Kustomer AI Agent Testing Checklist

Kustomer unifies conversation history, CRM data, and knowledge base content. Start by grouping recent conversations into specific customer intents.

Attach the knowledge base article, policy, or approved macro that should constrain each intent before it is cleared for automation.

Kustomer timeline and CRM attributes can enrich answers, but they also increase risk. Plan, order history, sentiment, or lifecycle stage should be restricted until permission and source rules are clear.

Test when the agent should answer with context, ask a clarifying question, or escalate.

Kustomer AI Agents can use roles, tools, and workflows that act across systems. Each role needs its own readiness check: data sources, action boundaries, failure states, and fallbacks.

An answer may be safe while a cross-system action remains unsafe if the target system state, permission, or required field is unresolved.

Kustomer AI should not handle high-value exceptions, billing disputes, legal complaints, account control, or security requests without explicit human routing.

Separate approved routine intents from restricted, blocked, and human-only work in the launch map.

Write clear handoff triggers for unresolved, sensitive, or missing-source intents. Define observability signals that tell operators when the agent is near a boundary.

Retest after knowledge updates, workflow changes, CRM schema edits, or policy revisions that affect approved intents.