A guide to agentic commerce: How AI shopping agents are reshaping brand loyalty

And according to Harvard Business Review & Talon.One research, 77% of executives say loyalty programs are extremely or very important to company leadership.

Brands have never cared more about loyalty. But the mechanism through which loyalty gets expressed is about to change fundamentally.

Agentic commerce is when autonomous AI systems handle the shopping journey on behalf of customers. They research, compare, and complete transactions with minimal human input, covering each step from discovery through checkout.

What makes this different from earlier forms of automated shopping is agency. These systems don't wait for a human to approve each step. They set a goal, evaluate options against structured criteria, and execute. 

For brands, that means purchase decisions are increasingly being made by systems that can't read your homepage, feel your brand story, or respond to emotional cues. They can only act on what they can query.

Major commerce platforms are already moving. Shopify has released its first agentic commerce capabilities, letting agents help buyers find products, add them to cart, and check out without leaving the experience. 

On the infrastructure side, Google and Shopify announced the Universal Commerce Protocol (UCP) at the National Retail Federation (NRF) in early 2026, an open-source shared standard for how AI agents, commerce platforms, and merchant systems work together. 

It followed OpenAI's release of the Agentic Commerce Protocol (ACP) in late 2025. While OpenAI has since dropped plans for a direct checkout inside ChatGPT, a recent report from the tech giant on user behavior found that people were increasingly using ChatGPT to explore products, compare options, and gather information.

This echoes what we are seeing here at Talon.One: agentic AI is already reshaping retail. But right now, most retailers have no way of controlling which loyalty program functionality and promotions get surfaced during the agentic shopping journey. The result is that agents often optimize for price, not the outcomes retailers actually care about: 

That's the core challenge. Your incentives need to be structured, centralized, and accessible in real time, or the agent can't factor them in. Whether your loyalty program can function within these new rails is now a real competitive question.

When agents make the purchase, the rules for being chosen change. Here's how.

Brands built most loyalty strategies to win a human's attention at the moment of choice. That mechanism is changing. Agents don't have habits, nostalgia, or emotional affinity. They extract structured data and make parameter-based decisions. Your brand story and the feeling customers get from your store? An agent can't process that.

Stephan Ritter, Director at Deloitte Digital, described where this is headed at Talon.One's INCENTIVIZE summit: "An AI agent will know a customer so well that it will streamline the entirety of anticipation of needs, discovery, and transaction into a single flow. The customer no longer needs to shop; the AI shops for them."

When that becomes the norm, brand value needs expression in formats agents can read, not just formats humans can feel.

Traditional loyalty mechanics, including point balances, tier benefits, and member-exclusive offers, are invisible to AI agents unless teams expose them as structured, machine-readable data.

When a customer's agent evaluates running shoes across retailers, it needs fast answers: 

If that information isn't available through an API the agent can query in milliseconds, the agent can't factor it in. The offer effectively doesn't exist.

Identity linking changes this. UCP introduced a capability that lets agents "sign in" to a retailer's system on behalf of a customer. That opens up loyalty earning and redemption through agent-led purchases and allows agents to surface personalized offers based on member eligibility. 

Without identity linking, every agent transaction is anonymous, and the loyalty value attached to that customer vanishes.

When agents handle discovery and purchasing, the moments where brands build loyalty preference shift downstream. Loyalty outcomes depend more heavily on post-purchase experiences like delivery reliability, support quality, and returns handling.

This creates a compounding dynamic. When an agent learns from satisfaction signals over time, loyalty value gets embedded in its decision logic rather than surfaced at checkout. Brands that invest in post-purchase quality build an advantage that grows. Brands that don't get quietly deprioritized.

Cart-native loyalty helps prevent loyalty value from disappearing mid-journey. It surfaces points, rewards, and member benefits throughout the cart flow in formats that both apps and agents can evaluate at the moment of purchase.

In agentic commerce, infrastructure is what makes great loyalty experiences possible. A well-designed program that agents can't read doesn't reach the customer. The sections below cover what agent-ready loyalty infrastructure looks like in practice.

When teams spread customer incentives across promotional tools, loyalty software, ecommerce platforms, and CRM systems, humans can work around the fragmentation. Agents can't. Łukasz Słoniewski, CEO at loyalty and martech agency Omnivy, frames it plainly: "Most organizations don't have a single source of truth. What they have is spaghetti. So 'connect to UCP' quickly turns into 'connect UCP to your chaos.'"

Luke Effenberger, Director of AI Solutions at Talon.One, puts the design requirement directly: "Agentic commerce requires fast and clear communication. AI agents need incentives that are structured, centralized, and available in real time, or they simply won't apply them."

UCP establishes a framework for how agents, commerce platforms, and merchant backends interact. It introduces 6 core capabilities, including product discovery, identity linking, checkout, and order execution, and gives merchants a way to expose structured data that agents can reliably act on.

For promotions, UCP provides a Discount Extension that governs how agents find eligible discounts, display them to users, and apply them at checkout. For loyalty, identity linking is the enabling mechanism: it lets agents associate a user with their merchant account and surface relevant rewards, balances, and personalized offers.

What UCP doesn't cover is the full range of modern loyalty mechanics. That's the gap Talon.One's Unified Incentives Protocol (UIP) addresses. UIP is a set of platform-agnostic standards that expose loyalty and promotion methods across all phases of the shopping lifecycle. 

