AI Assistant (Leat MCP)
July 20
→
Decision engine
The Decision Engine considers the customer, the current moment, the actions Governance permits, and the objective your business has set. It compares every available option, including doing nothing, then selects the action most likely to make a worthwhile difference.
Every decision begins with context from the Data layer, including the unified profile, activity history, latest event, loyalty status, and preferences the customer has shared. A purchase, lapse milestone, tier change, or e-commerce visit can start a decision. The engine receives the same complete customer context regardless of which channel produced the trigger.
Governance checks every possible action before the Decision Engine compares them. Eligibility, redemption rules, stacking restrictions, frequency caps, available budget, margin requirements, and consent determine which options remain. An action that breaks a rule, exceeds a limit, or does not apply to the customer never enters the final decision.
For every permitted action, the engine estimates the likely effect, expected cost, and confidence in the result. Each option is scored against the objective you have chosen, such as increasing visits, improving retention, protecting margin, or growing a category. Doing nothing is evaluated in the same way. If the customer is likely to return without an incentive, no action may be the strongest choice.
The selected action passes to Unified Execution for delivery through the appropriate channel. Leat records the customer context, permitted options, expected results, and chosen action, creating an explanation that can be reviewed later. The actual outcome then becomes available to Optimization, helping inform future decisions in similar situations.
Objective setting
Decisioning can focus on visits, retention, margin, category growth, or a weighted combination as business priorities change.
Available actions
Teams can define which rewards, messages, and customer actions the engine may choose for each segment or incentive type.
Decision trigger configuration
Purchases, lapse milestones, tier changes, and other customer events can start a decision for the relevant customer groups.
Confidence and fallback setting
Decisions below a chosen confidence level can follow a standard rule, use the safest option, or be sent to a person for approval.
Fixed journey testing
A fixed journey, such as a 90-day lapse message, can be compared with Decisioning using a customer group that remains on the existing rule.
Decision reviews
Teams can regularly review recent choices and their explanations to confirm that the engine is behaving as intended.
