Systems and methods for collection agent worklist generation

A non-transitory computer readable medium comprising computer executable instructions that, when executed by one or more processing units, cause the one or more processing units to:

• generate a series of collection worklist sets, wherein the generation of each collection worklist set in the series is separated from the generation of the preceding collection worklist set by a time interval, wherein each collection worklist in a collection worklist set is generated for one of a plurality of collection agents, wherein the plurality of collection agents are one or more software modules executing the collection worklist set, and wherein the generation of the collection worklist set for one time interval comprises;
• inputting a plurality of data inputs to the computing system by;
• generating with a first machine learning algorithm using customer data a list of multiple collection customers ordered according to a prioritized need for collection activity by accessing one or more databases comprising the customer data, by:
• configuring, on the computing system, a ranking engine to calculate a collection impact score for each of the collection customers by applying a weighted model, wherein the collection impact score is calculated by:
• calculating a plurality of derived ranking variables, including a Weighted Average Days Late (WADL) and a Weighted Clean Past Due (WCPD) from a plurality of direct ranking variables related to accounts receivable status of the collection customers;
• applying a weighting to the plurality of direct ranking variables and the plurality of derived ranking variables using the weighted model to create a plurality of weighted direct ranking variables and a plurality of weighted derived ranking variables;
• inputting the plurality of weighted direct ranking variables and the plurality of weighted derived ranking variables to the raking engine to calculate the collection impact score for each of the collection customer;
• and
• generating, using the configured ranking engine, a prioritized list of collection customers ordered based on the calculated collection impact score for each of the collection customer;
• obtaining a list of the plurality of collection agents;
• assigning a time available for each collection agent to perform collection actions within the time interval;
• entering a list of the collection actions;
• and
• associating a collection action duration with each of the collection actions;
• generating with a second machine learning algorithm, the collection worklist set for one time interval by processing the plurality of data inputs with the computing system by;
• calculating a collection action distribution, wherein the collection action distribution comprises a time allocation of the collection actions over a time interval;
• calculating an initial cadence, based upon the time available for each collection agent to perform collection actions within the time interval, and based upon the collection action distribution;
• calculating an action assignment based upon the initial cadence;
• assigning one of the collection customers to each of the collection actions on the action assignment, based at least in part upon the prioritized need for collection activity;
• generating an individual collection worklist for each collection agent from the action assignment;
• and
• delivering the collection worklist for each collection agent to each collection agent;
• monitor over multiple instances of the time interval, one or more collection agent productivity metrics by accessing one or more devices associated with the plurality of collection agents executing the collection worklist set;
• train a third machine learning algorithm to identify a positive trend or a negative trend in the one or more collection agent productivity metrics, based upon analysis, by the computing system, of the one or more collection agent productivity metrics;
• execute the third machine learning algorithm to autonomously identify a positive trend or a negative trend in the one or more collection agent productivity metrics, based upon analysis, by the computing system, of the one or more collection agent productivity metrics;
• train the third machine learning algorithm to change at least one of the plurality of data inputs to the computing system based upon the positive trend or the negative trend in the plurality of collection agent productivity metrics identified by the third machine learning algorithm;
• and
• execute the third machine learning algorithm to autonomously change at least one of the plurality of data inputs to the computing system based upon the positive trend or the negative trend in the plurality of collection agent productivity metrics identified by the third machine learning algorithm;
• wherein the third machine learning algorithm autonomously changes at least one of the plurality of data inputs to the computing system between generation of the series of collection worklist sets and generation of a next series of collection worklist sets;
• wherein executing the third machine learning algorithm to autonomously change at least one of the plurality of data inputs to the computing system comprises changing the first machine learning algorithm used to generate the list of multiple collection customers or changing customer data input into the first machine learning algorithm and changing the second machine learning algorithm used to generate the collection worklist set or changing at least one of the plurality of data inputs to the second machine learning algorithm;
• wherein the third machine learning algorithm autonomously modifies the ranking engine by changing one or more formulas used to calculate the plurality of derived ranking variables, the weighted model and a weighting factor applied by the weighted model, to calculate an updated collection impact score used for generation of an updated prioritized list for the next series of collection worklist sets;
• and
• execute the next series of collection worklist sets based on one or more changes implemented by the third machine learning algorithm to the first machine learning algorithm or the second machine learning algorithm, wherein the next series of collection worklist sets are different from the worklist collection set.

