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 .
Response to Arguments
Applicant's arguments filed 09/08/2025 have been fully considered but they are not persuasive.
Regarding the 101 rejections, on pages 12-14 of “Remarks” applicant contends that the amended claim 1 does not recite abstract ideas under Step 2A Prong 1. The examiner respectfully disagrees. Regarding claim 1, the previously identified limitations in the prior office action recite mental processes because under the broadest reasonable interpretation the identified limitations include a step of evaluation and judgement and could be performed mentally or with pen and paper which is either a mental process of evaluation/judgement (MPEP 2106)). For example, under the broadest reasonable interpretation, the limitation determines whether a workflow portion from a plurality of workflow portions is structured or unstructured can be performed mentally like making a determination on whether queries are formatted or not formatted. Similarly, under the broadest reasonable interpretation, the limitations of determining whether to make a structured or unstructured response based on the identified workflow type can be performed mentally like making formatted or unformatted outputs based on whether a query is formatted or not formatted. The judicial exception limitations in claim 1 are similar to the limitations in claims 8 and 14 and the reasons above apply to claims 8 and 14. Applicant argues that claim 8 cannot be performed mentally due to the mention of the judicial exceptions being performed on a no-code user interface. However, under the broadest reasonable interpretation, the no-code user interface is interpreted as merely reciting steps that apply generic computer components to perform an abstract idea, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)).
On pages 14-17 of “Remarks” applicant contends that the claims amount to significantly more than the judicial exceptions under Step 2B. The examiner respectfully disagrees. Applicant contends that the claimed invention’s ability to determine between structured and unstructured inputs can provide the improvement of “achieves results significantly faster and more reliably…eliminate the software engineering effort to capture the structured parameters from an unscripted, unstructured conversation” see Specification paragraph 46. However, the independent claims do not appear to claim the proposed technical improvements. For instance, claim 1, under the broadest reasonable interpretation, merely recites determining the type of workflow and then outputting different responses based on the type. Claim 1 does not appear to show the connection between the different types of workflows and achieving faster and more reliable results or eliminating the software engineering overhead. Additionally, applicant’s arguments appear to mention that the technical improvement is derived from the efficiency gained from having different outputs based on determining whether the workflow is structured or unstructured. It appears that the proposed improvement in this case is only realized because of the identified mental process limitation of determining between structured and unstructured workflows used in the claim. The judicial exception (mental process) itself cannot provide the improvement. See MPEP: “It is important to note, the judicial exception alone cannot provide the improvement. The improvement can be provided by one or more additional elements.” (MPEP 2106.05(a)). As noted in the previous office action, the independent claims recite additional elements that do not provide significantly more than the abstract idea. Therefore, applicant’s arguments regarding the 101 rejections are not persuasive.
Regarding the 103 rejections, applicant’s arguments about reference(s) Qin have been fully considered but are not persuasive.
Alleged no teaching of a list of structured parameters to be captured
In Remarks/Arguments pg. 10, applicant contends:
“Unlike independent claim 1, which recites "execute the workflow portion as unstructured based on the message, a task description and a list of structured parameters to be captured, when the workflow portion is determined to be unstructured, to produce an unstructured response via a generative artificial intelligence model," the cited references fail to disclose or suggest such a recitation. More specifically, the Office Action at pp. 27-28 acknowledges that Young does not disclose, but alleges that Qin discloses this claim recitation at paragraph [0019], which states "[i]n embodiments, the augmented prompt may include the original query, contextual information for answering the query, the retrieved augmentation information, and/or a request to answer the original query based on the contextual information and/or the retrieved augmentation information." The Office Action, however, is not clear on how this list of items in Qin is being mapped to the claim recitation "list of structured parameters to be captured". At best, Qin appears to merely disclose retrieval-augmented generation (RAG) with added contextual information for answering the original query, which is described in paragraph [0039]: In embodiments, contextual information 215 may describe the context of the user (e.g., user identifier, user role, user profile, user location, browsing history, etc.) and/or the context of the query (e.g., the current webpage, the product or service
associated with the current webpage, query timestamp information, etc.).
