Systems and methods for discovery of automation opportunities

1. A method for analyzing computer actions on a computer desktop system to identify computer-based processes which are automation candidates, the method comprising using one or more computers:

using a data gathering process executed by a processor on a computer of the one or more computers, gathering low-level user action computer information items describing user input to the computer, each low-level user action computer information item comprising an input type description, a username of the person inputting the user action, and screen window information; and

at a computer of the one or more computers:

applying pattern mining to determine a set of subprocesses, each subprocess comprising a series of low-level user actions, each user action associated with a user action vector, and each subprocess associated with a subprocess vector generated from user action vectors associated with low-level user actions comprised in the subprocess;

clustering the subprocess vectors to create a new plurality of processes;

grouping the new plurality of processes using contexts using a learning algorithm;

using the learning algorithm, continuously updating the new plurality of processes for each low-level user actions;

for each of the new plurality of processes, calculating an automation score using the actions comprised in the subprocesses comprised in the process;

determining whether the automation score of the new plurality of processes has passed a support threshold; and

once a support threshold has been passed, creating an automation bot through a remote server, wherein the automation bot automatically performs the new process within the plurality of new processes, by replacing the low-level user actions in the new process with automated bot actions.

2. The method of claim 1 , wherein calculating an automation score for a process is based on the time spent for each subprocess comprised in the process and a quality score for each low-level user action comprised in each subprocess comprised in the process.

3. The method of claim 1 , comprising creating a user action vector by:

creating a text string from at least the input type description, the name of the application executing the screen window, and the screen window information; and

calculating as the user action vector a word embedding of the text string.

4. The method of claim 1 comprising displaying a visualization for a process including the automation score.

5. The method of claim 1 wherein the pattern mining comprises sequential pattern mining.

6. The method of claim 1 wherein the learning algorithm comprises a TD-IDF (term frequency-inverse document frequency) learning algorithm.

7. The method of claim 1 comprising ordering subprocesses using a maximum likelihood estimator.

8. The method of claim 1 wherein determining whether the automation score of the new plurality of processes has passed a support threshold is performed using a PrefixSpan algorithm.

9. A method for identifying automation candidates on a computer desktop system, the method comprising:

using a data gathering process executed by a processor on a computer of one or more computers, gathering low-level user action computer information items describing user input to the computer, each low-level user action computer information item comprising a username of the person inputting the user action, and screen window information; and

at a computer of the one or more computers:

determining a set of subprocesses, each subprocess comprising a series of low-level user actions, each user action associated with a user action vector, and each subprocess associated with a subprocess vector generated from user action vectors;

clustering the subprocess vectors to create a new plurality of processes;

using a learning algorithm, updating the new plurality of processes for each low-level user actions;

for each of the new plurality of processes, calculating an automation score using the actions comprised in the subprocesses comprised in the process;

determining whether the automation score of the new plurality of processes has passed a support threshold; and

if a support threshold has been passed, creating an automation bot through a remote server, wherein the automation bot automatically performs the new process within the plurality of new processes, by replacing the low-level user actions in the new process with bot actions.

10. The method of claim 9 , wherein calculating an automation score for a process is based on the time spent for each subprocess comprised in the process and a quality score for each low-level user action comprised in each subprocess comprised in the process.

11. The method of claim 9 , comprising creating a user action vector by:

creating a text string from at least the input type description, the name of the application executing the screen window, and the window information; and

calculating as the user action vector a word embedding of the text string.

12. The method of claim 9 comprising displaying a visualization for a process including the score.

13. The method of claim 9 wherein the pattern mining comprises sequential pattern mining.

14. The method of claim 9 wherein the learning algorithm comprises a TD-IDF (term frequency-inverse document frequency) learning algorithm.

15. A system for analyzing computer actions on a computer desktop system to identify computer-based processes which are automation candidates, the system comprising:

a first processor configured to gather low-level user action computer information items describing user input to a computer, each low-level user action computer information item comprising an input type description and window information; and

a second processor configured to:

apply pattern mining to determine a set of subprocesses, each subprocess comprising a series of low-level user actions, each user action associated with a user action vector, and each subprocess associated with a subprocess vector generated from user action vectors associated with low-level user actions comprised in the subprocess;

cluster the subprocess vectors to create a new plurality of processes;

group the new plurality of processes using contexts using a learning algorithm;

using the learning algorithm, update the new plurality of processes for each low-level user actions;

for each of the new plurality of processes, calculate an automation score;

determine, using a PrefixSpan algorithm, whether the automation score of the new plurality of processes has passed a support threshold; and

once a support threshold has been passed, create an automation bot through a remote server, wherein the automation bot automatically performs the new process within the plurality of new processes, by replacing the low-level user actions in the new process with automated bot actions.

16. The system of claim 15 , wherein the second processor is configured to calculate an automation score for a process based on the time spent for each subprocess comprised in the process and a quality score for each low-level user action comprised in each subprocess comprised in the process.

17. The system of claim 15 , wherein the second processor is configured to create a user action vector by:

creating a text string from at least the input type description, the name of the application executing the screen window, and the screen window information; and

calculating as the user action vector a word embedding of the text string.

18. The system of claim 15 , wherein the second processor is configured to display a visualization for a process including the automation score.

19. The system of claim 15 , wherein the pattern mining comprises sequential pattern mining.

20. The system of claim 15 , wherein calculating an automation score is performed using a duration for the process.