Picture this: a customer tells an AI agent to book a business-class flight to Dubai, under £900. The agent searches, applies loyalty points, and books the ticket. Sound oddly specific? It isn’t hypothetical anymore.

And here’s the part that should worry you: your brand never sees any of it. No browse event. No session start. No abandoned cart. The order confirmation is the first time you know the sale happened.

Consumer adoption of agentic shopping is on track to jump from 19% to 46% by the end of 2026, according to Braze’s Retail Customer Engagement Review. That’s not a slow rollout. That’s roughly half of your customer base changing how they shop within a single fiscal year.

This isn’t a trend piece. It’s a CMO briefing on agentic commerce in Braze: what it breaks in your program, what to fix first, and how Braze’s AI stack addresses both. 

Let’s break it down. 

What Agentic Commerce actually means 

Agentic commerce is the use of autonomous AI agents to research, compare, and complete purchases through programmatic interfaces, with minimal human input at each step. Compare that to the browse-and-checkout eCommerce your funnel was built around, and you’ll see why this changes things.

An agent doesn’t just surface options and hand them off. It reasons through the purchase. If a price rises mid-session, it adapts. Your bestseller goes out of stock, it substitutes without asking. That’s the “autonomous” part, and it’s also why agents don’t behave like the customers your dashboards were built to track. 

Here’s the piece marketers miss most: Agents transact through APIs, not homepages. They don’t trigger retargeting pixels. They don’t generate session data. Most analytics stacks capture almost nothing until the order confirmation fires.

Three protocols are worth knowing here, because this infrastructure is already live, not something on the horizon:

• MCP (from Anthropic) connects agents to your product and inventory data. 

Autonomy with reasoning. A recommendation engine surfaces options. An agent selects one and acts. If a price rises during the session, the agent adapts. If a product goes out of stock, it routes to the alternative. The agent isn’t just presenting choices; it’s making them, within defined parameters. 

• ACP (from OpenAI and Stripe) standardizes how agents pay and check out. 

Programmatic interfaces, not browser sessions. Agents transact via APIs. They don’t browse your homepage, trigger your retargeting pixels, or generate session data. Your entire digital analytics stack, built around human browsing behavior, captures almost nothing about an agentic interaction until the order confirmation fires. 

• A2A (from Google) governs how agents talk to other agents. 

Interoperability at scale. Several communication standards are already in place: MCP (Anthropic) connects agents to data systems, including product catalogs and inventory; ACP (OpenAI and Stripe) standardizes how agents handle payments and checkout; and A2A (Google) covers direct communication between independent AI agents. 

That’s the plumbing. Now let’s talk about why it matters to you.

The numbers that should be in every CMO’s quarterly review

Metric	Source
Agentic shopping adoption: 19% → 46% by the end of 2026	Braze Retail CER 2026
Only 10% of consumers are willing to let agents act fully independently	Braze Retail CER 2026
71% of marketing leaders say agents have weakened direct customer connections	Braze Retail CER 2026
43% of consumers would stop engaging with a brand if personal data were misused	Braze Global CER 2026
99% of marketing leaders already use AI for customer engagement	Braze, April 2026
Only 5% of AI investments currently deliver a positive ROI	MIT, 2025
$3–5 trillion projected global agentic commerce value by 2030	McKinsey

The gap between the first and last stats is the real story. Almost every marketing team has adopted AI. Almost none is extracting meaningful commercial value from it. Agentic commerce is about to make that gap significantly more expensive to ignore with Braze coming into the picture.

How Agentic Commerce works: The mechanics marketers need to understand 

There are three stages to it. 

The buying journey becomes a conversation

Instead of a search query initiating a funnel, a natural language goal initiates an agent workflow. “Find me a sustainable gym kit under £80 that ships before Thursday” becomes a structured instruction the agent parses, executes across data sources, and resolves to a transaction.

If the request is too broad, the agent asks for more information. It also draws on stored preferences, past purchases, and remembered sizes, so the same prompt gets smarter results over time. The buying session becomes cumulative. The agent learns. Traditional new-visit retargeting logic becomes irrelevant.

