Granted Patent B1

US 12,737,370 · App. 18/888,945 · Granted Sep 15, 2026

Inventors: Vinayshekhar Bannihatti Kumar (San Jose, CA); Rashmi Gangadharaiah (San Jose, CA); Manoj Ghuhan Arivazhagan (Sunnyvale, CA); Sandesh Swamy (Seattle, WA); Sopan Khosla (Sunnyvale, CA); Siddharth Satish Goyal (Ashburn, VA); Narjessadat Seyeditabari (Daly City, CA); Deepak Seetharam Nadig (San Jose, CA); Pranjul Dubey (Lynnwood, WA); James W. Horsley (Carnation, WA)

Assignee: Amazon Technologies, Inc.

G06F16/248G06F16/2365

Abstract

Claims (84)

1 . A computer-implemented method, comprising:

receiving, a user query with respect to a computing resource of a provider network;

determining a plurality of organizational structure options for presenting content responsive to the user query;

generating an artificial intelligence (“AI”) prompt that includes the user query, a session context, the plurality of organizational structure options, and an instruction that an AI system provide a response to the user query that includes a visual structure selected from the plurality of organizational structure options for presenting the content included in the response and an action control that may be presented to a user such that the user may select the action control and cause an action to be performed with respect to the computing resource;

receiving, from the AI system, a visually structured response in accordance with the instruction; and

causing a presentation of:

the content of the visually structured response according to the visual structure specified in the visually structured response; and

a first action control corresponding to the computing resource referenced in the content.

2 . The computer-implemented method of claim 1 , wherein generating the AI prompt, includes:

generating a structure AI prompt with instructions that the AI system provide a visual structure for presentation of the content;

generating an actions AI prompt with instructions that the AI system provide a visual action for presentation of an action specified in the content; and

combining the visual structure presentation and the visual action presentation to generate the presentation.

3 . The computer-implemented method of claim 1 , further comprising:

determining a plurality of visual action options that may be used to present the action; and

including at least some of the visual action options in the AI prompt.

4 . The computer-implemented method of claim 1 , further comprising:

comparing a first embedding vector that semantically represents a first segment of a first document of a corpus of documents with a second embedding vector that semantically represents a second segment of a second document of the corpus to determine that the first embedding vector and the second embedding vector form a cluster that includes at least the first embedding vector and the second embedding vector;

in response to determining that the first embedding vector and the second embedding vector form the cluster:

processing, with the AI system, a first content of the first segment and a second content of the second segment to determine that the first content conflicts with the second content;

determining, with the AI system, a suggested revision to at least one of the first content or the second content so that the first content and the second content are aligned and do not conflict;

generating a canonical cluster segment that includes the suggested revision; and

presenting at least a portion of the suggested revision as responsive to the user query.

5 . The computer-implemented method of claim 1 , further comprising:

determining, based at least in part on the visually structured response, an action that may be performed with respect to the computing resource indicated in the content of the visually structured response; and

causing the presentation of the first action control that corresponds to the determined action, such that, when the first action control is selected, the action is performed with respect to the computing resource.

6 . The computer-implemented method of claim 5 , wherein the action modifies a state of the computing resource.

7 . A system, comprising:

one or more processors; and

a memory storing program instructions that, when executed by the one or more processors, cause the one or more processors to at least:

determine a plurality of organizational structure options for presenting content;

generate an artificial intelligence (“AI”) prompt with the plurality of organizational structure options and instructions that an AI system provide a visually structured response indicating an organizational structure selected from the plurality of organizational structure options as to how the content of the visually structured response is to be presented;

receive, from the AI system, the visually structured response that includes the content and the organizational structure indicating how the content is to be presented;

determine, based at least in part on the visually structured response, an action that may be performed with respect to a resource indicated in the content; and

cause, concurrent with a first presentation of the content according to the organizational structure, a second presentation of an action control such that, when the action control is selected, the action is performed with respect to the resource.

8 . The system of claim 7 , wherein the organizational structure is at least one of a table, a bullet list structure, a checklist structure, a graph structure, or a programming layout structure.

