Artificial intelligence assistant for customer service representatives

1. A method comprising:

beginning a workflow involving communication between a customer service representative and a customer, wherein the workflow is provided to the customer service representative using an interface, and wherein the workflow comprises a plurality of steps in the communication;

receiving, by an artificial intelligence (AI) assistant, an input via a communication session between a customer device and a customer service representative device, wherein the communication session is associated with the workflow, wherein the AI assistant is trained using machine learning on a database containing previously-received inputs, previously-delivered outputs, and the previous results of previously-delivered outputs;

tracking, by the AI assistant, progress of the customer service representative through the workflow;

processing, by the AI assistant, the input;

determining, by the processing, that at least a portion of the input includes sensitive information;

determining that an authorization level of the customer service representative is below a level required to receive the sensitive information;

determining, by the AI assistant, an output based on the processing of the input and the progress, wherein the output supplements at least one step in the workflow, wherein the output includes one of a suggested statement, change in tone, or correction for the customer service representative; and

sending, by the AI assistant, an indication of the output to the customer service representative device, wherein the output is censored to redact any sensitive information.

2. The method of claim 1 , wherein the indication of the output comprises at least one of causing a window comprising a text alert or an audio output.

3. The method of claim 1 , comprising:

storing the sensitive information to a secure database.

4. The method of claim 1 , wherein the processing the input comprises determining that a voice of the input matches a known voice of a user;

wherein the determining the output is based at least on an identity of the user.

5. The method of claim 1 , further comprising putting the customer device in communication with another customer service representative device.

6. The method of claim 1 , wherein the output comprises an indication of fraud.

7. The method of claim 1 , wherein the processing the input comprises translating the input from one language to another language.

8. The method of claim 1 , comprising:

determining a result of the output; and

storing at least a portion of the input, the output, and the result of the output to the database, wherein at least the portion of the input, the output, and the result are stored as previously-received inputs, previously-delivered outputs, and previous results.

9. An artificial intelligence (AI) assistant configured to:

receive an input via a communication session between a customer device and a customer service representative device, wherein the communication session is associated with a workflow for managing communication between a customer service representative associated with the customer service representative device and a customer associated with the customer device, wherein the AI assistant is trained using machine learning on a database containing previously-received inputs, previously-delivered outputs, and the previous results of previously-delivered outputs;

process the input using at least the artificial intelligence;

determine, based on processing the input using at least the artificial intelligence, that at least a portion of the input includes sensitive information;

determine that an authorization level of the customer service representative is below a level required to receive the sensitive information;

track progress of the customer service representative through the workflow;

determine an output based on the processing of the input and the progress, wherein the output supplements at least one step in the workflow, wherein the output includes one of a suggested statement, change in tone, or correction for the customer service representative; and

send an indication of the output to the customer service representative device, wherein the output is censored to redact any sensitive information.

10. The AI assistant of claim 9 , wherein the communication session comprises at least one of a web chat session, a text message, a selection via a user interface, or a voice communication.

11. The AI assistant of claim 9 , wherein the output comprises a request for information.

12. The AI assistant of claim 9 , wherein the processing the input comprises performing at least one of automatic speech recognition (ASR), computer speech recognition, or natural language processing (NLP).

13. The AI assistant of claim 9 , wherein the processing the input comprises determining a tone associated with the input; and

wherein the output comprises a suggestion to change tone.

14. The AI assistant of claim 9 , wherein the AI assistant comprises a machine learning (ML) algorithm trained on inputs and outputs.

15. A non-transitory computer-readable storage medium comprising instructions that, when executed, cause operations comprising:

beginning a workflow involving communication between a customer service representative and a customer, wherein the workflow is provided to the customer service representative using an interface, and wherein the workflow comprises a plurality of steps in the communication;

receiving, by an artificial intelligence (AI) assistant, an input via a communication session between a customer device and a customer service representative device, wherein the communication session is associated with the workflow, wherein the AI assistant is trained using machine learning on a database containing previously-received inputs, previously-delivered outputs, and the previous results of previously-delivered outputs;

tracking, by the AI assistant, progress of the customer service representative through the workflow;

processing, by the AI assistant, the input;

determining, by the processing, that at least a portion of the input includes sensitive information;

determining that an authorization level of the customer service representative is below a level required to receive the sensitive information;

determining, by the AI assistant, an output based on the processing of the input and the progress, wherein the output supplements at least one step in the workflow, wherein the output includes one of a suggested statement, change in tone, or correction for the customer service representative; and

sending, by the AI assistant, an indication of the output to the customer service representative device, wherein the output is censored to redact any sensitive information.

16. The non-transitory computer-readable storage medium of claim 15 , wherein the determining the output is further based on an identity of a user associated with the customer device.

17. The non-transitory computer-readable storage medium of claim 15 , wherein the sending the indication of the output to the customer service representative device comprises sending the indication of the output to the customer service representative device and to the customer device.

18. The non-transitory computer-readable storage medium of claim 15 , wherein the AI assistant comprises one or more computing devices external to the customer device and the customer service representative device.

19. The non-transitory computer-readable storage medium of claim 15 , wherein the instructions, when executed, cause operations comprising:

determining that the customer service representative made an incorrect statement or omitted a statement, wherein the output comprises a correction to the incorrect statement or an instruction to provide the omitted statement.

20. The non-transitory computer-readable storage medium of claim 15 , wherein the instructions, when executed, cause operations comprising:

determining, by the processing, a change in a tone of the communication between the customer service representative and the customer, wherein the change in tone is based on analyzing words used, volume, pace, inflection, and pitch,

wherein the output is based on the change in tone.