Agentic commerce: The next evolution of AI in retail

What is agentic commerce?

Agentic commerce refers to the use of AI agents that autonomously search, discover, compare, and purchase on behalf of users.

Unlike chatbots that simply execute tasks when prompted, AI agents have initiative. These sophisticated AI systems can understand intent, reason through context, and execute multi-step tasks without human intervention. This allows them to make hyper-personalized recommendations, navigate systems, and adapt to users on the fly.

Imagine having a personal shopper who’s available 24/7, learns from every interaction, and processes huge amounts of data instantly to assist you at every step of the buying journey.

These AI agents for retail can serve as:

• Personal shopping assistants that drive product discovery or reorder household essentials automatically

• In-store support that guides shoppers with real-time aisle information, stock checks, and product comparisons via kiosks or mobile apps

• Customer service agents that handle basic queries, returns, WISMO, order modifications, and route intelligently if needed

• Merchandizing optimizers that deliver personalized promotions by segment, behavior, or stock levels

• Procurement agents that identify suppliers, evaluate quotes, source materials, and manage complex B2B purchasing workflows

The key difference between AI agents and previous AI tools is agency. Rather than just recommending from static listings, they research and act independently on behalf of users. This leads to a faster, more convenient customer experience—but what’s in store for retailers, distributors, and manufacturers?

The impact of agentic commerce on retail

AI search mode: From browsing to delegation with agentic ecommerce

AI agents do more than search—they synthesize. This means consumers won’t need to compare products across sites, but will instead rely on agents for an internet-wide roundup of options for even basic purchases. According to Harvard Business Review, experts predict AI agents could replace Google within four years, as AI usage and functionality continue to grow beyond that of traditional search interfaces.

This will make AI agents the front door for online retail. As they aggregate product information and steer buying decisions, they decouple users from direct retailer channels.

This will create new dynamics:

• Agent selection criteria will determine market share

• Direct traffic will drop as agents bypass front-end experiences entirely

• Product visibility will rely on structured product data that agents can parse and compare

In this new competitive landscape, AI agents will become the interface for online shopping. Exemplifying this, Amazon's new "Buy for Me" feature allows AI agents to purchase products from third-party websites while users remain on Amazon.

Agentic commerce for B2B manufacturers and distributors

B2B commerce is well-suited to AI agents for retail. According to Bain Capital Ventures, the repetitive ordering, complex specifications, and multi-stakeholder approvals common in B2B environments are ideal for AI agents, which thrive on structure and consistency.

Leading payment providers have already invested billions in making agent-led purchasing possible. In April 2025, Visa introduced Intelligent Commerce, while Mastercard announced its pilot for Agent Pay—platforms both designed to enable agents to act autonomously on behalf of humans within defined guardrails.

Soon, we may see the rise of an agent-to-agent (A2A) economy, where AI systems representing both buyers and sellers handle an increasing share of price negotiations, inventory checks, fulfillment, or transaction workflows with little to no human oversight. This would reshape how manufacturers and distributors reach customers, while creating new efficiencies.

The CX opportunity of agentic commerce: Sales and support on every channel

Agentic commerce represents a fundamental shift in how brands will build relationships and support customers. AI agents can operate across formerly fragmented systems, uniting touchpoints into one seamless omnichannel journey. As they interact across channels—web, mobile, SMS, email, in-store kiosks—they gather and unify data into a single, evolving customer profile.

This enables a new standard of support:

• 24/7 omnichannel support: Customers get instant 24/7 help on every channel, where agents handle complex workflows (product questions, WISMO, returns, renewals) from start to finish, and use multimodal capabilities like voice AI.

• Smarter escalations: If a handoff is needed, agents deliver full conversation history, user context, and next actions to eliminate friction, reduce resolution times, and boost satisfaction.

• Personalized service at scale: With access to CRM, inventory, and behavioral data, agents can provide a more tailored experience that drives loyalty and sales, even identify upsell or cross-sell opportunities, and trigger promotions in real time.

Agentic commerce will allow retailers to scale high-quality experiences where intelligent support, purchasing, and returns are available to customers anytime, anywhere.

The product data challenge in agentic commerce

Unlike humans, AI agents don’t scan product pages or respond to marketing messages. Instead, they make decisions based on structured, machine-readable data they access through APIs to evaluate product information such as availability, pricing, ratings, and more.

To ensure AI agents can find and recommend products, organizations must adopt a more sophisticated approach to product data that goes beyond traditional SEO. It requires optimizing product content and data for AI agent ontology (AAO), supplying agents with the context they need to evaluate product fit, value, and make context-driven decisions that align with user intent.

For example:

The key difference here lies in addressing user intent. As an LLM-powered technology, AI agents perform best with context. This means product descriptions should connect features to real-world use cases and key value signals for users. This is why the agent-ready description above not only explains the product, but its suitability in different contexts (long distances, rainy climates, spring training, or all-weather daily wear).

