ML Recommendations on the Agentic Commerce Platform

ML-powered product recommendations are now available on the Agentic Commerce platform, enabling businesses to deliver personalized shopping experiences through AI agents.

The Opportunity

Every customer interaction is an opportunity to understand what they want and deliver it. Traditional recommendation systems rely on historical data and simple rules. They can suggest similar products or popular items, but they struggle with context, nuance, and the complex ways customers actually shop.

The Agentic Commerce platform changes that. ML recommendations powered by AI agents understand not just what customers buy, but why they buy it, what they're trying to accomplish, and what would help them complete their goals.

How It Works

The recommendation engine analyzes products, customer behavior, and purchase patterns to identify relationships between items. It can find products that are similar to what a customer is viewing, products that complement items in their cart, or alternatives that might better fit their needs.

But it goes further. The system understands personas—customer types defined by their preferences, behaviors, and goals. It can filter recommendations by budget, apply natural language queries to add context, and use agentic workflows that combine multiple signals to deliver highly personalized suggestions.

Recommendation Types

Similar Products: When a customer views a product, the system can find items with similar features, styles, or use cases. This helps customers explore options and find exactly what they're looking for.

Complementary Products: For items in a customer's cart, the system identifies products that work well together—accessories that enhance the main purchase, items frequently bought together, or products that complete a set.

Alternative Products: Sometimes customers want options. The system can find products that serve the same purpose but offer different features, price points, or styles, giving customers the ability to compare and choose.

Personalization and Context

The recommendation engine understands that every customer is different. It can apply persona-based filtering, using predefined customer types or natural language descriptions of customer characteristics. It respects budget constraints, filtering recommendations to fit what customers are willing to spend.

The system also understands context through natural language queries. A customer might be shopping for "gift ideas" or "office supplies for a small team"—the recommendation engine uses these queries to refine suggestions and deliver more relevant results.

Agentic Workflows

The most powerful feature is the agentic workflow, which combines multiple recommendation signals with AI reasoning. Instead of relying on a single algorithm, the system uses AI agents to understand customer intent, analyze product relationships, and deliver recommendations that feel natural and helpful.

This enables complex scenarios: an agent helping a customer find the perfect gift by understanding the recipient's preferences, a business agent identifying the right software tools for a team's workflow, or a shopping assistant suggesting products that complete a project the customer is working on.

Use Case

Here's how a business might use the recommendation engine to help customers find the perfect gift:

$ hyperfold recommend \
  --product prod_laptop_123 \
  --agentic \
  --model gpt-5 \
  --query-prompt "gift ideas for a creative professional who works from home" \
  --persona-prompt "tech-savvy, values design and aesthetics, budget-conscious" \
  --budget 100-500 \
  --complementary \
  --filter-prompt "items that enhance productivity and workspace comfort"

This command asks the recommendation engine to find complementary products for a laptop, using an agentic workflow that understands the customer is shopping for a gift. It applies persona characteristics, respects a budget range, and filters results to focus on productivity and workspace items. The result is a curated set of recommendations that feel personalized and relevant.

Business Impact

Businesses using ML recommendations on the Agentic Commerce platform see significant improvements in customer engagement and sales. Recommendations help customers discover products they might not have found otherwise, increasing average order values and reducing cart abandonment.

The system works at scale, handling thousands of products and millions of customer interactions. It learns from each interaction, continuously improving its understanding of product relationships and customer preferences.

Integration and Workflow

The recommendation engine integrates seamlessly with existing commerce workflows. It works with product catalogs, shopping carts, and customer profiles. It can be called programmatically through APIs or used interactively through AI agents that help customers shop.

The system respects business rules and constraints. Merchants can control which products are recommended, set minimum confidence thresholds, and configure how recommendations are displayed. This ensures recommendations align with business goals while delivering value to customers.

---

ML recommendations on the Agentic Commerce platform enable businesses to deliver personalized shopping experiences at scale.

Get started to learn how recommendations can transform your customer experience.