DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Information Disclosure Statement
The information disclosure statements (IDS) were submitted on 2/23/26 & 6/10/26. The submissions are in compliance with the provisions of 37 CFR § 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 1-20 are rejected under 35 U.S.C. 102(a) (2) as being anticipated by Zhang et al. (US 20250190449, hereinafter Zhang).
Regarding Claim 1, Zhang discloses a method for pairing user queries with agents ([0004], provide a response to a user prompt that includes a question related to analysis of a structured data set in a natural language), the method performed by one or more processors of a computing system ([0025], computing device 102 includes one or more processors 104) and comprising:
receiving, from a user over a communications network, a plurality of queries ([0052], the agent orchestrator 218 performs recursive agent assignment, such as via a process of sequentially and iteratively choosing multiple generative AI agents from a pool of generative AI agents based on a defined order; [0043], After the computing device 102 has executed the ensemble model of one or more of the generative AI agents 126 and generated the response 110, the computing device 102 may output a graphical user interface (GUI) 174 that includes one or more visual elements (referred to herein collectively as “the visual elements 176”) based on the response 110]);
identifying a plurality of agents to which each query of the plurality of queries may be assigned ([0041], the agent orchestrator 124 may identify or extract certain features (e.g., the orchestrator features 118) from the prompt 170, the structured data set 154 (or descriptions thereof), and descriptions and historic performance of the generative AI agents 126, and based on the orchestrator features 118, the agent orchestrator 124 may select a single agent, or multiple agents, of the generative AI agents 126 that are likely to have the best performance in performing analytics tasks based on the prompt 170);
pairing each query with a corresponding agent of the plurality of agents based at least in part on a comparison of the respective query with agent descriptions associated with the plurality of agents ([0041], The agent features included in the orchestrator features 118 may include expertise level associated with the generative AI agents 126, historical performance metrics associated with the generative AI agents 126, response times recorded during operation of the generative AI agents 126, other information or metrics related to the generative AI agents 126; [0052], agent orchestrator 218 selects one or multiple generative AI agents based on the nature of the prompt 202 and the related use case, thus providing context-specific agent recommendations based on their individual characteristics and functionality and supports a generic architecture to plug in any new generative AI agent with minimal overhead; [0077], orchestrator 704 selects a single agent to perform a task, multiple agents to perform tasks concurrently, or a sequency of agents to perform sequential tasks, in order to generate the answer 718) ); and
transmitting, via a communications interface coupled to the computing system, each query to its corresponding agent ([0052], agent orchestrator 218 may ensure that the generative AI agents work in harmony by routing prompts and inputs to the most appropriate generative AI agent based on the context and expertise of each generative AI agent and prevents conflicts or redundancies between generative AI agents, and monitor the performance and health of individual generative AI agents, including making adjustments or dynamic changing agent selection when necessary, to ensure high availability and reliability; [0085], FIG. 9, decision node 906 is configured to route the user prompt and features 910 to the selected generative AI agent(s) responsible for handling the task).
PNG
media_image1.png
402
466
media_image1.png
Greyscale
Regarding Claim 2, Zhang discloses the method of claim 1, wherein each query is associated with a respective context and is paired with the corresponding agent based on the respective context (([0052], the agent orchestrator 218 is configured to perform recursive agent assignment, such as via a process of sequentially and iteratively choosing multiple generative AI agents from a pool of generative AI agents based on a defined order; [0087], The chain of AI thoughts 1012 represents AI “thoughts” during the process of selecting the response 1004 from the agent outputs 1002. As an initial thought, the chain of AI thoughts 1012 may include a single agent 1014 decision or a combined agents 1016 decision and determines the output of the highest-ranked agent is sufficient to answer the user question represented by the prompt).
Regarding Claim 3, Zhang discloses the method of claim 2, wherein the context includes one or more textual comments provided by the user ([0085], This determination is dynamic and is based on the prompt provided by the user, as well as based on factors associated with the agents themselves, such as whether a prompt requires a visualization (and if so, which agents are capable of outputting visualizations), the size of the data set and the performance of various agents for similar sized data sets (e.g., if a metric associated with an agent for a particular data set size satisfies a threshold, than it is selected),).