The first building block is a live UCP loyalty extension that gives agents visibility into what matters at the moment of purchase: 

It also supports card-based programs, so loyalty use cases work even without identity linking. More extensions covering personalized promotions and offer management are in development.

Static loyalty programs are a poor fit for agents who re-evaluate options on every transaction. As agentic commerce scales, many loyalty strategies will shift toward real-time, context-aware rewards that calculate and apply benefits based on current conditions: 

This requires incentive decisioning that operates at machine speed. Batch loyalty updates that take hours to propagate won't work when an AI agent may evaluate your offer against 10 competitors in the same second.

Agentic commerce introduces a measurement layer most brands haven't built yet. When an AI agent executes a purchase, trust may attach to the agent or its ecosystem rather than to the retailer. That means tracking 2 distinct layers.

Human loyalty metrics cover emotional affinity, direct channel engagement, and the explicit preferences customers encode in their agents. 

Agent loyalty metrics are a newer territory. This includes generative engine optimization (GEO) discoverability scores, brand mention frequency in AI recommendations, and conversion differences in agent-mediated purchases compared to human-initiated ones.

Every loyalty KPI now needs channel attribution that can distinguish between human-initiated and agent-initiated transactions.

You don't need to rebuild everything. Start with an honest audit of whether your current stack is consumable by systems that don't click through your app, and go from there.

Here's how to approach agentic commerce across five key actions.

The starting point is simple: can an external system query your loyalty program rules, a customer's eligibility, and your available promotions through an API and get a response in under 200ms? If the answer is no, agents can't factor your incentives in.

The audit covers 4 areas:

Gaps in any of these areas mean agents are making decisions about your brand without seeing your full value proposition.

Agentic commerce doesn't tolerate fragmentation. Many teams still manage incentives across a promotions platform, separate loyalty software, ecommerce campaigns, point-of-sale systems, and CRM. Humans can work around that complexity. Agents default to price when incentives are unclear or unavailable.

Forward-looking brands have centralized promotion and loyalty decisioning into a single system, giving every channel, whether human or agent-led, the same rules to follow. The result is better margin control and more predictable outcomes across the board.

Agents compare discounts quickly and precisely. Brands that compete on price alone in agent-led environments get commoditized fast. 

The ones that win are those building loyalty value that goes beyond monetary savings: tier status, access, recognition, and experiences that agents can factor into decisions without triggering a race to the bottom. 

Non-financial rewards are particularly powerful here, as they differentiate without eroding margin.

In agentic commerce, identity linking is how brands stay connected to their customers through agent-mediated transactions. 

Sabina Radziush, Product Lead at Talon.One, explains the stakes: "Without UCP's identity linking, agent purchases become anonymous transactions. With it, merchants retain ownership of the customer record and allow agents to apply the right benefits in real time. For commerce teams, the action is clear: use loyalty as the incentive to drive identity linking early, or risk losing visibility as agentic shopping scales."

Leading brands are making opt-in worthwhile by rewarding customers for identifying themselves and granting consent across every channel, including agent-led shopping.

Agentic systems need access through endpoints, not user interfaces. That makes composable architecture less of a nice-to-have and more of a prerequisite for agentic commerce readiness. 

The MACH Alliance (Microservices-based, API-first, Cloud-native, Headless) benchmark shows 77% successful AI deployment for mature MACH adopters, compared to 36% for those new to MACH. That 2x gap shows a structural advantage. When your stack is API-first by design, making incentives agent-readable is a natural extension of how it already works.

If your promotional logic lives in spreadsheets or depends on manual interpretation, agents can't consume it. Incentive rules need a standardized, machine-readable expression to function in agent-led environments.

Agentic commerce doesn't require a full rebuild. It requires the right foundation: incentives that are centralized, machine-readable, and fast enough for real-time agent decisioning. Brands that get there first won't just be agent-ready. 

They'll have cleaner margin control, faster campaign execution, and a loyalty program that works across every channel, human or agent-led.

The practical starting point is consolidation. If your promotions, loyalty, and offers live across multiple systems, fix that first. Agents can't reason over fragmented incentives, and neither can your own marketing team.

Talon.One unifies loyalty, promotions, and gamification into a single incentives infrastructure built for exactly this kind of operational precision. Marketing teams can build, test, and launch campaigns without filing engineering tickets, while the platform processes incentive logic at sub-50ms. 

With the Unified Incentives Protocol, Talon.One extends that infrastructure directly into agentic channels, giving agents visibility into member balances, tier status, and earning logic that would otherwise be invisible to agent-led shopping.

What is agentic commerce? 

Agentic commerce is when autonomous AI systems research, compare, and complete purchases on behalf of customers. They can act end-to-end with minimal human input.

How does agentic commerce affect loyalty programs?

If you don't expose loyalty program benefits through machine-readable APIs, agents can't factor them into recommendations. That can make loyalty program mechanics invisible to the AI agents evaluating purchase options on a customer's behalf.

What do API-first loyalty programs mean? 

API-first loyalty programs mean teams expose loyalty program benefits, rules, and entitlements through structured endpoints. AI shopping agents can discover, evaluate, and apply them in real time.

How should brands prepare their loyalty programs for agentic AI? 

Audit your stack for agent-readiness, move toward MACH-aligned API-first architecture, standardize rules for third-party consumption, invest in real-time decisioning, and build governance frameworks.