Paragraph [0046] further mentions that contextual information can include "e.g., the current webpage, the product or service of the current webpage, temporal information, location, etc." Nothing in this contextual information, however, discloses, suggests or even relates to "a list of structured parameters to be captured". In other words, providing contextual information to better retrieve documents to augment the query (e.g., from dataset(s) 112) or to generate an augmented prompt (e.g., augmented prompt 236) entirely fails to disclose or suggest structured parameters to be captured by such prompts. Neither Cai nor Baeuml remedy the deficiencies of Qin and Young.”
The relevant claim limitations appear to be: a task description and a list of structured parameters to be captured, in claim 1. As noted in the previous Office Action, Qin teaches:
(Qin, ⁋19, “In on implementation, a query may be used to retrieve pieces of augmentation information that may be included in a prompt to the LLM…In embodiments, the augmented prompt may include the original query, contextual information for answering the query, the retrieved augmentation information, and/or a request to answer the original query based on the contextual information and/or the retrieved augmentation information.”).
In other words, the elements that make up a prompt are interpreted as list of structured parameters to be captured because, under the broadest reasonable interpretation, a prompt is interpreted as having set fields and requirements to be entered before passing the prompt to a language model. Therefore, the applicant’s arguments are not persuasive.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding claim 1, in step 1 of the 101 analysis set forth in MPEP 2106, the claim recites A processor-readable non-transitory medium. The claim recites an article of manufacture. An article of manufacture method is one of the four statutory categories of invention.
In Step 2A, Prong 1 of the 101 analysis set forth in MPEP 2106, the examiner has determined that the following limitations recite a process that, under broadest reasonable interpretation, covers a mental process or mathematical concept but for the recitation of generic computer components:
determine whether a workflow portion from a plurality of workflow portions is structured or unstructured; (i.e., the broadest reasonable interpretation includes a step of evaluation and judgement and could be performed mentally like determining if a query is formatted or not formatted, which is either a mental process of evaluation/judgement (MPEP 2106)).
execute the workflow portion as structured based on the message when the workflow portion is determined to be structured, to produce a structured response; (i.e., the broadest reasonable interpretation includes a step of evaluation and judgement and could be performed mentally like giving a response to a formatted query, which is either a mental process of evaluation/judgement (MPEP 2106)).
and execute the workflow portion as unstructured based on the message, a task description and a list of structured parameters to be captured, when the workflow portion is determined to be unstructured, to produce an unstructured response… (i.e., the broadest reasonable interpretation includes a step of evaluation and judgement and could be performed mentally like giving a response to an unformatted query, which is either a mental process of evaluation/judgement (MPEP 2106)).
If the claim limitations, under their broadest reasonable interpretation, covers activities classified under Mental processes: concepts performed in the human mind (including observation, evaluation, judgement, or opinion) (see MPEP 2106.04(a)(2), subsection (III)) or Mathematical concepts: mathematical relationships, mathematical formulas or equations, or mathematical calculations (see MPEP 2106.04(a)(2), subsection (I)). Accordingly, the claim recites an abstract idea.
In Step 2A, Prong 2 of the 101 analysis, set forth in MPEP 2106, the examiner has determined that the following additional elements do not integrate this judicial exception into a practical application:
A processor-readable non-transitory medium storing code representing instructions to be executed by a processor, the code comprising code to cause the processor to: (i.e., the generic computer components recited in this limitation merely add the words “apply it”, or an equivalent, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f))).
repeat the following during a session with a user compute device: receive a message from the user compute device; (i.e., the broadest reasonable interpretation of receiving a message is mere data gathering, which is an insignificant extra solution activity (MPEP 2106.05(g))).
…via a generative artificial intelligence model. (i.e., the generic computer components recited in this limitation merely add the words “apply it”, or an equivalent, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f))).