Autonomous execution replaces the checkout flow

Once a goal is set, the agent moves through a multi-step workflow, scanning retailers, comparing prices in real time, checking inventory, applying available discounts, and completing the transaction. Agent autonomy can be designed via a tiered structure: routine or low-value purchases can run fully automated. Higher-value transactions might require the person to approve before the final step.

The implication for brands: the customer’s willingness to allow autonomous execution depends on trust. Trust is built through consistent, relevant, respectful direct brand interactions over time. In other words, your lifecycle marketing program today is directly influencing how much latitude customers will give agents to act on your behalf tomorrow.

Agentic payments are already live

OpenAI and Stripe’s Agentic Commerce Protocol (ACP) powers ChatGPT’s instant checkout, allowing purchases to be completed directly within a chat interface. Google’s AP2 verifies that an agent is genuinely authorized before a transaction goes through. Mastercard’s Agent Pay does the same across its global network. Stripe also generates a temporary card number for each agent transaction, so the user’s actual payment details are never passed to a retailer.

This isn’t theoretical infrastructure. These systems are live, in production, processing real transactions. The payments layer has moved faster than most marketing teams’ awareness of it. 

Where Agentic Commerce is already playing out (real-life examples)

Retail and eCommerce

The most immediately disrupted sector. In retail, agents are already handling recurring purchases, real-time price comparison, and cross-channel fulfillment coordination. Monitoring household staples and reordering automatically within set budgets. Comparing prices across multiple retailers in real time before committing, and coordinating online ordering with in-store pickup availability.

For retail brands on Braze, the critical question is: Does your post-purchase Canvas fire on a transaction event, regardless of how the transaction was initiated? If it only fires on events that your SDK triggered, agent-initiated purchases may slip through entirely.

Travel and hospitality

In travel, agents handle the full booking workflow and the moments when plans fall apart, automatically rebooking when conditions change, and processing refunds within pre-approved limits without human intervention. The rebooking moment is where customer loyalty is built or lost. 

Dayuse, the global hospitality platform, is already ahead of this. After deploying the BrazeAI Agent Console™ to generate individualized messages at the moment of booking confirmation, pulling real-time context including booking history, stated preferences, and language, Dayuse saw a 90% increase in booking conversion rate and an additional 23% uplift in repeat engagement. 

Subscriptions and digital services

In subscription services, agents monitor usage and either optimize plans or switch providers on the user’s behalf. An agent can switch a customer to a competitor as easily as they can renew them. Retention depends on offering demonstrably better value.

B2B and procurement

61% of procurement leaders cite geopolitical and supply risks as their top concerns. Agents respond to exactly those risks in real time, identifying and activating alternative sourcing when supply disruptions occur, and making decisions that used to take days in only minutes.

For B2B brands on Braze, this raises the question of whether your Canvas journeys are scoped for the actual decision-maker vs. the agent that may be conducting the vendor evaluation on their behalf. 

The two problems Agentic Commerce creates for brands

When we work through agentic commerce readiness with Braze clients at Mavlers, we find the challenge almost always breaks down into two distinct problems, and most teams focus on only one.

Problem 1: Discoverability: Can agents find you?

This is Generative Engine Optimization, or GEO: making sure your product data is machine-readable, your attributes are standardized, and your inventory is accurate in real time. If an agent can’t parse your catalog, it doesn’t fail politely. It skips you and deprioritizes you next time. 

Most brands are beginning to think about this. Fewer have solved it.

Problem 2: The relationship layer: What survives the handoff?

The relationship layer. This one gets less airtime, and it’s the sharper problem. Remember that 71% stat? When an agent completes the transaction, the browse session never happens. The cart event never fires. The checkout page never loads. The order confirmation becomes the first touchpoint you actually own. 

Think about what that means for your Canvas architecture. If your welcome flow only triggers at the start of a human session, an agent-acquired customer falls through the cracks entirely. So ask yourself: does your onboarding flow fire on anything other than a human-triggered session start? If the answer is no, you already have a blind spot. 