A computer system comprising:

• one or more processors;
• a non-transitory memory communicatively coupled to the one or more processors and storing instructions executable by the one or more processors to cause the one or more processors to:
• generate a series of collection worklist sets, wherein the generation of each collection worklist set in the series is separated from the generation of the preceding collection worklist set by a time interval, wherein each collection worklist in a collection worklist set is generated for one of a plurality of collection agents, wherein the plurality of collection agents are one or more software modules executing the collection worklist set, and wherein the generation of the collection worklist set for one time interval comprises;
• inputting a plurality of data inputs to the computing system by;
• generating with a first machine learning algorithm using customer data a list of multiple collection customers ordered according to a prioritized need for collection activity by accessing one or more databases comprising the customer data, by:
• configuring, on the computing system, a ranking engine to calculate a collection impact score for each of the collection customers by applying a weighted model, wherein the collection impact score is calculated by:
• calculating a plurality of derived ranking variables, including a Weighted Average Days Late (WADL) and a Weighted Clean Past Due (WCPD) from a plurality of direct ranking variables related to accounts receivable status of the collection customers;
• applying a weighting to the plurality of direct ranking variables and the plurality of derived ranking variables using the weighted model to create a plurality of weighted direct ranking variables and a plurality of weighted derived ranking variables;
• and
• inputting the plurality of weighted direct ranking variables and the plurality of weighted derived ranking variables to the raking engine to calculate the collection impact score for each of the collection customer;
• and
• generating, using the configured ranking engine, a prioritized list of collection customers ordered based on the calculated collection impact score for each of the collection customer;
• obtaining a list of the plurality of collection agents;
• assigning a time available for each collection agent to perform collection actions within the time interval;
• entering a list of the collection actions;
• and
• associating a collection action duration with each of the collection actions;
• generating with a second machine learning algorithm, the collection worklist set for one time interval by processing the plurality of data inputs with the computing system by;
• calculating a collection action distribution, wherein the collection action distribution comprises a time allocation of the collection actions over a time interval;
• calculating an initial cadence, based upon the time available for each collection agent to perform collection actions within the time interval, and based upon the collection action distribution;
• calculating an action assignment based upon the initial cadence;
• assigning one of the collection customers to each of the collection actions on the action assignment, based at least in part upon the prioritized need for collection activity;
• generating an individual collection worklist for each collection agent from the action assignment;
• and
• delivering the collection worklist for each collection agent to each collection agent;
• monitor over multiple instances of the time interval, one or more collection agent productivity metrics by accessing one or more devices associated with the plurality of collection agents executing the collection worklist set;
• train a third machine learning algorithm to identify a positive trend or a negative trend in the one or more collection agent productivity metrics, based upon analysis, by the computing system, of the one or more collection agent productivity metrics;
• execute the third machine learning algorithm to autonomously identify a positive trend or a negative trend in the one or more collection agent productivity metrics, based upon analysis, by the computing system, of the one or more collection agent productivity metrics;
• train the third machine learning algorithm to change at least one of the plurality of data inputs to the computing system based upon the positive trend or the negative trend in the plurality of collection agent productivity metrics identified by the third machine learning algorithm;
• and
• execute the third machine learning algorithm to autonomously change at least one of the plurality of data inputs to the computing system based upon the positive trend or the negative trend in the plurality of collection agent productivity metrics identified by the third machine learning algorithm;
• wherein the third machine learning algorithm autonomously changes at least one of the plurality of data inputs to the computing system between generation of the series of collection worklist sets and generation of a next series of collection worklist sets;
• wherein executing the third machine learning algorithm to autonomously change at least one of the plurality of data inputs to the computing system comprises changing the first machine learning algorithm used to generate the list of multiple collection customers or changing customer data input into the first machine learning algorithm and changing the second machine learning algorithm used to generate the collection worklist set or changing at least one of the plurality of data inputs to the second machine learning algorithm;
• wherein the third machine learning algorithm autonomously modifies the ranking engine by changing one or more formulas used to calculate the plurality of derived ranking variables, the weighted model and a weighting factor applied by the weighted model, to calculate an updated collection impact score used for generation of an updated prioritized list for the next series of collection worklist sets;
• and
• execute the next series of collection worklist sets based on one or more changes implemented by the third machine learning algorithm to the first machine learning algorithm or the second machine learning algorithm, wherein the next series of collection worklist sets are different from the worklist collection set.