Since the claim does not contain any other additional elements, that amount to integration into a practical application, the claim is directed to an abstract idea.
In Step 2B of the 101 analysis set forth in the 2019 PEG, the examiner has determined that the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception:
Regarding limitation (V), under the broadest reasonable interpretation, recite steps of mere data gathering, which has been recognized by the courts as being well-understood, routine, and conventional functions. Specifically, the courts have recognized computer functions directed to mere data gathering as well-understood, routine, and conventional functions when they are claimed in a merely generic manner or as insignificant extra-solution activity when considering evidence in view of Berkheimer v. HP, Inc., 881 F.3d 1360, 1368, 125 USPQ2d 1649, 1654 (Fed. Cir. 2018), see USPTO Berkheimer Memorandum (April 2018)).
Examiner uses Berkheimer: Option 2, a citation to one or more of the court decisions discussed in MPEP 2106.05(d)(II) as noting well-understood, routine, and conventional nature of the additional elements:
Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information); TLI Communications LLC v. AV Auto. LLC, 823 F.3d 607, 610, 118 USPQ2d 1744, 1745 (Fed. Cir. 2016) (using a telephone for image transmission); OIP Techs., Inc., v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014) (computer receives and sends information over a network). See MPEP 2106.05(d)(II).
Further, limitation (IV) merely recites steps that apply generic computer components to perform an abstract idea, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Similarly, limitation (VI) merely recite steps that apply a generic generative machine learning model, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Considering additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Regarding claim 2, it is dependent upon claim 1 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 2 recites maintain state while executing the workflow portion as structured and while executing the workflow portion as unstructured. Under the broadest reasonable interpretation, the limitation recites storing data while executing portions which is a step of evaluation and judgement which can be performed mentally or with pen and paper. The steps of evaluation and judgement are mental processes thus, claim 2 does not solve the deficiencies of claim 1.
Regarding claim 3, it is dependent upon claim 1 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 3 recites each workflow portion from the plurality of workflow portions is associated with at least one required parameter. Under the broadest reasonable interpretation, the limitation recites associating an element to another element which is a step of evaluation and judgement which can be performed mentally or with pen and paper. The steps of evaluation and judgement are mental processes thus, claim 3 does not solve the deficiencies of claim 1.
Regarding claim 4, it is dependent upon claim 3 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 4 recites each workflow portion from the plurality of workflow portions is further associated with at least one optional parameter. Under the broadest reasonable interpretation, the limitation recites associating an element to an optional element which is a step of evaluation and judgement which can be performed mentally or with pen and paper. The steps of evaluation and judgement are mental processes thus, claim 4 does not solve the deficiencies of claim 3.
Regarding claim 5, it is dependent upon claim 1 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 5 recites wherein the plurality of workflow portions is defined in a no-code user interface. Under the broadest reasonable interpretation, which merely recite steps that apply a generic visual user interface, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Therefore, claim 5 does not solve the deficiencies of claim 1.
Regarding claim 6, it is dependent upon claim 1 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 6 recites the session is a first session,the generative artificial intelligence model is a large language model (LLM),the unstructured response produced via the LLM includes a plurality of structured values associated with the list of structured parameters and during a second session between an entity and the LLM, the second session being shorter than the first session. Under the broadest reasonable interpretation, merely recite steps that apply a generic LLM to generate a response based on unformatted data, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Therefore, claim 6 does not solve the deficiencies of claim 1.
Regarding claim 7, it is dependent upon claim 1 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 7 recites wherein: the generative artificial intelligence model is a large language model (LLM). Under the broadest reasonable interpretation, merely recite steps that apply a generic LLM to generate a responses, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Claim 7 also recites and the code to execute the workflow portion as unstructured includes code to: send a plurality of prompts to the LLM based on the task description and the list of structured parameters associated with the workflow portion that is unstructured and that is from the plurality of workflow portions, and update state information associated with that workflow portion and output from the LLM based on the plurality of prompts. Under the broadest reasonable interpretation, the limitation recites sending a plurality of prompts to a model and updating data using outputs which are steps of evaluation and judgement which can be performed mentally or with pen and paper. The steps of evaluation and judgement are mental processes thus, claim 7 does not solve the deficiencies of claim 1.
