Determining sequences of interactions, process extraction, and robot generation using generative artificial intelligence / machine learning models
Use of generative artificial intelligence (AI)/machine learning (ML) models is disclosed to determine sequences of user interactions with computing systems, extract common processes, and generate robotic process automation (RPA) robots. The generative AI/ML model may be trained to recognize matching n-grams of user interactions and/or a beneficial end state. Recorded real user interactions may be analyzed, and matching sequences may be implemented as corresponding activities in an RPA workflow.
1 . A computer-implemented method, comprising:
providing, by a computing system, a generative artificial intelligence (AI)/machine learning (ML) model with data comprising time-ordered interactions of a plurality of users with respective user computing systems;
training the generative AI/ML model, by the computing system, to recognize related sequences of user interactions that pertain to tasks in the time-ordered sequences of user interactions of the plurality of users by comparing n-grams of sequences of user interactions in recorded data from the computing systems over a sliding window to find the related sequences, the n-grams comprising two or more values of n; and
deploying the trained generative AI/ML model.
2 . The computer-implemented method of claim 1 , wherein the deployed generative AI/ML model is configured to be called by one or more robotic process automation (RPA) robots.
3 . The computer-implemented method of claim 1 , wherein the training further comprises training the generative AI/ML model to determine a lowest value of n such that a majority of sequences of at least size n pertain to tasks performed by the users.
4 . The computer-implemented method of claim 1 , wherein a minimum number of related sequences is required for the generative AI/ML model to determine that the sequence pertains to a task.
5 . The computer-implemented method of claim 1 , wherein the training further comprises training the generative AI/ML model to determine a highest value of n such that n-grams above the highest value of n are not considered.
6 . The computer-implemented method of claim 1 , further comprising:
training the generative AI/ML model to use one or more importance metrics when determining whether a task is found in the related sequences, by the computing system.
7 . The computer-implemented method of claim 6 , wherein the one or more importance metrics comprise whether a related sequence of actions generates at least a certain amount of revenue, saves at least a certain amount of revenue, increases efficiency and/or speed of a task by at least a certain amount, speeds up a customer acquisition process, reduces a number of communications that are required, or a combination thereof.
8 . The computer-implemented method of claim 1 , further comprising:
training the generative AI/ML model or a robotic process automation (RPA) designer application to associate user interactions with RPA activities.
9 . The computer-implemented method of claim 8 , further comprising:
analyzing recorded real user interactions of a plurality of users with respective computing systems and determining sets of related sequences in the recorded real user interactions, by the generative AI/ML model; and
generating one or more respective robotic process automation (RPA) workflows comprising activities that implement user interactions of the determined sets of related sequences.
10 . The computer-implemented method of claim 9 , further comprising:
generating respective RPA robots implementing the one or more generated RPA workflows; and
deploying the one or more generated RPA robots to one or more user computing systems.
11 . The computer-implemented method of claim 9 , further comprising:
making the one or more generated RPA workflows accessible to one or more RPA designer applications.
12 . The computer-implemented method of claim 9 , wherein the determining of the sets of related sequences comprises:
generating a probability graph comprising associations between sequences of user interactions, by the generative AI/ML model;
pruning the probability graph to remove unrelated user interactions and sequences, by the generative AI/ML model; and
determining that sequences of user interactions match as part of a set of the set of related sequences based on the pruned probability graph, by the generative AI/ML model.
13 . The computer-implemented method of claim 12 , wherein the determination of whether the sequences match is performed using a Levenshtein distance, fuzzy matching, or a combination thereof.
14 . The computer-implemented method of claim 9 , further comprising:
determining a most efficient sequence of a set of related sequences to generate an RPA workflow based on one or more efficiency metrics; and
generating the respective RPA workflow using the most efficient sequence.
15 . The computer-implemented method of claim 14 , wherein a metric of the one or more efficiency metrics for the determination of the most efficient sequence is based on a number of user interactions in the sequence, a time associated with user interactions in the sequence, or a combination thereof.
16 . The computer-implemented method of claim 1 , wherein the generative AI model is a large language model (LLM), a generative adversarial network (GAN), a variational autoencoder (VAE), or a transformer.
17 . A non-transitory computer-readable medium storing a computer program, the computer program configured to cause at least one processor to:
analyze recorded real user interactions of a plurality of users with respective computing systems and determine sets of related sequences in the recorded real user interactions, by a generative artificial intelligence (AI)/machine learning (ML) model; and
generate one or more respective robotic process automation (RPA) workflows comprising activities that implement user interactions of the determined sets of related sequences, wherein
a minimum number of related sequences is required for the generative AI/ML model to determine that the sequence pertains to a task.
