# US 12,120,267 B2 · Federated intelligent contact center concierge service > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,120,267 B2 - Document ID: US12120267B2 - Status: Granted - Kind: B2 - Application number: 17/748,025 - Filed: 2022-05-18 - Published: 2023-11-23 - Granted: 2024-10-15 - Earliest claimed: 2022-05-18 - Assignees: Avaya Management L.P. - Applicants: Avaya Management L.P. - Inventors: Kevin R. Plain; John Young - Web page: https://patentgrep.com/patent/US12120267B2 ## CPC - H04M 3/4936: {Speech interaction details (speech recognition per se G10L15/00)} - G10L 15/063: {Training} - G10L 15/16: using artificial neural networks - H04L 9/321: {involving a third party or a trusted authority} - H04L 9/50: {using hash chains, e.g. blockchains or hash trees} - H04M: TELEPHONIC COMMUNICATION (circuits for controlling other apparatus via a telephone cable and not involving telephone switching apparatus G08) - G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING - H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M) ## Abstract Customers often call a contact center to resolve an issue only to find the contact center cannot resolve the customer's issue and instead requires a third party to perform some action. By monitoring a customer's communication with an agent, an automated process may generate a workflow to address the issue comprising at least one action to be performed by a third party. The third party is then contacted, in real-time or offline, and provided with their actions to be performed. Records are written to a blockchain describing authentications, actions to take and by whom, and the completion of such actions. The resulting success, or lack thereof, is provided as feedback to further refine the accuracy of the automatic determinations of the workflow and/or the identity of the particular third parties. ## Claims ### Claim 1 A system, comprising: a server; a network interface to a network; and wherein at least one processor of the server performs: connecting, via the network, to a customer communication device for a communication comprising an audio portion comprising speech provided by a customer utilizing the customer communication device and wherein the communication is a real-time communication; analyzing the speech to identify a work item; upon determining the work item requires an action from a third party, analyzing the speech to identify the third party; constructing a workflow comprising a number of steps that, when each step has been completed, resolves the work item; providing at least one third-party step of the number of steps to the third party that, when complete, performs the action; and performing at least one host step of the number of steps. ### Claim 2 (depends on 1) The system of claim 1, wherein analyzing the speech to identify the work item comprises providing the speech to a neural network trained to identify work items and receiving the work item therefrom. ### Claim 3 (depends on 2) The system of claim 2, wherein the neural network is trained, comprising at least one processor executing a computer-implemented method of training a neural network for work item identification comprising: collecting a set of work items from a database; applying one or more transformations to each work item including substituting a word with a synonymous word, substituting a word with a synonymous phrase, substituting a purpose for a topic associated with the purpose, substituting the topic associated with the purpose of the communication with the purpose, inserting at least one redundant word, removing at least one redundant word, removing a first unique topic, and adding a second unique topic to create a modified set of work items; creating a first training set comprising the collected set of work items, the modified set of work items, and a set of customer provided content absent work items; training the neural network in a first stage of training using the first training set; creating a second training set for a second stage of training comprising the first training set and customer provided content absent work items that are incorrectly detected as comprising a work item after the first stage of training; and training the neural network in the second stage of training using the second training set. ### Claim 4 (depends on 1) The system of claim 1, wherein analyzing the speech to identify the third party comprises providing the speech to a neural network trained to identify third parties and receiving the third party therefrom. ### Claim 5 (depends on 4) The system of claim 4, wherein the neural network is trained, comprising at least one processor executing a computer-implemented method of training a neural network for work item identification comprising: collecting a set of third-party actions from a database; applying one or more transformations to a third-party action including substituting a third-party service for a third-party name, substituting a third-party name for a third-party service, substituting a third-party action for a third-party name, substituting a third-party name for a third-party action, substituting a work item topic for a third-party name, and substituting a third-party name for a work item subject to create a modified set of work items; creating a first training set comprising the collected set of third-party actions, the modified set of work items, and a set of communication content absent a third-party action; training the neural network in a first stage of training using the first training set; creating a second training set for a second stage of training comprising the first training set and host actions that are incorrectly determined as third-party actions after the first stage of training; and training the neural network in the second stage of training using the second training set. ### Claim 6 (depends on 1) The system of claim 1, wherein the communication includes an automated agent engaging in the communication with the customer. ### Claim 7 (depends