Regarding claim 8, in step 1 of the 101 analysis set forth in MPEP 2106, the claim recites A processor-readable non-transitory medium. The claim recites an article of manufacture. An article of manufacture method is one of the four statutory categories of invention.
In Step 2A, Prong 1 of the 101 analysis set forth in MPEP 2106, the examiner has determined that the following limitations recite a process that, under broadest reasonable interpretation, covers a mental process or mathematical concept but for the recitation of generic computer components:
and define a plurality of workflow portions of a workflow based on the plurality of structured workflow portion indicators, the at least one unstructured workflow portion indicator, and the plurality of connection indicators. (i.e., the broadest reasonable interpretation includes a step of evaluation and judgement and could be performed mentally like determining if a query is formatted or unformatted based on an indicator, which is either a mental process of evaluation/judgement (MPEP 2106)).
If the claim limitations, under their broadest reasonable interpretation, covers activities classified under Mental processes: concepts performed in the human mind (including observation, evaluation, judgement, or opinion) (see MPEP 2106.04(a)(2), subsection (III)) or Mathematical concepts: mathematical relationships, mathematical formulas or equations, or mathematical calculations (see MPEP 2106.04(a)(2), subsection (I)). Accordingly, the claim recites an abstract idea.
In Step 2A, Prong 2 of the 101 analysis, set forth in MPEP 2106, the examiner has determined that the following additional elements do not integrate this judicial exception into a practical application:
A processor-readable non-transitory medium storing code representing instructions to be executed by a processor, the code comprising code to cause the processor to: (i.e., the generic computer components recited in this limitation merely add the words “apply it”, or an equivalent, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f))).
…through a no-code user interface… (i.e., the generic computer components recited in this limitation merely add the words “apply it”, or an equivalent, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f))).
receive,…, a plurality of structured workflow portion indicators associated with a plurality of structured workflow portions, receive,…, an indicator of at least one unstructured workflow portion indicator, (i.e., the broadest reasonable interpretation of receiving a plurality of indicators is mere data gathering, which is an insignificant extra solution activity (MPEP 2106.05(g))).
the at least one unstructured workflow portion indicator configured to receive a task description and a list of structured parameters that is to be identified by a generative artificial intelligence model, (i.e., the generic computer components recited in this limitation merely add the words “apply it”, or an equivalent, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f))).
receive,…, a plurality of connection indicators, each connection indicator from the plurality of connection indicators identifying a workflow order between at least two collectively of (1) at least one structured workflow portion from the plurality of structured workflow portions or (2) the at least one unstructured workflow portion, (i.e., the broadest reasonable interpretation of receiving a plurality of connection indicators is mere data gathering, which is an insignificant extra solution activity (MPEP 2106.05(g))).
Since the claim does not contain any other additional elements, that amount to integration into a practical application, the claim is directed to an abstract idea.
In Step 2B of the 101 analysis set forth in the 2019 PEG, the examiner has determined that the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception:
Regarding limitations (IV) and (VI), under the broadest reasonable interpretation, recite steps of mere data gathering, which has been recognized by the courts as being well-understood, routine, and conventional functions. Specifically, the courts have recognized computer functions directed to mere data gathering as well-understood, routine, and conventional functions when they are claimed in a merely generic manner or as insignificant extra-solution activity when considering evidence in view of Berkheimer v. HP, Inc., 881 F.3d 1360, 1368, 125 USPQ2d 1649, 1654 (Fed. Cir. 2018), see USPTO Berkheimer Memorandum (April 2018)).
