# US 12,762,000 B2 · Reduce recurring issues and incidents by remediation artificial intelligence > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,762,000 B2 - Document ID: US12762000B2 - Status: Granted - Kind: B2 - Application number: 18/515,854 - Filed: 2023-11-21 - Published: 2025-05-22 - Granted: 2026-10-13 - Earliest claimed: 2023-11-21 - Assignees: Paypal, Inc. - Applicants: PayPal, Inc. - Inventors: Sneha Venkatachalam; Jacqueline Lanxin Gao; Soodong Park; Ravi Retineni; Maneesh Divana; Yiheng Zhou - Web page: https://patentgrep.com/patent/US12762000B2 ## CPC - G06F 11/079: {Root cause analysis, i.e. error or fault diagnosis (in a hardware test environment G06F11/22; in a software test environment G06F11/36)} - G06F 11/0721: {within a central processing unit [CPU]} - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract Systems, methods, and computer program products for using artificial intelligence (AI) models, machine learning, and large language models to identify root causes of issues in a computing environment where multiple applications and computing devices operate. One or more AI models may determine adverse trends from one or more issue metrics, where an issue metric corresponds to issues occurring in a computing system. The AI models may identify issues corresponding to the adverse trends. From the identified issues, the AI models may determine root causes from the issues and the impacted area information and process information from the issues. From the issues and the impacted area information and process information from the issues the AI models may determine recommendations for rectifying the issues. The relations between the root causes, the impacted area information, the process information, and the issues may be formatted and displayed as a traversable network graph. ## Claims ### Claim 1 A method comprising: determining, using at least one artificial intelligence (AI) model, adverse trends from at least one issue metric, wherein the at least one issue metric is associated with issues occurring in a computing system; identifying a plurality of issues corresponding to the adverse trends in the computing system; determining, using an ensemble of AI models, a plurality of root causes from the plurality of issues, wherein the determining further comprises: generating, using a first AI model in the ensemble of AI models, similar historical issues from the plurality of issues; generating, using a second AI model in the ensemble of AI models and in parallel with the first AI model, root cause information from the plurality of issues; and generating, using a third AI model in the ensemble of AI models, the plurality of root causes for the plurality of issues from the similar historical issues and the root cause information; determining, using the at least one AI model, impacted area information and process information from the plurality of issues; determining, using the at least one AI model, at least one recommendation for rectifying the plurality of issues from the plurality of root causes, the impacted area information and the process information; and generating a traversable network graph comprising the plurality of root causes, the impacted area information, the process information, and the plurality of issues. ### Claim 2 (depends on 1) The method of claim 1, further comprising: generating, using the at least one AI model, a narrative summary summarizing at least one adverse trend in the adverse trends; and displaying the narrative summary on a user interface. ### Claim 3 (depends on 1) The method of claim 1, further comprising: generating, using the at least one AI model, root cause labels from the plurality of root causes, the impacted area information and the process information; generating root cause embeddings from the root cause labels; grouping, using a similarity function, the root cause labels into a plurality of groups, wherein each group in the plurality of groups includes a subset of the root cause labels; refining subsets of the root cause labels in the plurality of groups, wherein the refining identifies at least one root cause in the each group; and generating a root cause summary for the at least one root cause in the each group. ### Claim 4 (depends on 1) The method of claim 1, wherein determining the impacted area information and the process information from the plurality of issues further comprises: generating, using a fourth AI model in the at least one AI model, the impacted area information; and generating, using a fifth AI model in the at least one AI model, the process information. ### Claim 5 (depends on 1) The method of claim 1, wherein determining the at least one recommendation further comprises: generating, using the first AI model in the at least one AI model and the plurality of root causes, historical issues associated with historical root causes; and generating the at least one recommendation using a fourth AI model in the at least one AI model, the historical issues, the impacted area information and the process information. ### Claim 6 (depends on 1) The method of claim 1, wherein the traversable network graph further comprises a plurality of entities associated with the plurality of root causes, the plurality of entities connecting the plurality of root causes to the impacted area information or the process information. ### Claim 7 (depends on 6) The method of claim 6, wherein a size of a geometric figure representing an entity in the plurality of entities corresponds to a number of issues in the plurality of issues associated with the entity. ### Claim 8 (depends on 6) The method of claim 6, further comprising: displaying the traversable network graph on a user interface; receiving an input selecting an entity in the plurality of entities corresponding to a root cause in the plurality of root causes displayed in the traversable network graph; and in response to the input, displaying geometric shapes in the traversable network graph that correspond to the entity, a subset of issues of the plurality of issues corresponding to the entity, and the root cause. ### Claim 9 (depends on 8) The method of claim 8, further comprising: receiving a second input expanding the selected entity; and in response to the second input, displaying, in the traversable network graph, a subset of impacted areas and process information that correspond to