Express
Auto-Pay at Deviation
Customer misses their normal monthly payment pattern. Platform detects deviation, checks preference rules, and autonomously executes payment from the designated account.
An ML clustering system that learns each customer's behavioral pattern, detects deviations in real time, and autonomously triggers preference-matched account management actions — falling back to SMS when internet isn't available.
Traditional account management systems fire alerts when something has already gone wrong: a missed payment, a declined transaction, a dropped balance. They don't learn individual behavioral patterns, can't predict what's coming, and have no way to act on a customer's own preferences — they escalate to a human or do nothing.
A computing platform trains an ML clustering model on historical user activity data — identifying each user's individual behavioral pattern. When the system detects a deviation, it surfaces the anticipated transaction, retrieves the user's stored preference rules, and autonomously triggers the matching action.
An intelligent virtual assistant collects those preference rules upfront via NLP conversation and hierarchical tree traversal — so when a deviation occurs, the system already knows what the user wants to happen.
The system is built in two phases: a training phase that creates user-specific behavior models from historical data, and a monitoring phase that continuously evaluates live activity against those models.
When deviation is detected, the platform queries a repository of the user's pre-stored preference rules and determines the appropriate action — without requiring human review for routine cases.
The platform trains a first ML model using historical user activity across a population. It extracts attributes from that data and applies a clustering algorithm — classifying users by behavioral pattern rather than demographic segment or account type.
A second ML model powers the intelligent virtual assistant — trained to analyze user responses, traverse a hierarchical preference tree, and extract structured preference rules from natural language input.
The intelligent virtual assistant conducts an interactive session to elicit preference rules from the user. It asks questions in sequence, analyzes responses with NLP, traverses a hierarchical tree structure to surface follow-up questions, and stores the resulting rules in a user data repository.
When a deviation is later detected, the system retrieves those rules and checks whether they apply to the current situation — enabling autonomous action that reflects the user's own stated preferences.
User specifies a bridge loan amount that can be triggered automatically if an anticipated transaction would cause a shortfall.
User designates an alternate account or funding source to draw from if the primary account lacks sufficient funds.
Specifies which outstanding balances can be paid automatically and under what conditions — e.g., pay minimum vs. full balance.
Designates a trusted party who can authorize account actions on the user's behalf if the user is unreachable.
Specifies preferred outreach channels — internet (first mode) or SMS (second mode) — and whether fallback is permitted.
When the ML model detects a behavioral deviation, the platform identifies the anticipated transaction, retrieves the user's preference rules, and triggers the appropriate action. Select a deviation type to see the full response chain.
Before triggering any action, the platform checks whether the user can connect to the internet. This is a structural check baked into the core claim — the preference rule application determines which communication mode to use, not just whether to communicate.
This dual-mode architecture ensures that users without reliable internet access — rural customers, customers during outages, international travelers — still receive the autonomous management actions they configured.
App notification, in-app messaging, push alert, portal communication — rich-channel delivery when connectivity is confirmed.
Plain text SMS fallback — triggered automatically when internet connectivity check fails. Preference rules specify whether SMS fallback is permitted.
The combination of behavioral ML, structured preference collection, and autonomous action execution enables proactive account management across a wide range of financial events — without requiring customer-initiated contact.
Six organizations — including Mastercard, Visa, and Truist Bank — have cited this patent in work spanning autonomous transaction management, AI-driven user decisioning, and preference-based financial automation. The breadth of assignees signals that this framework is being used as a reference across multiple sectors of financial AI.