9 . The system of claim 7 , wherein the program instructions that, when executed by the one or more processors, further cause the one or more processors to at least:

determine a first segment of a first document of a document corpus that conflicts with a second segment of a second document of the document corpus;

send a consolidation AI prompt to an AI system that includes a first content of the first segment, a second content of the second segment, and an instruction that the AI system generate a suggested content that accurately describes the first content and the second content such that the first content and the second content are aligned;

update the first content and the second content to include the suggested content; and

cause, concurrent with the first presentation, a third presentation of at least a portion of the suggested content.

10 . The system of claim 9 , wherein the program instructions that, when executed by the one or more processors, further cause the one or more processors to at least:

generate a first embedding vector that semantically represents the first content of the first segment;

generate a second embedding vector that semantically represents the second content of the second segment; and

determine, based at least in part on a distance between the first embedding vector and the second embedding vector, that the first embedding vector and the second embedding vector form a cluster.

11 . The system of claim 7 , wherein the program instructions that, when executed by the one or more processors, cause the one or more processors to at least:

receive, a query with respect to an activity that may be performed during a session;

determine a session summary of the session; and

generate the AI prompt based at least in part on the query and the session summary.

12 . The system of claim 7 , wherein the program instructions that, when executed by the one or more processors, further cause the one or more processors to at least:

determine, based on two or more activities of a session, a session trajectory of the session;

generate a next best activity AI prompt that includes an indication of the session trajectory, a session summary of the session, and an instruction that a next best recommended activity be determined for the session;

receive, from the AI system and responsive to the next best activity AI prompt, a predicted next best action for the session; and

present the predicted next best action.

13 . The system of claim 7 , wherein the program instructions that, when executed by the one or more processors, further cause the one or more processors to at least:

determine a start activity and a destination activity for a flow;

determine that the start activity has been completed during a session;

update a session flow to include an indication of activities performed after the start activity;

determine that the destination activity has been completed as part of the session; and

in response to determination that the destination activity has been completed, generate at least one friction score for the flow.

14 . The system of claim 13 , wherein the friction score is at least one of an error rate, a page latency, a time to complete the flow, a number of page changes during the flow, or a number of activities of the flow.

15 . The system of claim 13 , wherein the program instructions that, when executed by the one or more processors, further cause the one or more processors to at least:

generate a complexity score indicative of a complexity of the session flow between the start activity and the destination activity.

16 . The system of claim 13 , wherein the program instructions that, when executed by the one or more processors, further cause the one or more processors to at least:

determine, based at least in part on the session flow and a plurality of other session flows between the start activity and the destination activity, a most efficient flow of activities between the start activity and the destination activity.

17 . A computer-implemented method, comprising:

determining a plurality of organizational structure options for presenting content;

generating an artificial intelligence (“AI”) system prompt with the plurality of organizational structure options and instructions that the AI system provide a response with the content and an organizational structure selected from the plurality of organizational structure options for presenting the content;

receiving, from the AI system, a response that includes the content and the organizational structure;

determining an action that may be performed with respect to at least one resource indicated in the content;

generating an action control representative of the action such that selection of the action control causes the action to be performed with respect to the at least one resource; and

causing a concurrent presentation of the content according to the organizational structure and the action control.

18 . The computer-implemented method of claim 17 , further comprising:

determining that the action control has been selected;

causing the action to be performed with respect to the at least one resource;

determining, based at least in part on performance of the action, a session trajectory of a session; and

determining, based at least in part on the session trajectory, a predicted next best action for the session.

19 . The computer-implemented method of claim 18 , wherein determining the predicted next best action is further determined based at least in part on one or more of the performance of the action, a session summary of the session, a user context, or a best practices determined for a flow that corresponds with the session trajectory.

20 . The computer-implemented method of claim 17 , further comprising:

generating, for a session, a session flow that includes an indication of activities performed during the session; and

determining, from the session flow, a friction score indicative of at least a number of activities between a start activity and a destination activity of the session flow.

21 . The computer-implemented method of claim 17 , further comprising:

receiving a query corresponding to the session; and

wherein generating the AI system prompt further includes:

generating the AI system prompt with instructions that the AI system provide a response to the query with content and the organizational structure for presenting the content.

Assignments (2)

References Cited (42)