the entity, and the subset of issues that correspond to the subset of impacted areas and process information. ### Claim 10 A system comprising: a non-transitory memory storing instructions; and one or more hardware processors coupled to the non-transitory memory and configured to read the instructions from the non-transitory memory to cause the system to perform operations comprising: determining, using a first generative artificial intelligence (AI) model, adverse trends from at least one issue metric; identifying a plurality of issues corresponding to the adverse trends in a computing environment; determining, using a first AI model in an ensemble of generative AI models, a plurality of root causes from the plurality of issues; determining, using a second AI model in the ensemble of generative AI models, impacted area information and process information from the plurality of issues; determining, using a third generative AI model in the ensemble of generative AI models, the plurality of root causes, the impacted area information and the process information, at least one recommendation for rectifying the plurality of issues; generating a traversable network graph comprising the plurality of root causes, the impacted area information, the process information, and the plurality of issues, wherein the impacted area information comprises impacted systems and products, and process information comprises impacted processes or functions; and displaying the traversable network graph on a user interface. ### Claim 11 (depends on 10) The system of claim 10, further comprising: generating, using a fourth generative AI model, narrative summaries summarizing a predefined number of the adverse trends; and displaying the narrative summaries on the user interface. ### Claim 12 (depends on 10) The system of claim 10, further comprising: generating, using at least one AI model, root cause labels from the plurality of root causes, the impacted area information and the process information; generating root cause embeddings from the root cause labels; grouping, using a similarity function, the root cause labels into a plurality of groups, wherein each group in the plurality of groups includes a subset of the root cause labels; refining subsets of the root cause labels in the plurality of groups, wherein the refining identifies at least one root cause in the each group; and generating a root cause summary for the at least one root cause in the each group. ### Claim 13 (depends on 10) The system of claim 10, wherein determining the impacted area information and the process information from the plurality of issues further comprises: generating, using the second AI model, the impacted area information and the process information. ### Claim 14 (depends on 10) The system of claim 10, wherein determining the at least one recommendation further comprises: generating, using the third generative AI model and the plurality of root causes, historical issues associated with historical root causes; and generating, using the third generative AI model, the historical issues, the impacted area information and the process information, the at least one recommendation. ### Claim 15 (depends on 10) The system of claim 10, wherein the traversable network graph further comprises a plurality of entities associated with the plurality of root causes, the plurality of entities connecting the plurality of root causes to the impacted area information and the process information, and wherein a size of a geometric figure representing an entity in the entities corresponds to a number of issues in the plurality of issues associated with the entity. ### Claim 16 A non-transitory machine-readable medium having stored thereon machine-readable instructions executable to cause a machine to perform operations comprising: determining, using a plurality of large language models (LLMs), adverse trends from at least one issue metric, wherein the at least one issue metric is associated with issues occurring in a computing system; identifying a plurality of issues corresponding to the adverse trends in the computing system; determining, using the plurality of LLMs, a plurality of root causes from the plurality of issues, wherein the determining further comprises: generating, using a first LLM in the plurality of LLMs, similar historical issues from the plurality of issues; generating, using a second LLM in the plurality of LLMs and in parallel with the first LLM, root cause information from the plurality of issues; and generating, using a third LLM in the plurality of LLMs, the plurality of root causes for the plurality of issues from the similar historical issues and the root cause information; determining, using the plurality of LLMs, impacted area information from the plurality of issues in parallel with determining process information from the plurality of issues; determining, using the plurality of LLMs, at least one recommendation for rectifying the plurality of issues from the plurality of root causes and the impacted area information and the process information; and generating a traversable network graph comprising the plurality of root causes, the impacted area information and the process information, and the plurality of issues. ### Claim 17 (depends on 16) The non-transitory machine-readable medium of claim 16, wherein the traversable network graph further comprises a plurality of entities associated with the plurality of root causes, the plurality of entities connecting the plurality of root causes to the impacted area information or the process information. ### Claim 18 (depends on 16) The non-transitory machine-readable medium of claim 16, wherein the impacted area information comprises impacted systems and products, and process information comprises impacted processes or functions. ### Claim 19 (depends on 16) The non-transitory machine-readable medium of claim 16, further comprising: receiving a selection of a node in the traversable network graph corresponding to an impacted area information; and automatically displaying nodes in the traversable network graph corresponding to at least one issue and at least one root cause associated with the selected node. ### Claim 20 (depends on 1) The method of claim 1, wherein the impacted area information comprises impacted systems and products, and process information comprises impacted processes or functions.