Augmented intelligence assistant for agents

A system comprising:

• one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:
• collecting, from a database, (i) aggregated information of customers, (ii) intents of the customers that identify specific customer support services that the customers received from past interactions with customer support agents of a company, from among multiple support services that the customer support agents of the company are able to provide to the customers, (iii) sentiments of the customers that reflect an automatically determined mood or opinion that the customers holds about the company, (iv) previous actions performed by the customer support agents for the customers, and (v) outcomes of the previous actions performed by the customer support agents for the customers that reflect whether the customers were satisfied by the previous actions performed by the customer support agents for the customers;
• creating training sets that, for each of the customers, include (i) information about the customer, (ii) an intent of the customer, (iii) a sentiment of the customer, (iv) a previous action performed by an agent for the customer, and (v) an outcome of the previous action that was performed by the agent for the customers;
• training, using the training sets, a neural network to output a recommendation to be performed by a given agent for use with a given customer, based an input that includes (i) for input information about a given customer, (ii) an input intent of the given customer that identifies a specific customer support service that the given customer intends to receive from a customer support agent of the company, from among multiple support services that the customer support agent is able to provide to the given customer, and (iii) an input sentiment of the given customer that reflects an automatically discerned mood or opinion that the given customer holds about the company, wherein training the neural network comprises assigning a higher weight to recommended actions that are associated with positive outcomes in training sets, and assigning a lower weight to recommended actions that are associated with negative outcomes in the training sets;
• obtaining interaction information regarding an interaction between a customer and an agent;
• while the interaction is occurring, automatically transcribing the interaction information from one format into a different format;
• identifying the customer from the transcribed interaction information;
• aggregating information of the customer from multiple databases, including querying one or more databases that are external to the company and that are not controlled by the company;
• determining, based on the transcribed interaction information and the aggregated information of the customer, an intent of the customer that identifies a particular customer support service that the customer likely intends to receive from the customer support agent of the company, from among the multiple support services that the customer support agent of the company is able to provide to the customer;
• automatically determining, based on the transcribed interaction information and the aggregated information of the customer, a sentiment of the customer that reflects a mood or opinion that the customer holds about the company;
• providing, to the trained neural network, an input feature set that includes (i) the aggregated information of the customer, (ii) the intent of the customer, and (iii) the sentiment of the customer;
• receiving, from the trained neural network based on the input feature set, an output that identifies a particular recommended action to be performed by the agent for the customer;
• and
• while the interaction is occurring, generating an interface to provide to the agent based at least on the particular recommended action.

A non-transitory computer-readable medium storing software comprising instructions executable by one or more computers which, upon such execution, cause the one or more computers to perform operations comprising:

• collecting, from a database, (i) aggregated information of customers, (ii) intents of the customers that identify specific customer support services that the customers received from past interactions with customer support agents of a company, from among multiple support services that the customer support agents of the company are able to provide to the customers, (iii) sentiments of the customers that reflect an automatically determined mood or opinion that the customers holds about the company, (iv) previous actions performed by the customer support agents for the customers, and (v) outcomes of the previous actions performed by the customer support agents for the customers that reflect whether the customers were satisfied by the previous actions performed by the customer support agents for the customers;
• creating training sets that, for each of the customers, include (i) information about the customer, (ii) an intent of the customer, (iii) a sentiment of the customer, (iv) a previous action performed by an agent for the customer, and (v) an outcome of the previous action that was performed by the agent for the customers;
• training, using the training sets, a neural network to output a recommendation to be performed by a given agent for use with a given customer, based an input that includes (i) for input information about a given customer, (ii) an input intent of the given customer that identifies a specific customer support service that the given customer intends to receive from a customer support agent of the company, from among multiple support services that the customer support agent is able to provide to the given customer, and (iii) an input sentiment of the given customer that reflects an automatically discerned mood or opinion that the given customer holds about the company, wherein training the neural network comprises assigning a higher weight to recommended actions that are associated with positive outcomes in training sets, and assigning a lower weight to recommended actions that are associated with negative outcomes in the training sets;
• obtaining interaction information regarding an interaction between a customer and an agent;
• while the interaction is occurring, automatically transcribing the interaction information from one format into a different format;
• identifying the customer from the transcribed interaction information;
• aggregating information of the customer from multiple databases, including querying one or more databases that are external to the company and that are not controlled by the company;
• determining, based on the transcribed interaction information and the aggregated information of the customer, an intent of the customer that identifies a particular customer support service that the customer likely intends to receive from the customer support agent of the company, from among the multiple support services that the customer support agent of the company is able to provide to the customer;
• automatically determining, based on the transcribed interaction information and the aggregated information of the customer, a sentiment of the customer that reflects a mood or opinion that the customer holds about the company;
• providing, to the trained neural network, an input feature set that includes (i) the aggregated information of the customer, (ii) the intent of the customer, and (iii) the sentiment of the customer;
• receiving, from the trained neural network based on the input feature set, an output that identifies a particular recommended action to be performed by the agent for the customer;
• and
• while the interaction is occurring, generating an interface to provide to the agent based at least on the particular recommended action.