Examiner uses Berkheimer: Option 2, a citation to one or more of the court decisions discussed in MPEP 2106.05(d)(II) as noting well-understood, routine, and conventional nature of the additional elements:
Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information); TLI Communications LLC v. AV Auto. LLC, 823 F.3d 607, 610, 118 USPQ2d 1744, 1745 (Fed. Cir. 2016) (using a telephone for image transmission); OIP Techs., Inc., v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014) (computer receives and sends information over a network). See MPEP 2106.05(d)(II).
Further, limitations (II) and (III) merely recites steps that apply generic computer components to perform an abstract idea, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Similarly, limitation (V) merely recites steps that apply a generic generative machine learning model, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Considering additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Regarding claim 9, it is dependent upon claim 8 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 9 recites the code to receive the plurality of structured workflow portion indicators includes code to receive the plurality of structured workflow portion indicators through a drag-and-drop function of the no-code user interface, the code to receive the indicator of the at least one unstructured workflow portion includes code to receive the indicator of the at least one unstructured workflow portion through the drag-and-drop function of the no-code user interface, and the code to receive the plurality of connection indicators includes the code to receive the plurality of connection indicators through the drop-and-drop function of the no-code user interface. Under the broadest reasonable interpretation, the limitations merely recite steps of mere data gathering, which has been recognized by the courts as being well-understood, routine, and conventional functions (MPEP 2106.05(d)). Therefore, claim 9 does not solve the deficiencies of claim 8.
Regarding claim 10, it is dependent upon claim 8 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 10 recites code to cause the processor to: receive a plurality of structured values based on the task description and the list of structured parameters, each structured value from the plurality of structured values being associated with a structured parameter from the plurality of structured parameters. Under the broadest reasonable interpretation, the limitations merely recite steps of mere data gathering, which has been recognized by the courts as being well-understood, routine, and conventional functions (MPEP 2106.05(d)). Therefore, claim 10 does not solve the deficiencies of claim 8.
Regarding claim 11, it is dependent upon claim 8 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 11 recites the generative artificial intelligence model is a large language model (LLM), and the task description indicates a task for completion by the LLM during execution of the at least one unstructured workflow portion, the task description associated with at least one structured value from a plurality of structured values that is obtained during a session by the LLM with an entity and that is associated with the plurality of structured parameters. Under the broadest reasonable interpretation, merely recite steps that apply a generic LLM to generate a response based on unformatted data, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Therefore, claim 11 does not solve the deficiencies of claim 8.
Regarding claim 12, it is dependent upon claim 8 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 12 recites the generative artificial intelligence model is a large language model (LLM), the task description listing a plurality of tasks for completion by the LLM during execution of the at least one unstructured workflow portion, each task from the plurality of tasks associated with at least one structured value from a plurality of structured values obtained during a session by the LLM with an entity,. Under the broadest reasonable interpretation, merely recite steps that apply a generic LLM to generate a response based on unformatted data, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Claim 12 further recites the code further comprises code to cause the processor to: receive, from the LLM, the plurality of structured values in response to the session by the LLM. Under the broadest reasonable interpretation, the limitation merely recites steps of mere data gathering, which has been recognized by the courts as being well-understood, routine, and conventional functions (MPEP 2106.05(d)). Therefore, claim 12 does not solve the deficiencies of claim 8.
Regarding claim 13, it is dependent upon claim 8 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 13 recites a connection indicator from the plurality of connection indicators identifies a first workflow order between a structured workflow portion from the plurality of structured workflow portions and an unstructured workflow portion from the at least one unstructured workflow portion, the first workflow order being the structured workflow portion, then the unstructured workflow portion, and then returning to the structured workflow portion. Under the broadest reasonable interpretation, the limitation recites determining a workflow order using formatted and unformatted queries which is a step of evaluation and judgement which can be performed mentally or with pen and paper. The steps of evaluation and judgement are mental processes thus, claim 13 does not solve the deficiencies of claim 8.
Regarding claim 14, in step 1 of the 101 analysis set forth in MPEP 2106, the claim recites A apparatus. The claim recites an apparatus. An apparatus is one of the four statutory categories of invention.
