The Current State: From Browsing to Buying with AI

We're witnessing the early stages of a fundamental shift in how people shop. What started as simple chatbots answering customer service questions has evolved into sophisticated AI agents capable of making purchasing decisions on behalf of consumers. The evidence is mounting across major platforms:

Amazon's "Buy for Me" Agent represents perhaps the most aggressive move into agentic commerce. Amazon is testing an AI agent that shops third-party sites when Amazon doesn't sell something, powered by Amazon's Nova AI models and Anthropic's Claude. This isn't just a recommendation—it's active procurement.

Shopify's Agentic Commerce Platform uses the Model Context Protocol (MCP) to let AI agents like ChatGPT and Claude interact directly with merchant storefronts—searching products, managing carts, and handling checkout across their entire ecosystem. This MCP infrastructure democratizes AI shopping, allowing small boutiques to offer the same agent-powered experience as major retailers.

Why Agents Will Reshape the Funnel

The traditional marketing funnel—Intent → Awareness → Conversion → Payment → Loyalty—assumes human decision-making at every stage. AI agents are changing this equation dramatically, but not uniformly across all stages.

High Influence Zones: Intent and Awareness

AI agents excel at the top of the funnel, where their influence is strongest and most valuable to businesses. Unlike traditional marketing that waits for consumers to express intent, agents can proactively predict and even create needs:

• Predictive Intent Generation: "Based on your calendar, you have a work trip next week. Should I book your usual hotel and order travel-sized toiletries?"
• Cross-Category Discovery: "Your running shoes are due for replacement based on your weekly mileage. I found three options that match your gait pattern."
• Latent Need Recognition: "Your grocery patterns suggest you're cooking more Asian cuisine lately. These specialty ingredients are on sale this week."

This represents a fundamental shift from reactive commerce (responding to expressed needs) to predictive commerce (anticipating unstated needs).

The Conversion Sweet Spot

At the conversion stage, agents provide moderate but sophisticated influence. They excel at optimization tasks—finding the best specifications, comparing alternatives, and analyzing value propositions. However, this is where the question of who designs the agent becomes critical.

Trust: The Missing Layer

Here's where the agent revolution hits its biggest obstacle. Payment represents the irreversible decision point—the moment where recommendation becomes financial commitment. The Salesforce data reveals the critical trust factors: shoppers rank data privacy and security protections, the ability to easily turn agents on/off, and requiring approval before any purchase as their top requirements for trusting AI agents.

The friction isn't just psychological; it's practical. The generational divide is stark: Gen Z shoppers are 2.7x more likely than baby boomers to want product recommendations from AI agents (63% vs 23%), yet even among younger users, trust mechanisms remain essential.

Users develop personalized trust thresholds based on historical accuracy of agent recommendations. A user might fully automate $20 grocery purchases but require validation for anything over $100, or any purchase in an unfamiliar category—similar to how developers review AI-generated code before deployment.

Post-purchase, agents regain high influence through automated reordering, predictive restocking, and loyalty optimization. This is particularly powerful for habitual purchases where decision fatigue is high and brand switching costs are low.

‍

Who's Actually Making the Decisions?

The most critical question in agentic commerce isn't technological—it's economic. Who owns and operates your shopping agent determines what you buy. The current landscape reveals four distinct types of agents, each with fundamentally different incentives:

Marketplace Agents (like Amazon's "Buy for Me") optimize for platform metrics. Their primary goals are maximizing gross merchandise volume (GMV), pushing higher-margin products, optimizing inventory turnover, and keeping users within their ecosystem. When Amazon's agent recommends a product, it's likely considering not just your needs, but also Amazon's business objectives—inventory levels, profit margins, and competitive positioning.

LLM Provider Agents (like ChatGPT shopping or Claude) optimize for user satisfaction and continued engagement. Their revenue comes from API usage, subscriptions, and maintaining user trust rather than individual transactions. This creates different incentives—they're more likely to recommend the objectively best product for your needs, even if it's not the most profitable for any single retailer.

Third-Party Specialized Agents focus on niche optimization and values-based shopping. A sustainability-focused agent might only recommend products meeting specific environmental criteria, while a price-optimization agent might always choose the cheapest option. These agents typically operate on commission models or subscription fees, creating yet another set of incentives.

Hybrid/Custom Enterprise Agents serve business customers with customized logic, workflow integration, and data privacy controls. These agents align with specific organizational procurement policies and approval workflows.