“No browse session, no welcome flow trigger, no on-site behavioral data to power your retention program. The lifecycle layer is the part of your strategy that determines whether an AI-acquired customer ever buys again.”

Three challenges CMOs face that no one is talking about honestly

Your data is probably not agent-ready

If a retailer’s product catalog spans multiple systems with inconsistent attributes, incomplete specifications, or outdated pricing, AI agents can’t evaluate those products reliably. Fragmented data limits both discoverability and interoperability.

The same problem exists inside Braze. Agents making personalization decisions based on user context need well-populated profiles. Sparse custom attributes, missing event data, and incomplete catalog fields all degrade agent output, because agent quality is directly proportional to the richness of the data it can access. A data audit is the first step before any meaningful agent deployment.

Consumer trust hasn’t caught up with the technology

Only 10% of consumers are willing to let those agents operate without oversight. 43% of consumers would stop engaging with a brand entirely if their personal data were misused. The technology is ahead of consumer comfort, and that gap matters commercially. Brands that push agentic experiences on customers who haven’t yet opted in to agentic delegation will damage the trust they’re trying to build. Moving at the customer’s pace, not the technology’s, is a strategic choice, not a concession.

Brand errors in agentic contexts become brand failures

If agents hallucinate, quote outdated pricing, or cite unreliable sources, customers don’t see a system error. They see a brand failure. The brand owns the outcome even when an agent causes it. This means quality control inside agentic workflows isn’t a nice-to-have; it’s a commercial necessity. Every agent deployed in a customer-facing context needs defined guardrails, fallback logic, and, for high-stakes interactions, a QA agent reviewing outputs before they reach the customer. 

How Braze powers the agentic commerce layer: A practitioner’s breakdown

Here’s the part that matters most: how do you actually fix this inside Braze? Each piece maps directly back to the two problems we named earlier.

1. BrazeAI Agent Console™ is the central environment for building custom agents inside your Catalogs and Canvases. It generates content in real time at the moment of send, routes users based on qualitative signals instead of rigid boolean logic, and turns free-form conversational replies into structured profile attributes.

2. BrazeAI Operator™ is the conversational, in-dashboard layer that makes agent configuration possible without a prompt engineer on staff. That matters most for agency-supported teams. The client describes what they need, the operator scaffolds the build, and the agency tightens it for production. That’s exactly how we work with our Braze clients.

3. BrazeAI Decisioning Studio™ sits on top of Agent Console and optimizes channel, timing, and content variant using reinforcement learning. Luxury Escapes used Agent Console on its own to replace manual segmentation with an AI agent that weighed ten behavioral signals at once, and it delivered a 10% lift in revenue per user, driven entirely by conversion, plus a 7% increase in total transaction value (Source: Braze customer case study; Braze Q1 FY27 earnings call, May 2026).

Two more pieces close the loop. 

• The Braze–ChatGPT integration gives your brand a presence, product carousels, and storefront details inside the agent’s native environment, which addresses discoverability head-on. 
• And cross-channel orchestration across email, push, in-app, SMS, and WhatsApp fires a post-purchase Canvas within minutes of any transaction, whether human- or agent-initiated. That’s the mechanism that turns an agent-mediated sale into a direct relationship.

Wrapping up 

That brings us to the business end of this article, where it’s fair to say that agentic commerce in Braze doesn’t erase the value of lifecycle marketing. 

It concentrates that value into the post-purchase relationship, because the pre-purchase journey now largely belongs to an agent, not your website. 

So here’s the way forward: Audit your Canvas architecture for anything that only triggers on human behavior, and audit your catalog for anything an agent can’t read. 

The brands with a structural advantage over the next two years are the ones treating direct customer relationships and agent-ready data as a strategic asset now, before agentic volume forces the issue. 

If you’d like a second set of eyes on how your Canvas architecture and catalog readiness hold up against agentic commerce, our team at Mavlers is always happy to take a look. Let’s talk.