Express
Real-Time Queue Rebalancing
Phone queue reaches 25-minute wait. Platform intercepts queued users, collects query attributes, and routes those with web-resolvable issues to chat — where agents have capacity.
An ML-driven platform that intercepts customers waiting in congested service channels, collects their query attributes upfront via an IVA, then routes them to an underutilized channel — where the agent already knows what they need. The customer never repeats themselves.
Customer service infrastructure routes users to whichever channel they contacted first — regardless of whether that channel is congested. A phone queue might have a 30-minute wait while email and chat have agents available. The system has no mechanism to rebalance.
A computing platform monitors wait times across all customer service channels simultaneously. When it identifies a congested channel, it intercepts the waiting customer via an IVA, collects the query attributes directly, and selects a less-congested channel to redirect them to.
Critically: the platform forwards those attributes to the enterprise agent at the receiving channel before the customer arrives. The agent is briefed. The customer never has to re-explain their issue. An ML model learns wait-time patterns and ingests external events to proactively pre-position resources before demand spikes materialize.
The platform operates as a traffic layer across all customer service channels — not inside any single channel. It monitors, intercepts via IVA, and routes. The ML model handles both real-time selection and proactive resource allocation from external event signals.
The IVA is the bridge: it contacts the user on the congested channel, collects attributes, and facilitates the move to the second channel — without requiring the user to initiate a new contact.
The ML model in this patent plays two distinct roles — and is explicitly trained for both. First: detecting patterns of estimated wait times and query attributes to make real-time channel selection decisions. Second: determining resource allocation for channels, adjusted dynamically by external events.
The external event integration is unique: the platform identifies events from one or more external data sources — weather, geopolitical, market news — that may impact wait times, feeds them to the ML model, and uses the output to pre-allocate service resources before demand spikes arrive.
The patent explicitly covers routing among telephone, web interface, video teleconference, email, and the IVA itself. Special routing logic applies when a licensed professional is needed, or when the user has 5G connectivity enabling video. Select a scenario below to see how the platform routes.
The most critical element of the patent's customer experience design is the attribute forwarding step. When a user is redirected to a second channel, the platform transmits their query attributes — collected during the IVA intercept — directly to the enterprise agent at the receiving channel.
The agent already knows the user's issue, account context, and query type before the conversation begins. No "can you describe your issue?" No starting over. The user's time is preserved on both ends of the redirection.
The platform isn't constrained to a specific interface type — it routes among any combination of channels in the enterprise's service infrastructure. The same ML model handles routing decisions for phone, web, video, email, and IVA sessions.
Two specialized routing rules are explicitly claimed: if the query requires a licensed professional, the platform selects the channel associated with one. If the user has 5G connectivity, the platform recommends a video channel for a higher-quality service experience.
Voice call — typically the highest-volume channel and most likely to experience congestion spikes.
Chat or portal-based interaction — often underutilized when phone queues are long.
Live video with agent. Platform recommends this when user's 5G connectivity is detected.
Asynchronous channel — suitable for non-urgent queries; rarely at capacity during real-time spikes.
The IVA serves both as the intercept mechanism and as a possible destination channel — the platform can route the query to be fully resolved via IVA when no human agent is needed.
The combination of ML-based wait-time prediction, external event ingestion, cross-channel routing, and attribute forwarding creates a service infrastructure that manages itself — redistributing load before customers experience it, not after.
This patent has been cited by an individual inventor building communication coordination systems and by Verint Americas — a major enterprise contact center software company — in a 2024 patent on asynchronous task-oriented virtual assistants. Verint's citation directly validates the query-attribute-forwarding architecture.