Product images, videos, and media assets also need to be readable to AI agents for retail through structured metadata. For example, describing a shoe as "Women’s waterproof red running shoes designed for road and trail running in wet weather” rather than simply "red running shoes for women" helps AI agents understand both the aesthetic and the appropriate context.

If product data quality is poor, there are potentially massive costs in the agentic economy:

• Loss of revenue because agents can’t “see” products to recommend
• Poor recommendations that arise from ambiguous or incomplete product attributes
• Broken automations due to poor data quality that stall fulfillment

Businesses already lose an average of $15 million annually due to poor product data, a number that will balloon as agents become the primary path to purchase.

How to prepare for agentic commerce

Agentic commerce is no longer experimental—it’s fast-becoming the foundational infrastructure of a new era of AI in retail. Concurrent to top payment providers developing agent-friendly infrastructure in April 2025, OpenAI updated ChatGPT's web search capabilities to allow personalized product recommendations and direct purchase links.

With adoption accelerating, here are a few best practices to consider as you create your AI strategy for agentic commerce:

1. Optimize product data

AI agents rely on structured, machine-readable product data to operate effectively. To support this, retailers must audit their product catalog to identify data gaps and inconsistencies. Next, organizing product information with standardized classifications, attributes, and metadata is essential. By adopting consistent naming conventions and aligning with emerging agent-oriented ontology (AOO) standards, you enable interoperability and visibility across AI agent platforms.

To ensure AI accountability and drive improvement, businesses can implement data quality scoring across their catalog. This creates measurable benchmarks that align with the structured data requirements AI agents depend on. A product information management (PIM) system plays a critical role in this process, helping retailers manage, normalize, and enrich product data from multiple sources—ultimately improving data accuracy, consistency, and AI agent performance.

2. Develop agent-friendly infrastructure

For AI agents to act independently, your infrastructure must speak their language: API. By exposing product data and functionality through APIs, you give agents access to enable everything from product discovery to pricing, availability, and order execution.

An API-first approach ensures every system and channel reflects real-time data, which eliminates delays and desynchronization that lead to poor outcomes. It also sets the foundation for integration with agentic commerce leaders. Tools like PayPal’s Agent Toolkit and frameworks like OpenAI’s Agent SDK, LangChain, and Vercel’s AI SDK depend on robust, well-documented APIs to retrieve data and trigger business logic.

Access alone isn’t enough, though. Product content must be machine-readable. Implement structured data markup (like Schema.org) across your catalog to ensure agents can parse and act on listings. If they can’t read your products, they can’t recommend or buy them—leaving your brand invisible to the AI-powered future of retail.

3. Build AI-ready content and context

For your product content to surface in agent-led journeys, its description must not only be structured but also enriched and contextual. This means going beyond SEO keywords:

• Supply structured product metadata (prices, stock, specs, reviews)

• Build signal-rich interfaces that agents can interpret easily

• Focus on contextual attributes like product quality, reviews, or sustainability that help AI evaluate fit, function, and value

• Treat metadata as a marketing channel with standardized attributes and use-case indicators that help with agentic product discovery and recommendation

When you make products agent-ready, they’re not just discoverable—they’re ready to convert.

4. Train teams for AI-driven sales

To succeed in the era of agentic commerce, sales and marketing teams must understand how AI agents differ from human shoppers. This means investing in training that covers data quality standards, structured listing requirements, and the shifting dynamics of digital discovery. Unlike humans, AI agents respond to storytelling or brand loyalty—they prioritize structured, machine-readable signals of value. As a result, niche or challenger brands with stronger data hygiene may outperform larger household names.

Preparing for this shift means adapting go-to-market strategies to favor structured data over traditional messaging and narratives. It also means equipping internal teams with the tools to create and manage agent-optimized product content.

For manufacturers and distributors, enabling seamless data flow through retail partners is equally critical. Providing templates and self-serve tooling for agent-friendly listings helps standardize information across the supply chain and ensures products are both visible and competitive in agent-driven environments.

Getting started with agentic commerce

Agentic commerce is a paradigm shift in how brands sell to and support their customers. As AI agents for retail increasingly take on the work of researching, comparing, and purchasing for people, businesses that adapt quickly will remain visible, responsive, and trusted—while laggards fall by the wayside.

With leading payment platforms already investing in agent-friendly architecture, it’s no surprise that Gartner predicts that 33% of enterprises will deploy AI agents by 2028.

To succeed, retailers and ecommerce leaders must anchor their strategy around two key pillars:

• Discoverability: Product data must be structured, enriched, and machine-readable so agents can find and compare your offerings.

• Scalability: Your infrastructure must support real-time, omnichannel AI agent interaction without latency or fragmentation.

Sendbird’s AI agent platform delivers on both fronts.

Offering enterprise-grade infrastructure trusted across 7 billion+ conversations each month, built-in agent-first APIs, and a robust suite of features for AI trust and safety—we help retailers build, evaluate, and scale the next generation of AI-powered customer experiences.

To learn more, contact sales or book a demo.