Customer AI

Customer AI in Adobe Experience Platform generates propensity scores that predict individual customer actions, enabling marketers to build highly targeted segments based on predicted behavior rather than historical patterns alone.

How Customer AI Works

Customer AI analyzes historical experience events (page views, purchases, email clicks) stored in the Real-Time Customer Profile to build predictive models. These models generate per-customer propensity scores that update continuously.

Propensity Score Types

Building a Customer AI Model

1. Define the Goal: Select what you want to predict (purchase, churn, specific event) and the prediction time window (7, 14, 30 days).
2. Select Population: Choose which customers to score - all profiles, specific segments, or profiles matching certain criteria.
3. Choose Events: Select which experience events to use as training signals (page views, add-to-cart, past purchases).
4. Configure Schedule: Set how often the model re-trains and re-scores profiles (weekly or monthly recommended).
5. Review & Launch: Validate model configuration, review data availability, and launch training.
6. Monitor Performance: Review model accuracy metrics (AUC, lift charts) and adjust configuration if needed.

Activating Customer AI Scores

Influential Factors

Customer AI provides transparency into which factors drive each propensity score:

• Top Factors: View the most influential variables (recency of purchase, email engagement, browse frequency) for each score.
• Factor Direction: Understand whether each factor increases or decreases the propensity (e.g., "days since last purchase" increases churn risk).
• Segment-Level Insights: Analyze which factors are most important for specific customer segments to refine marketing strategy.
• Model Transparency: Use influential factors to validate that the model is learning meaningful patterns, not artifacts.