# US 12,763,195 B2 · Conversational artificial intelligence system with multi-agent orchestration and parallel supervisory processing > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,763,195 B2 - Document ID: US12763195B2 - Status: Granted - Kind: B2 - Application number: 19/547,381 - Filed: 2026-02-23 - Published: 2026-07-02 - Granted: 2026-10-13 - Earliest claimed: 2023-05-15 - Assignees: HealthGPT, Inc. - Applicants: HealthGPT, Inc. - Inventors: Munjal Shah; Vishal Parikh; Meenesh Bhimani; Subhabrata Mukherjee; Alex Miller; Saad Godil; Debajyoti Datta; Paul Gamble; Rae Lasko - Web page: https://patentgrep.com/patent/US12763195B2 ## CPC - G16H 40/20: for the management or administration of healthcare resources or facilities, e.g. managing hospital staff or surgery rooms - G10L 15/183: using context dependencies, e.g. language models - G10L 15/22: Procedures used during a speech recognition process, e.g. man-machine dialogue - G10L 25/66: for extracting parameters related to health condition (detecting or measuring for diagnostic purposes A61B5/00) - G10L 15/16: using artificial neural networks - G10L 15/1822: {Parsing for meaning understanding} - G10L 15/19: Grammatical context, e.g. disambiguation of the recognition hypotheses based on word sequence rules - G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA - G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING ## Abstract A conversational artificial intelligence system includes a plurality of large language models configured to perform different functions. A routing system routes requests to a large language model best suited for a task associated with the request. A conversation manager coordinates a flow of a conversation between the system and a user using the plurality of large language models. A supervisory module operates in parallel with the conversation to audit responses generated during the conversation and provides feedback that contributes to construction of downstream responses to the user. One or more safety modules prevent the system from performing actions that are outside of defined constraints. The routing system may route requests based on latency requirements, domain, or classification tasks, and may fail over to a different large language model when a first large language model degrades. ## Claims ### Claim 1 A conversational artificial intelligence system, comprising: a plurality of large language models, wherein individual large language models in the plurality are configured to perform different functions; a routing system configured to route a request to a large language model, selected from the plurality of large language models, that is best suited for a task associated with the request; a conversation manager configured to coordinate a flow of a conversation between the system and a user using the plurality of large language models; a supervisory module configured to operate in parallel with the conversation and audit responses generated during the conversation, the supervisory module providing feedback that contributes to construction of downstream responses to the user; and one or more safety modules configured to prevent the system from performing actions that are outside of defined constraints. ### Claim 2 (depends on 1) The conversational artificial intelligence system of claim 1, wherein the routing system is configured to route the request based on at least one of a latency requirement, a domain of the request, or a classification task. ### Claim 3 (depends on 1) The conversational artificial intelligence system of claim 1, wherein the conversation manager is configured to orchestrate the plurality of large language models to process speech, interpret meaning, and generate responses in parallel. ### Claim 4 (depends on 1) The conversational artificial intelligence system of claim 1, wherein the different functions include at least two of: intake, scheduling, order processing, pre-operative preparation, discharge, or navigation. ### Claim 5 (depends on 1) The conversational artificial intelligence system of claim 1, wherein the supervisory module is implemented using a large language model that is different from a large language model used by the conversation manager to generate the responses. ### Claim 6 (depends on 1) The conversational artificial intelligence system of claim 1, wherein the system comprises a plurality of supervisory modules, individual supervisory modules configured to perform different supervisory functions. ### Claim 7 (depends on 1) The conversational artificial intelligence system of claim 1, wherein the supervisory module is configured to determine whether a response is to be clarified and, in response to a determination that the response is to be clarified, provide feedback to the conversation manager causing the conversation manager to provide a clarification to the user during the conversation. ### Claim 8 (depends on 1) The conversational artificial intelligence system of claim 1, wherein the one or more safety modules are configured to detect and intercept threat vectors. ### Claim 9 (depends on 1) The conversational artificial intelligence system of claim 1, further comprising redundancy across model providers, wherein the routing system is configured to fail over to a different large language model when a first large language model degrades. ### Claim 10 (depends on 1) The conversational artificial intelligence system of claim 1, wherein the conversation is an audio-based conversation, and the system further comprises an automatic speech recognition module and a text-to-speech module. ### Claim 11 (depends on 1) The conversational artificial intelligence system of claim 1, wherein the supervisory module is further configured to transfer the conversation to a human when the supervisory module determines that a clarification is insufficient. ### Claim 12 (depends on 1) The conversational artificial intelligence system of claim 1, wherein the routing system routes the request to a large language model that is fine-tuned for a specific domain. ### Claim 13 (depends on 1) The conversational artificial intelligence system of claim 1, wherein the conversation manager coordinates the flow of the conversation across a plurality of channels including at least two of: voice, chat, email, or messaging. ### Claim 14 A computer-implemented method for operating a conversational artificial intelligence system that includes a plurality of large language models, the method comprising: configuring individual large language models in the plurality to perform different functions; routing, by a routing system, a request to a large language model, selected from the plurality of large language models, that is best suited for a task associated with the request; coordinating, by a conversation manager, a flow of a conversation between the system and a user using the plurality of large language models; auditing, by a supervisory module operating in parallel with the conversation, responses generated during the conversation, and providing feedback that contributes to construction of downstream responses to the user; and preventing, by one or more safety modules, the system from performing actions that are outside of defined constraints. ### Claim 15 (depends on 14) The method of claim 14, wherein the routing comprises routing the request based on at least one of a latency requirement, a domain of the request, or a classification task. ### Claim 16 (depends on 14) The method of claim 14, further comprising orchestrating the plurality of large language models to process speech, interpret meaning, and generate responses in parallel. ### Claim 17 (depends on 14) The method of claim 14, wherein the auditing comprises determining whether a response is to be clarified and, in response to a determination that the response is to be clarified, providing feedback causing the conversation manager to provide a clarification to the user during the conversation. ### Claim 18 (depends on 14) The method of claim 14, further comprising failing over, by the routing system, to a different large language model when a first large language model degrades. ### Claim 19 A non-transitory computer-readable medium storing instructions that, when executed by one or more processors, cause the one or more processors to: route, by a routing system, a request to a large language model, selected from a plurality of large language models configured to perform different functions, that is best suited for a task associated with the request; coordinate, by a conversation manager, a flow of a conversation between a conversational artificial intelligence system and a user using the plurality of large language models; audit, by a supervisory module operating in parallel with the conversation, responses generated during the conversation, and provide feedback that contributes to construction of downstream responses to the user; and prevent, by one or more safety modules, the system from performing actions that are outside of defined constraints. ### Claim 20 (depends on 19) The non-transitory computer-readable medium of claim 19, wherein the plurality of large language models includes at least one commercial large language model and at least one proprietary large language model.