18 . The non-transitory computer-readable medium of claim 17 , wherein the determining of the sets of related sequences comprises:
generating a probability graph comprising associations between sequences of user interactions, by the generative AI/ML model;
pruning the probability graph to remove unrelated user interactions and sequences, by the generative AI/ML model; and
determining that sequences of user interactions match as part of a set of the set of related sequences based on the pruned probability graph, by the generative AI/ML model.
19 . The non-transitory computer-readable medium of claim 17 , wherein the computer program is further configured to cause the at least one processor to:
determine a most efficient sequence of the set of related sequences to generate an RPA workflow based on one or more efficiency metrics; and
generate the respective RPA workflow using the most efficient sequence.
20 . The non-transitory computer-readable medium of claim 19 , wherein a metric of the one or more efficiency metrics for the determination of the most efficient sequence is based on a number of user interactions in the sequence, a time associated with user interactions in the sequence, or a combination thereof.
21 . The non-transitory computer-readable medium of claim 17 , wherein
the determining of the sets of related sequences in the recorded real user interactions comprises using one or more importance metrics, and
the one or more importance metrics comprise whether a related sequence of actions generates at least a certain amount of revenue, saves at least a certain amount of revenue, increases efficiency and/or speed of a task by at least a certain amount, speeds up a customer acquisition process, reduces a number of communications that are required, or a combination thereof.
22 . The non-transitory computer-readable medium of claim 17 , wherein the generative AI model is a large language model (LLM), a generative adversarial network (GAN), a variational autoencoder (VAE), or a transformer.
23 . A computing system, comprising:
memory storing computer program instructions; and
at least one processor configured to execute the computer program instructions, wherein the computer program instructions are configured to cause the at least one processor to:
analyze recorded real user interactions of a plurality of users with respective computing systems and determine related sequences in the recorded real user interactions, by a generative artificial intelligence (AI)/machine learning (ML) model;
determine a most efficient sequence of the related sequences;
generate a graph or data model comprising associations between sequences of user interactions, by the generative AI/ML model, and
generate one or more respective robotic process automation (RPA) workflows comprising activities that implement user interactions of the determined most efficient related sequence.
24 . The computing system of claim 23 , wherein the determination of the most efficient sequence is based on a number of user interactions in the sequence, a time associated with user interactions in the sequence, or a combination thereof.
25 . The computing system of claim 23 , wherein
the determining of the related sequences in the recorded real user interactions comprises using one or more importance metrics, and
the one or more importance metrics comprise whether a related sequence of actions generates at least a certain amount of revenue, saves at least a certain amount of revenue, increases efficiency and/or speed of a task by at least a certain amount, speeds up a customer acquisition process, reduces a number of communications that are required, or a combination thereof.
26 . The computing system of claim 23 , wherein a minimum number of related sequences are required for generation of a respective RPA workflow.
27 . The computing system of claim 23 , wherein
a graph is generated comprising the associations between the sequences of user interactions, and
the graph is a probability graph.
28 . The computing system of claim 23 , wherein the generative AI model is a large language model (LLM), a generative adversarial network (GAN), a variational autoencoder (VAE), or a transformer.
29 . A computer-implemented method, comprising:
providing, by a computing system, a generative artificial intelligence (AI)/machine learning (ML) model with data comprising time-ordered interactions of a plurality of users with respective user computing systems;
training the generative AI/ML model, by the computing system, to recognize related sequences of user interactions that pertain to tasks in the time-ordered sequences of user interactions by comparing n-grams of sequences of user interactions in recorded data from the computing systems over a sliding window to find the related sequences, the n-grams comprising two or more values of n; and
deploying the trained generative AI/ML model, wherein
the training further comprises training the generative AI/ML model to determine a highest value of n such that n-grams above the highest value of n are not considered.
30 . The computer-implemented method of claim 29 , wherein the training further comprises training the generative AI/ML model to determine a lowest value of n such that a majority of sequences of at least size n pertain to tasks performed by the users.
31 . The computer-implemented method of claim 29 , wherein a minimum number of related sequences is required for the generative AI/ML model to determine that the sequence pertains to a task.
32 . The computer-implemented method of claim 29 , wherein the generative AI model is a large language model (LLM), a generative adversarial network (GAN), a variational autoencoder (VAE), or a transformer.