Regarding Claim 4, Zhang discloses the method of claim 2, wherein the context includes a browsing history of the user within a user assistance page associated with the computing system [(0085], decision node 906 may route the user prompt and features 910 to a single agent if a single agent is selected, to multiple agents if multiple agents are selected to perform tasks in parallel (e.g., concurrently), or to the first one or more agents of multiple agents if multiple agents are selected to perform tasks in a sequence).
Regarding Claim 5, Zhang discloses the method of claim 2, wherein the context is based at least in part on a type of application from which the plurality of queries are received from the user ([0086], The agent ranker 1010 is configured to rank the agent outputs 1002 (e.g., the agent responses) based on various criteria, such as the agent output itself, the relevance of the response to the user's question, historical performance of the generative AI agent that generated the response, other criteria, or a combination thereof. In some implementations, at least some of the various criteria may be weighted, and the ranking may be based on a weighted combination of the various criteria).
Regarding Claim 6, Zhang discloses the method of claim 2, wherein the plurality of queries are received during a portion of a conversation between the user and an automated assistant, and the context is based at least in part on one or more previous portions of the conversation ([0086], The agent ranker 1010 is configured to rank the agent outputs 1002 (e.g., the agent responses) based on various criteria, such as the agent output itself, the relevance of the response to the user's question, historical performance of the generative AI agent that generated the response, other criteria).
Regarding Claim 7, Zhang discloses the method of claim 2, wherein different agents of the plurality of agents are configured to generate responses to different queries associated with different contexts or different groups of contexts ([0052], the agent orchestrator 218 is configured to perform recursive agent assignment, such as via a process of sequentially and iteratively choosing multiple generative AI agents from a pool of generative AI agents based on a defined order).
Regarding Claim 8, Zhang discloses the method of claim 1, wherein each of the plurality of agents is associated with a corresponding large language model (LLM) ([0059], Professional agents 204 representative of LLM instances trained or fine-tuned to play a specific role in the workflow such as each professional agent 204 may be individually trained or fine-tuned to solve specific tasks related to user prompt 201 and are fine-tuned or trained to identify instructions assigned to it shared message pool 206 and understand the specific requirements and deliverables in the instructions. Each professional agent 204 agent may have access to one or more analytical models or tools 118 based on their role; [0099], Coordination agent 202 may initiate the first message-including defining the recipient professional agent (i.e., the professional agent(s) supposed to respond to that message), together with the requirements and deliverables; trained using query-and-response training data associated with a unique context or a unique group of contexts).
Regarding Claim 9, Zhang discloses the method of claim 1, wherein the comparison further includes comparing the queries with the agent descriptions associated with the plurality of agents using a large language model (LLM) ([0033] One type of optional disambiguation operations performed by the computing device 102 include data disambiguation operations to account for or improve performance of the generative AI agents 126 when performing analytics on the structured data 172. Other generative AI agents, such as large language model (LLM)-based agents, are often not able to handle a complex dataset due to multiple problems. For example, when a dataset has more than 10-15 columns, other LLM-based agents may “hallucinate” (e.g., return incorrect data in response to a question) or choose wrong fields from the dataset to answer questions. As another example, poorly labelled columns, such as abbreviated or proprietary metric-labeled columns, may cause other LLM-based agents to hallucinate or answer incorrectly. Conventional LLM-based agents also have reduced answer accuracy for questions that require personalization or slicing and filtering of data combined with calculations.
Regarding Claim 10, Zhang discloses the method of claim 1, wherein the comparison comprises: generating an embedded representation of each query of the plurality of queries; and comparing the embedded representations of each query with an embedded representation of the agent descriptions associated with the plurality of agent ([0038], if the prompt 170 is “Give me the average speed of top 3 runners”, and an entity that is mapped to a similar intent but that is not included in the prompt 170 is a “date/time” entity, the computing device 102 may provide a prompt to one of the generative AI models 128 to cause output of text content that indicates a date of the entry of the prompt, such as “on Nov. 5, 2023”, and this may be added to the prompt 170 to generate “Give me the average speed of the top 3 runners on Nov. 5, 2023” as the modified prompt 170]).
Regarding Claims 11-20, Computing system claims 11-20 of using the corresponding method claimed in claims 1-10, and the rejections of which are incorporated herein for the same reasons as used above.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Samuel D Fereja whose telephone number is (469)295-9243. The examiner can normally be reached 8AM-5PM.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, DAVID CZEKAJ can be reached at (571) 272-7327. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/SAMUEL D FEREJA/Primary Examiner, Art Unit 2487
