Granted Patent B2

US 12,743,639 · App. 18/208,912 · Granted Sep 22, 2026

Inventors: Sandeep Verma (Gurugram, IN); Nandini Rathaur (Mumbai, IN); Amit Mishra (Chennai, IN); Srinivas Dundigalla (Waxhaw, NC); Pavan Chayanam (Alamo, CA)

Assignee: Bank of America Corporation

G06N5/022G06F16/24575G06F16/25H04M3/5166

Abstract

Claims (96)

1 . A method for an artificial intelligence (“AI”) agent answering a call within a voice AI container, the method comprising:

receiving, at the voice AI container in a first receiving step, the call from a human caller, the call comprising a query;

parsing, by the AI agent, the query to identify context associated with the query; identifying, by the AI agent in a first identifying step using the context, an intent of the query; and

determining, by the AI agent in a first determining step, whether a first valid mapping stored within a backend database is associated with the intent of the query, the backend database being located within the voice AI container;

wherein the context comprises a content of the query and at least one of: a time the call was placed, a location from which the call was placed, and a history of previous calls placed; and

wherein the first valid mapping comprises a set of steps taken to satisfy the intent of the query;

wherein:

there is a plurality of valid mappings stored in the backend database;

each of the plurality of valid mappings is associated with a respective one of a plurality of intents of the query;

each of the plurality of valid mappings are created by the following steps:

identifying, by a machine learning agent located within the voice AI container in a third identifying step, the respective one of the plurality of intents of the query;

navigating, by the machine learning agent in a second navigating step, an agent screen, located within the voice AI container, to gather information to satisfy the respective one of the plurality of intents of the query;

capturing agent screen navigation data during the second navigating step;

associating the agent screen navigation data with the respective one of the plurality of intents of the query;

creating, in a second creating step, a respective automated screen navigation script using the agent screen navigation data and an associated intent of the query;

creating, in a third creating step, a respective one of the plurality of valid mappings using the respective automated screen navigation script and the associated intent of the query; and

storing the respective one of the plurality of valid mappings in the backend database.

2 . The method of claim 1 , further comprising the steps of:

receiving, by the AI agent in a second receiving step, the first valid mapping from the backend database; and

loading, by the AI agent in a first loading step, the first valid mapping to an agent screen located within the voice AI container;

wherein:

the first valid mapping comprises a first automated screen navigation script; and

the first loading step comprises loading the first automated screen navigation script on the agent screen and navigating, in a first navigating step, the agent screen using the first automated screen navigation script to gather information required to satisfy the intent of the query.

3 . The method of claim 1 , further comprising the steps of:

identifying, by the AI agent in a second identifying step, a human agent that is actively assigned to the voice AI container;

loading, in a second loading step, an agent voice and speech profile associated with the human agent from an agent speech profile database located in the voice AI container; and

relaying, by the AI agent, using the agent voice and speech profile associated with the human agent, the information to the human caller.

4 . The method of claim 3 , wherein the agent voice and speech profile is created by the following steps:

recording voice samples from the human agent;

sending the voice samples to a voice cloning software;

creating, in a first creating step, using the voice cloning software, the agent voice and speech profile;

associating the agent voice and speech profile with the human agent; and

storing the agent voice and speech profile in the agent speech profile database.

5 . The method of claim 1 wherein the context further comprises grammar and syntax of the query, a language which was used by the human caller, a pace that the human caller was speaking, an amount of words the human caller used for the query and a tone of the human caller.

6 . The method of claim 2 , further comprising the steps of:

updating the backend database to associate the query with the context and the first valid mapping.

7 . A method for an artificial intelligence (“AI”) agent answering a call within a voice AI container, the method comprising:

receiving, at the voice AI container in a first receiving step, the call from a human caller, the call comprising a query;

parsing, by the AI agent, the query to identify context associated with the query; identifying, by the AI agent in a first identifying step using the context, an intent of the query; and

determining, by the AI agent in a first determining step, absence of a first valid mapping associated with the intent of the query, the absence being limited to any location within a backend database, the backend database being located within the voice AI container;

wherein the context comprises a content of the query and at least one of: a time the call was placed, a location from which the call was placed, and a history of previous calls placed;

wherein the first valid mapping comprises a set of steps taken to satisfy the intent of the query;

identifying, by the AI agent in a second identifying step, a machine learning agent that is actively assigned to the voice AI container;

passing, by the AI agent, control of the call to the machine learning agent; and

the machine learning agent responding, in a first responding step, to the human caller.

8 . The method of claim 7 , wherein:

the first responding step comprises the machine learning agent, using the intent of the query identified in the first identifying step, determining, in a second determining step, whether a second valid mapping stored in the backend database is associated with the intent of the query; and

upon the determination, in the second determining step, that the second valid mapping is associated with the intent of the query the machine learning agent passes the call back to the AI agent with the second valid mapping.

