Agentic AI for Decision Explainability in CDH: Understand Why Offers Surface (or Don’t)

AI
1

This article walks you through designing and implementing a Decision Reasoning Agent layered on top of Pega Customer Decision Hub. A demo scenario is included that can be adapted to your own NBA use case.

Decision Reasoning Agent

Not just a trace viewer — an intelligent assistant that analyzes decision outcomes end-to-end. It can:

  • Interpret qualification logic

  • Analyze engagement and contact policies

  • Evaluate suppression conditions

  • Explain arbitration prioritization

  • Compare evaluated vs. surfaced actions

Instead of manually reviewing strategies and traces, the agent synthesizes decision context and produces business-readable explanations.

When a stakeholder asks:

“Why didn’t this offer show?”

The agent evaluates:

  • Customer attributes

  • Eligibility & applicability rules

  • Suppression conditions

  • Arbitration weights

  • Adaptive model influence

It then explains the outcome clearly and identifies the dominant factor. If no action surfaces, it indicates whether the cause was eligibility failure, suppression, arbitration loss, or configuration.

How It Works

Implemented using:

  • Conversational Agent rule — interprets diagnostic requests

  • Agent + Tool rules + Data Pages — fetch, update, and process data across decision layers

  • Structured prompts — guide reasoning and generate business-readable explanations

Runtime Flow (CPV via Data Pages)

  1. User inputs — CustomerID, Channel, Direction, Action

  2. Agent retrieves data — via Data Pages configured to expose customer profile, engagement context, and strategy metadata (similar to CPV)

  3. Evaluates all actions — qualified and disqualified, based on eligibility rules, suppression policies, and arbitration logic

  4. Generates layered explanation:

    • Qualification result

    • Suppression reasoning

    • Arbitration comparison

    • Final surfaced outcome

  5. Summarizes findings — in business-friendly language for stakeholders or operational dashboards

Outcome: Fast, automated root cause identification and full transparency across complex NBA ecosystems — without manually navigating CPV or tracing every strategy.

Demo Scenario

For a Web Inbound interaction where a specific action did not surface, the agent:

  • Evaluates all relevant actions

  • Confirms action qualification

  • Detects suppression by contact policy

  • Explains arbitration score comparison

  • Highlights the primary suppression driver

Analysis completes in seconds, eliminating manual trace navigation.

Why This Matters

Modern CDH implementations include:

  • Multi-layered engagement policies

  • Complex arbitration strategies

  • Adaptive model influence

  • Cross-channel suppression logic

Manual investigation is time-consuming and requires deep expertise. The Decision Reasoning Agent provides:

  • Faster triage

  • Business-readable explanations

  • Improved governance

  • Reduced dependency on senior architects