Power Conversational Experience
Learning Objectives
After completing this unit, you'll be able to:
- Explain how your recommenders work inside a conversational AI experience.
- Describe the three key actions an AI agent uses to deliver personalized suggestions.
- Identify how intent detection enhances recommender output based on customer utterances.
From Channels to Conversations
You’ve come a long way. So far, you’ve explored the building blocks of recommenders, created both rule-based and objective-based recommenders, and activated them across web, mobile, and email using a Personalization Point. Now, it’s time to go beyond traditional channels.
Until now, recommendations have appeared in predefined placements, such as a homepage carousel or an email. But customers increasingly expect more interactive experiences, where they can ask questions and receive relevant suggestions in real time.
At Cloud Kicks, Linda wants to take this a step further by bringing recommendations into a conversational experience, so customers can interact naturally and receive suggestions tailored to their preferences. For example, a shopper says, “I need some new basketball shoes.” Instead of responding with a generic suggestion, the system interprets the request, considers the shopper’s behavior, and returns personalized recommendations using the recommenders already built.
It’s time to explore how to use these recommenders to power intelligent, conversational product suggestions.
Agent Actions: How Your Recommender Gets Conversational
Agentforce Adaptive Websites combine real-time customer data with conversational AI to deliver personalized experiences. Instead of manually filtering, customers just ask the AI agent questions in plain language and receive personalized answers.
Agentforce uses three core Agent Actions to turn a conversation into personalized results.
Agent Actions | What It Does | Why It Matters |
|---|---|---|
Get Context | Retrieves customer data such as browsing history and preferences. | Helps the agent understand who the shopper is. |
Understand User Intent | Interprets the shopper’s request and extracts key details. | Turns a question into meaningful inputs for recommendations. |
Get Recommendations | Calls the Personalization Point to fetch products. | Returns personalized results using your existing recommenders. |
These three actions work together in a single conversation flow, orchestrated by a Prompt Template in Salesforce Prompt Builder.
Tying It All Together with Prompt Builder
A Prompt Template is the instruction set that tells the AI agent how to act, when to use each action, and how to format the response for the customer. Think of it as the agent's playbook.
At Cloud Kicks, admin Linda uses Prompt Builder to configure the agent for product recommendations. She gives it a clear set of instructions:
You are a helpful shopping assistant for Cloud Kicks. When a customer asks for product recommendations: Use the Get Context action to retrieve their browsing history and affinities. Use the Understand User Intent action to extract their preferences such as category, price range, style. Use the Get Recommendations action to fetch personalized sneaker suggestions from the Maximize Revenue recommender. Present the recommendations in a friendly, conversational tone with product names, prices, and direct links to add to cart.
The agent follows this flow when a shopper asks for recommendations. The agent first retrieves customer context, then learns the shopper’s request, and calls the recommender to return relevant products. It presents the results in a friendly, conversational format with product details and links.
The Prompt Template makes sure the agent stays on-brand, respects the customer's intent, and delivers recommendations in a way that feels natural, and not robotic.
Why This Matters
Agentforce Adaptive Websites bring a new level of personalization.
- Conversational discovery: Shoppers don’t have to browse, they can ask the agent what they’re looking for and get instant, tailored suggestions.
- Reuses existing logic: The same recommenders power both the web carousel and the conversational experience. You don’t build a separate recommendation engine for the agent.
- Intent-aware: The agent doesn’t just respond to keywords, it understands what the customer is trying to accomplish and adjusts recommendations accordingly.
For Cloud Kicks, this means after-hours shoppers who land on the site without a rep available can still get 1:1 personalized guidance—driving conversions even when the team is offline.
Wrap Up
In this badge, you explored how recommenders work in Salesforce Personalization, from building rule-based and objective-based recommenders to making them available across web, mobile, and email using a Personalization Point. You also learned how the same recommenders power conversational experiences with Agentforce Adaptive Websites, using real-time context and intent to deliver personalized suggestions.
One recommender. Multiple channels. Consistent personalization everywhere. Now you’re ready to build recommenders that meet customers wherever they engage, keeping every interaction relevant and connected.