9 . The method of claim 8 , wherein after the call is passed back to the AI agent, the first responding step further comprises the steps of:

updating the backend database to associate the query, context and intent of the query with the second valid mapping;

receiving, by the AI agent in a second receiving step, the second valid mapping from the backend database;

loading, by the AI agent in a first loading step, the second valid mapping to an agent screen located within the voice AI container wherein the second valid mapping comprises a first automated screen navigation script and wherein the first loading step comprises loading the first automated screen navigation script on the agent screen and navigating, in a first navigating step, the agent screen using the first automated screen navigation script to gather information required to satisfy the intent of the query;

identifying, by the AI agent in a third identifying step, a human agent that is actively assigned to the voice AI container;

loading, in a second loading step, an agent voice and speech profile associated with a human agent assigned to the voice AI container from an agent speech profile database located within the voice AI container; and

relaying, by the AI agent using the agent voice and speech profile in a first relaying step, the information to the human caller.

10 . The method of claim 7 , wherein the context further comprises grammar and syntax of the query, a language used by the human caller, a pace that the human caller was speaking, an amount of words the human caller used and a tone of the human caller.

11 . The method of claim 7 , wherein the first responding step comprises the steps of:

navigating, by the machine learning agent in a second navigating step, an agent screen located within the voice AI container to gather information required to satisfy the intent of the query identified in the first identifying step; and

relaying, by the machine learning agent in a second relaying step, the information to the human caller.

12 . The method of claim 11 wherein:

agent screen navigation data is captured during the second navigating step;

the agent screen navigation data is associated with the intent of the query;

the agent screen navigation data and the associated intent of the query are used to create a second automated screen navigation script;

a third valid mapping, using the second automated screen navigation script and the associated intent of the query, is created and stored in the backend database; and

the backend database is updated to associate the query, the context and the intent of the query with the third valid mapping.

13 . The method of claim 10 , wherein an agent voice and speech profile is created by the following steps:

recording voice samples from a human agent;

sending, in a second sending step, the voice samples to a voice cloning software;

creating, in a first creating step, using the voice cloning software, the agent voice and speech profile;

associating the agent voice and speech profile with the human agent; and

storing the agent voice and speech profile in an agent speech profile database.

14 . The method of claim 7 , wherein:

there is a plurality of valid mappings stored in the backend database; and

each of the plurality of valid mappings is associated with a respective one of a plurality of intents of the query.

15 . The method of claim 14 , wherein each of the plurality of valid mappings are created by the following steps:

identifying, in a fourth identifying step by a machine learning agent located within the voice AI container, the respective one of the plurality of intents of the query;

navigating, in a fourth navigating step by the machine learning agent, an agent screen, located within the voice AI container, to gather information to satisfy the respective one of the plurality of intents of the query;

capturing agent screen navigation data during a second navigating step;

associating the agent screen navigation data with the respective one of the plurality of intents of the query;

creating, in a second creating step, a respective automated screen navigation script using the agent screen navigation data and an associated intent of the query;

creating, in a third creating step, a respective one of the plurality of valid mappings using the respective automated screen navigation script and the associated intent of the query; and

storing the respective one of the plurality of valid mappings in the backend database.

16 . A method for an artificial intelligence (“AI”) agent answering a call within a voice AI container, the method comprising:

receiving, at the voice AI container in a first receiving step, the call from a human caller, the call comprising a query;

parsing, by the AI agent, the query to identify context associated with the query; identifying, by the AI agent in a first identifying step using the context, an intent of the query;

determining, by the AI agent in a first determining step, whether a first valid mapping stored within a backend database is associated with the intent of the query, the backend database being located within the voice AI container;

receiving, by the AI agent in a second receiving step, the first valid mapping from the backend database;

loading, by the AI agent in a first loading step, the first valid mapping to an agent screen located within the voice AI container;

wherein:

the first valid mapping comprises a first automated screen navigation script; and

the first loading step comprises loading the first automated screen navigation script on the agent screen and navigating, in a first navigating step, the agent screen using the first automated screen navigation script to gather information required to respond to the intent of the query;

identifying, by the AI agent in a second identifying step, a human agent that is actively assigned to the voice AI container;

loading, in a second loading step, an agent voice and speech profile associated with the human agent from an agent speech profile database located in the voice AI container; and

relaying, by the AI agent, using the agent voice and speech profile associated with the human agent, the information to the human caller;

wherein the context comprises a content of the query and at least one of: a time the call was placed, a location from which the call was placed, a history of previous calls placed, grammar and syntax of the query, a language which was used by the human caller, a pace that the human caller was speaking, an amount of words the human caller used for the query and a tone of the human caller; and

wherein the first valid mapping comprises a set of steps taken to respond to the intent of the query.

Assignments (1)

Continuity (1)

References Cited (12)