on 1) The system of claim 1, wherein the communication includes an agent communication device utilized by an agent to engage in the communication with the customer. ### Claim 8 (depends on 1) The system of claim 1, wherein the action is encoded as encrypted blocks and added to a blockchain. ### Claim 9 (depends on 8) The system of claim 8, wherein at least one of the encrypted blocks comprises a self-executing smart contract in response to determining an corresponding action has been completed. ### Claim 10 (depends on 1) The system of claim 1, wherein upon determining the work item requires the action from the third party, joining a third-party communication device to the communication and conducting a portion of the communication therewith. ### Claim 11 A method, comprising: connecting, via a network, to a customer communication device for a communication comprising an audio portion comprising speech provided by a customer utilizing the customer communication device and wherein the communication is a real-time communication; analyzing the speech to identify a work item; upon determining the work item requires an action from a third party, analyzing the speech to identify the third party; constructing a workflow comprising a number of steps that, when each step has been completed, resolves the work item; and providing at least one third-party step of the number of steps to the third party that, when complete, performs the action. ### Claim 12 (depends on 11) The method of claim 11, wherein analyzing the speech to identify the work item comprises providing the speech to a neural network trained to identify work items and receiving the work item therefrom. ### Claim 13 (depends on 12) The method of claim 12, wherein the neural network is trained, comprising at least one processor executing a computer-implemented method of training a neural network for work item identification comprising: collecting a set of work items from a database; applying one or more transformations to each work item including substituting a word with a synonymous word, substituting a word with a synonymous phrase, substituting a purpose for a topic associated with the purpose, substituting the topic associated with the purpose of the communication with the purpose, inserting at least one redundant word, removing at least one redundant word, removing a first unique topic, and adding a second unique topic to create a modified set of work items; creating a first training set comprising the collected set of work items, the modified set of work items, and a set of customer provided content absent work items; training the neural network in a first stage of training using the first training set; creating a second training set for a second stage of training comprising the first training set and customer provided content absent work items that are incorrectly detected as comprising a work item after the first stage of training; and training the neural network in the second stage of training using the second training set. ### Claim 14 (depends on 11) The method of claim 11, wherein analyzing the speech to identify the third party comprises providing the speech to a neural network trained to identify third parties and receiving the third party therefrom. ### Claim 15 (depends on 14) The method of claim 14, wherein the neural network is trained, comprising at least one processor executing a computer-implemented method of training a neural network for work item identification comprising: collecting a set of third-party actions from a database; applying one or more transformations to a third-party action including substituting a third-party service for a third-party name, substituting a third-party name for a third-party service, substituting a third-party action for a third-party name, substituting a third-party name for a third-party action, substituting a work item topic for a third-party name, and substituting a third-party name for a work item subject to create a modified set of work items; creating a first training set comprising the collected set of third-party actions, the modified set of work items, and a set of communication content absent a third-party action; training the neural network in a first stage of training using the first training set; creating a second training set for a second stage of training comprising the first training set and host actions that are incorrectly determined as third-party actions after the first stage of training; and training the neural network in the second stage of training using the second training set. ### Claim 16 (depends on 11) The method of claim 11, wherein the communication includes an automated agent engaging in the communication with the customer. ### Claim 17 (depends on 11) The method of claim 11, wherein the communication includes an agent communication device utilized by an agent to engage in the communication with the customer. ### Claim 18 (depends on 11) The method of claim 11, wherein the action is encoded as encrypted blocks and added to a blockchain. ### Claim 19 (depends on 11) The method of claim 11, wherein upon determining the work item requires the action from the third party, joining a third-party communication device to the communication and conducting a portion of the communication therewith. ### Claim 20 A system, comprising: means to connect a customer communication device for a communication comprising an audio portion comprising speech provided by a customer utilizing the customer communication device and wherein the communication is a real-time communication; means to analyze the speech to identify a work item; means to, upon determining the work item requires an action from a third party, analyze the speech to identify the third party; means to construct a workflow comprising a number of steps that, when each step has been completed, resolves the work item; and means to provide at least one third-party step of the number of steps to the third party that, when complete, performs the action.