Next-Best-Action Rule

Next-best-action logic lives entirely in the heads of senior advisors who intuitively know which product to recommend based on years of client interaction — nothing is written down, and recommendations vary wildly between advisors.

None — AI has no decision rules to execute; recommendation engines cannot be built when eligibility criteria, propensity thresholds, and priority rankings exist only as tribal knowledge.

A few next-best-action rules have been written down in scattered documents — one advisor created a cheat sheet for mortgage cross-sell eligibility, another keeps a personal spreadsheet of product priority rankings, but there is no consistent format or central location.

AI can surface these ad-hoc documents via search, but cannot execute recommendations because rule definitions are inconsistent, incomplete, and frequently contradictory across authors.

Next-best-action rules are documented in a standard template covering eligibility criteria, propensity thresholds, channel constraints, and priority rankings — the product marketing team maintains a rule library in SharePoint with consistent formatting across all active offers.

AI can parse and display rules for advisor reference, but cannot dynamically evaluate them against client data because rules reference external systems (CRM segments, risk scores) without explicit linkage definitions.

Next-best-action rules are formally connected to client data sources — each rule's eligibility criteria reference specific CRM fields, propensity model outputs, and regulatory restriction flags, enabling the recommendation engine to evaluate rules against any client profile programmatically.

AI can evaluate next-best-action rules against client profiles and generate ranked recommendations, but cannot self-optimize rule parameters because rules are stored as static documents rather than machine-executable logic.

Next-best-action rules are encoded in a machine-readable decision engine using DMN tables — eligibility criteria, propensity thresholds, channel constraints, regulatory restrictions, and priority rankings are all parameterized, version-controlled, and directly executable by AI recommendation systems.

AI can autonomously evaluate, rank, and recommend actions for each client context in real time; it can also propose rule modifications based on outcome data, though human approval is still required before changes go live.

Next-best-action rules are dynamic and self-tuning — the decision engine continuously adjusts propensity thresholds, priority rankings, and eligibility boundaries based on real-time outcome data, market conditions, and regulatory changes, with full audit trails and governance controls on every automated adjustment.

AI fully manages the next-best-action rule lifecycle: creating new rules from observed patterns, retiring underperforming rules, adjusting parameters in real time, and explaining every recommendation decision to advisors and compliance teams.

Ceiling of the CMC framework for this dimension.