Using compliance analysis to manage customer interactions

1 . A method comprising:

receiving, using processing circuitry, interaction data of at least one interaction of a customer with an agent of an institution;

preprocessing, using the processing circuitry, the received interaction data;

generating, using the processing circuitry, a multidimensional numerical feature vector from the preprocessed interaction data, the multidimensional numerical feature vector being input to a machine learning model trained with historical data of a plurality of interactions between customers and agents of the institution;

using the processing circuitry, determining, using the machine learning model, that a regulated topic is represented in the at least one interaction based on the generated multidimensional numerical feature vector, the determining including the machine learning model mapping the multidimensional numerical feature vector to a regulated-topic representation;

in response to determining that the regulated topic is represented in the at least one interaction, generating, using the processing circuitry, a banker compliance score for the agent based on the at least one interaction and the regulated topic;

determining, using the processing circuitry, whether the banker compliance score satisfies a compliance standard specific to a branch location of the agent; and

in response to determining that the banker compliance score fails to satisfy the compliance standard specific to the branch location of the agent, automatically outputting, to an agent computing device, using the processing circuitry, a recommendation identifying a remediation for the branch location for future customer interactions.

2 . The method of claim 1 , further comprising:

generating a benchmark score for the institution based on a proportion of a number of interactions with the customer or a group of customers in which the compliance standard is not met.

3 . The method of claim 2 , further comprising generating at least one of a training plan or a remediation plan for a branch of the institution having a score lower than the benchmark score.

4 . The method of claim 2 , wherein a threshold for determining whether the compliance standard is met is based on a risk tolerance level of the institution.

5 . The method of claim 2 , wherein the compliance standard includes a regulation that governs financial interactions.

6 . The method of claim 5 , wherein the regulation regulates the financial interactions with a protected group.

7 . The method of claim 1 , wherein natural language is input as text by the agent of the institution.

8 . The method of claim 1 , wherein the interaction data includes at least one of a voice recording of the customer, of an interaction of the customer with the agent, or of the agent of the institution.

9 . The method of claim 1 , wherein the interaction data includes at least one of a text communication or a voice communication by at least one customer.

10 . The method of claim 1 , wherein determining that the regulated topic is represented in the at least one interaction includes implementing a natural language processing (NLP) algorithm.

11 . The method of claim 1 , wherein the recommendation includes a mitigation recommendation to mitigate an adverse event associated with the customer.

12 . A system for managing customer interactions with an institution, the system comprising:

a device configured to receive interaction data of at least one interaction of a customer with an agent of the institution; and

a processor configured to:

preprocess the received interaction data;

generate a multidimensional numerical feature vector from the preprocessed interaction data, the multidimensional numerical feature vector being input to a machine learning model trained with historical data of a plurality of interactions between customers and agents of the institution;

determine, using the machine learning model, that a regulated topic is represented in the at least one interaction based on the generated multidimensional numerical feature vector, the determining including the machine learning model mapping the multidimensional numerical feature vector to a regulated-topic representation;

in response to determination that the regulated topic is represented in the at least one interaction, generate a banker compliance score for the agent based on the at least one interaction and the regulated topic;

determine whether the banker compliance score satisfies a compliance standard specific to a branch location of the agent; and

in response to determining that the banker compliance score fails to satisfy with the compliance standard specific to the branch location of the agent, automatically output, to an agent computing device, a recommendation identifying a remediation for the branch location for future customer interactions.

13 . The system of claim 12 , wherein the processor is further configured to:

generate a benchmark score for the institution based on a proportion of a number of interactions with the customer or a group of customers in which the compliance standard is not met.

14 . The system of claim 13 , wherein the processor is further configured to generate at least one of a training plan or a remediation plan for branches of the institution having a score lower than the benchmark score.

15 . The system of claim 13 , wherein the compliance standard includes a regulation that governs financial interactions, and wherein the regulation regulates the financial interactions with a protected group.

16 . The system of claim 12 , wherein the interaction data includes at least one of a voice recording of the customer, of an interaction of the customer with the agent, or of the agent of the institution, and wherein determine that the regulated topic is represented in the at least one interaction includes implementing a natural language processing (NLP) algorithm.

17 . A non-transitory computer-readable medium including instructions that, when executed on a processor, cause the processor to perform operations including:

receiving interaction data of at least one interaction of a customer with an agent of an institution;

preprocessing the received interaction data;

generating a multidimensional numerical feature vector from the preprocessed interaction data, the multidimensional numerical feature vector being input to a machine learning model trained with historical data of a plurality of interactions between customers and agents of the institution;

determining, using the machine learning model, that a regulated topic is represented in the at least one interaction based on the generated multidimensional numerical feature vector, the determining including the machine learning model mapping the multidimensional numerical feature vector to a regulated-topic representation;

in response to determining that the regulated topic is represented in the at least one interaction, generating a banker compliance score for the agent based on the at least one interaction and the regulated topic;

determining whether the banker compliance score satisfies a compliance standard specific to a branch location of the agent; and

in response to determining that the banker compliance score fails to satisfy the compliance standard specific to the branch location of the agent, automatically outputting, to an agent computing device, a recommendation identifying a remediation for the branch location for future customer interactions.

18 . The non-transitory computer-readable medium of claim 17 , wherein the operations further include:

generating a benchmark score for the institution based on a proportion of a number of interactions with the customer or a group of customers in which the compliance standard is not met; and

generating at least one of a training plan or a remediation plan for branches of the institution having a score lower than the benchmark score, wherein a threshold for determining whether the compliance standard is met is based on a risk tolerance level of the institution.

19 . The non-transitory computer-readable medium of claim 17 , wherein the compliance standard includes a regulation that governs financial interactions, and wherein the regulation regulates the financial interactions with a protected group.

20 . The non-transitory computer-readable medium of claim 17 , wherein determining that the regulated topic is represented in the at least one interaction includes implementing a natural language processing (NLP) algorithm.