Customer Satisfaction & Retention Agent - Data App

An agent that assists the Marketing and Sales team to enhance customer satisfaction and retention by proactively addressing dissatisfied customers who might churn. This includes:

• Identifying the sentiment of each incoming customer review.

• Detecting the language of the review to tailor the communication in the customer's native language.

• Predicting what customers are likely to churn and selecting a personalized gift.

• Composing a customized apology email and exposing it to the the human for review and sending.

  • Optionally, the agent can also send the email autonomously without human supervision, if instructed to do so in the User message.

TECHNICAL REQUIREMENTS FOR TOOLS:

Before using the agent, make sure to:

• Load your own OpenAI API key in your Secret store on KNIME Hub and pass it to the tools that use an LLM to analyze or draft text: Sentiment_Analyzer_Tool and Retention_Email_Writer_Tool

• [Optional] Provide email credentials to the Email_Sender_Tool, if you want the agent to actually send the emails.

Disclaimer: AI agents are powered by LLMs, which generate responses based on patterns in data rather than fixed rules. As a result, their behavior is non-deterministic, meaning outputs may vary even for similar inputs. This may lead to situations where the agent fails to trigger expected tool calls or behaves in ways that differ from prior interactions. Please account for this variability when designing or using AI agents.

Provide agent with data

Possible questions

1. What customers have left a negative review?

2. Who are the first three non-English (no "en") speaking customers in the review data?

3. Which customers are predicted as churners ("1") and have a predicted churn probability above 0.8?

4. What custom gifts did customers with a high probability of churning (> 0.8) receive?

5. TASK:

   • Identify customers predicted as churners ("1") and have a predicted churn probability above 0.8

   • Write a retention email in the customer's language

   • Show me the email preview and ask for my confirmation before sending the email to the customers

Chat with agent