Pega Next-Best-Action: Moving from Rules to Real-Time AI Decisioning

Pega's Next-Best-Action Designer has always been powerful. The shift from rule-based eligibility filters to AI-driven propensity models is where that power compounds.

The Problem with Rules at Scale

A typical enterprise NBA configuration accumulates hundreds of eligibility rules over time. Nobody fully understands the interaction effects. New rules get added defensively. The result: a system that's conservative by default and increasingly hard to change without regression testing every downstream action.

"

One insurance client had 847 active eligibility rules across 12 action groups. After migrating to adaptive models, they replaced 640 of them with three propensity models — and improved conversion 28%.

Adaptive Models vs. Predictive Models

Pega's adaptive models learn continuously from customer responses — no batch retraining, no MLOps pipeline, no separate model registry. For organisations without a mature data science function, this is transformative. For those with one, adaptive models handle the real-time layer while predictive models handle the longer-horizon propensity scoring.

Getting Started

Start with one action group, one channel. Let adaptive models run alongside your existing rules for 30 days and compare outcomes. The data from that comparison builds the business case for broader rollout — and usually speaks for itself.