The implications are profound: the same shopping request could yield entirely different recommendations depending on which type of agent you're using. A marketplace agent might suggest a higher-margin alternative, while an LLM provider agent might recommend the most popular option, and a specialized agent might prioritize sustainability or price above all else.

The Agent Economy: New Rules, New Revenue

This shift to agentic commerce creates entirely new revenue streams and business models that didn't exist in traditional e-commerce:

For Retailers, the opportunities include agent-specific pricing (offering different prices to agents versus human shoppers), agent influence fees (paying to get prioritized in agent recommendations), API access premiums (charging for agent-friendly product data feeds), and sponsored agent placement (ensuring visibility in agent search results). Smart retailers are already experimenting with "agent-first" product information—structured data that helps agents make better recommendations while highlighting key selling points.

For Platforms, new revenue streams emerge from agent orchestration fees (charging for facilitating multi-platform transactions), cross-platform transaction fees, selling agent performance analytics, and providing trust and verification services for agent-mediated purchases. The companies that build the infrastructure connecting agents to commerce will likely capture significant value.

For Agent Providers, monetization extends beyond simple transaction commissions to include premium agent personalities (pay extra for an agent trained on specific influencer preferences), specialized industry knowledge (agents optimized for restaurant procurement or medical supplies), and white-label agent solutions for retailers wanting to offer branded shopping assistants.

Perhaps most significantly, we're seeing the emergence of agent influence fees—retailers paying to ensure their products are recommended by popular agents, similar to how Google Ads works today but applied to conversational commerce.

How Agentic Commerce is Actually Built

The current landscape reveals three distinct approaches to agentic commerce, each with different technical architectures and business models:

1. Direct Integration Protocols (Anthropic's MCP): Conversational agents access merchant APIs directly
2. Agent-to-Agent Frameworks (Google's A2A protocol): Systems where agents communicate with other agents to complete transactions
3. Hybrid Web-Browsing Approaches: Agents that can navigate traditional e-commerce websites as well as use specialized APIs

Each approach creates different monetization opportunities for retailers. MCP requires merchants to build agent-friendly APIs. Agent-to-agent frameworks create new intermediary roles. Web-browsing agents can work with existing infrastructure but may miss agent-specific optimization opportunities.

How to Win in Agentic Commerce

We're not just witnessing the automation of shopping—we're seeing the emergence of a new influence economy where the power to shape purchasing decisions is shifting from traditional marketing channels to AI systems.

The companies that understand this shift earliest—and design their strategies around agent-mediated commerce—will capture disproportionate value in the coming decade. Those that don't risk being disintermediated by more agent-friendly competitors.

The question isn't whether agents will reshape retail—it's whether your business will shape how agents recommend products to consumers, or whether you'll be subject to the recommendations of agents controlled by your competitors.

In our next section, we'll explore the trust and control frameworks that will determine which products consumers are willing to delegate to agents, and which require human oversight.

Ready to Navigate the Agent Revolution?

The transformation to agentic commerce is accelerating faster than most businesses realize. The companies that act now—understanding both the market dynamics and technical implementation—will have significant first-mover advantages.

For Business Leaders: Book a discovery call with our team to assess how agentic commerce will impact your industry and develop a strategic roadmap for the next 18 months.

For Teams & Organizations: Join our upcoming "Agentic Commerce Masterclass" where we'll dive deep into implementation strategies, competitive positioning, and ROI frameworks for AI shopping agents.

For Enterprise Teams: Schedule a custom workshop for your leadership team. We'll analyze your specific market position, competitive threats, and opportunities in the agent economy—from both market design and AI implementation perspectives.

References

1. Amazon "Buy for Me" AI Shopping Agent Testing. TechCrunch, April 2025.
   
   
2. Shopify Agentic Commerce Platform Launch. Shopify Developer Documentation, 2024.
   
   
3. Shopify Storefront MCP Server Integration. Shopify Developer Documentation, 2024.
   
   
4. "How MCP-UI Powers Shopify's New Commerce Widgets in Agents." The New Stack, August 2025.
   
   
5. "The race is on to make AI agents do your online shopping for you." TechCrunch, December 2024.
   
   
6. "Salesforce reveals AI agent retail trends for 2025." Salesforce, March 2025.
   
   
7. "Connected Shoppers Report, 6th Edition." Salesforce, 2024.
   
   
8. Anthropic Model Context Protocol (MCP) Documentation. Anthropic, 2024.
   
   
9. "Announcing the Agent2Agent Protocol (A2A)." Google Developers Blog, April 2025.

‍