrails. Retrofitting governance onto a live agent is far harder than designing it in. 

How can we help you? 

The retailers that successfully move forward with agentic commerce adopt a narrow focus. They start with one use case that solves for real purchasing friction, scoped so that every system and data dependency it touches can be modernized and tied to measurable outcomes. Once that use case is proven out, then they expand. 

In our work with retailers, the gap is rarely vision. It's the foundation under it. We can help you pinpoint that first use case, define the minimum data and architecture required to support it, and prove the model before scaling.