Express
First-Time User Onboarding
System detects a new user (via credentials or facial recognition) and proactively surfaces step-by-step guidance before they encounter the first decision point.
A computing platform that detects when ATM users are struggling — via idle time, eye tracking, object detection, and facial recognition — and proactively delivers targeted assistance before the user ever asks for it.
Self-service ATM infrastructure was built to reduce cost — not to monitor user experience in real time. A new customer staring at the menu, a foreign visitor unsure of the interface, an elderly user holding a check and not knowing where to insert it: all face the same outcome. The machine waits. No help arrives.
A computing platform receives real-time presence data from ATM sensors — cameras, microphones, device detectors — and identifies when a specific user has a need for assistance. It then generates targeted commands that cause the ATM to execute the exact assistance action matched to that user's situation.
The system doesn't wait for a "help" button press. It proactively classifies the user's behavioral state from sensor data and acts — adapting the interface in real time to remove the friction that caused confusion.
The ATM feeds user presence data — not just session activity — to a central computing platform. The platform identifies need, generates a targeted assistance command, and routes it back to the ATM for immediate execution.
The loop is closed locally: no human in the middle. Assistance fires automatically based on sensor state, classified by the platform in real time.
The platform classifies user need by reading multiple sensor streams simultaneously. Each sensor type reveals a different dimension of user difficulty — from temporal (idle time) to behavioral (eye tracking, object holding) to contextual (foreign institution linkage).
The platform doesn't return a generic help screen — it matches the assistance action to the specific signal that triggered it. Select a sensor type below to see what assistance the platform generates.
The platform generates commands that direct the ATM to execute one or more of six distinct assistance modes — each designed for a different failure mode of the self-service experience.
Dynamically highlights navigation options in color — directing attention to the correct next step for the user's current context.
Plays a contextual tutorial based on the user's actual interaction sequence with the ATM — not a generic walkthrough.
Initiates a live video assistance call with a representative, displayed on the ATM's own screen.
Routes the video call to the user's linked mobile device instead of the ATM screen — supporting privacy or accessibility needs.
Displays step-by-step instructions for completing the specific function the user appears to be attempting.
Informs the user about recently updated functions or replacement functions available on the ATM that may resolve their issue.
Detecting struggle isn't enough — the platform also identifies who is struggling. A first-time user needs onboarding. A foreign institution customer needs interface localization. A user holding a check needs deposit guidance. User identity modulates the assistance type.
The platform captures user presence information via facial recognition cameras, drive-up cameras (license plate data), device detection, and ATM credentials — creating a multi-signal user context before generating any command.
The combination of multi-sensor detection, user context profiling, and targeted assistance commands enables adaptive ATM experiences across a wide range of use cases — including accessibility, internationalization, and real-time agent escalation.
Granted Aug 2023, this patent has been cited by Wells Fargo in a 2025 publication on customer-specific ATM content delivery — validating that the proactive assistance framework is directly relevant to how other institutions are building next-generation ATM infrastructure.