In Step 2A, Prong 1 of the 101 analysis set forth in MPEP 2106, the examiner has determined that the following limitations recite a process that, under broadest reasonable interpretation, covers a mental process or mathematical concept but for the recitation of generic computer components:
the structured-and-unstructured-logic module configured to determine, for each workflow portion from the plurality of workflow portions, whether the workflow portion is structured or unstructured, (i.e., the broadest reasonable interpretation includes a step of evaluation and judgement and could be performed mentally like determining if a query is formatted or not formatted, which is either a mental process of evaluation/judgement (MPEP 2106)).
in response to determining that a workflow portion from the plurality of workflow portions is structured, the structured-and-unstructured-logic module configured to send to the orchestration engine an indication that that workflow portion is structured, (i.e., the broadest reasonable interpretation includes a step of evaluation and judgement and could be performed mentally like flagging a portion as formatted, which is either a mental process of evaluation/judgement (MPEP 2106)).
in response to determining that a workflow portion from the plurality of workflow portions is unstructured, the structured-and-unstructured-logic module configured to send to the orchestration engine a task description and a list of structured parameters associated with that workflow portion, (i.e., the broadest reasonable interpretation includes a step of evaluation and judgement and could be performed mentally like flagging a portion as unformatted and requires further processing, which is either a mental process of evaluation/judgement (MPEP 2106)).
in response to receiving the indication that a workflow portion from the plurality of workflow portions is structured, the orchestration engine configured to execute that workflow portion as structured to produce a structured response, (i.e., the broadest reasonable interpretation includes a step of evaluation and judgement and could be performed mentally like giving a response to a formatted portion, which is either a mental process of evaluation/judgement (MPEP 2106)).
and (2) update state information associated with that workflow portion and output from the LLM based on the plurality of prompts. (i.e., the broadest reasonable interpretation includes a step of evaluation and judgement and could be performed mentally like storing data related to each portion, which is either a mental process of evaluation/judgement (MPEP 2106)).
If the claim limitations, under their broadest reasonable interpretation, covers activities classified under Mental processes: concepts performed in the human mind (including observation, evaluation, judgement, or opinion) (see MPEP 2106.04(a)(2), subsection (III)) or Mathematical concepts: mathematical relationships, mathematical formulas or equations, or mathematical calculations (see MPEP 2106.04(a)(2), subsection (I)). Accordingly, the claim recites an abstract idea.
In Step 2A, Prong 2 of the 101 analysis, set forth in MPEP 2106, the examiner has determined that the following additional elements do not integrate this judicial exception into a practical application:
An apparatus, comprising: a processor; and a memory coupled to the processor, the memory storing a structured-and-unstructured- logic module and an orchestration engine, (i.e., the generic computer components recited in this limitation merely add the words “apply it”, or an equivalent, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f))).
in response to receiving the task description and the list of structured parameters associated with a workflow portion that is from the plurality of workflow portions and this is unstructured, (i.e., the broadest reasonable interpretation of receiving a description and parameters is mere data gathering, which is an insignificant extra solution activity (MPEP 2106.05(g))).
the orchestration engine configured to (1) send a plurality of prompts to a large language model (LLM) based on the task description and the list of structured parameters associated with that workflow portion (i.e., the generic computer components recited in this limitation merely add the words “apply it”, or an equivalent, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f))).
Since the claim does not contain any other additional elements, that amount to integration into a practical application, the claim is directed to an abstract idea.
In Step 2B of the 101 analysis set forth in the 2019 PEG, the examiner has determined that the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception:
Regarding limitation (VII), under the broadest reasonable interpretation, recite steps of mere data gathering, which has been recognized by the courts as being well-understood, routine, and conventional functions. Specifically, the courts have recognized computer functions directed to mere data gathering as well-understood, routine, and conventional functions when they are claimed in a merely generic manner or as insignificant extra-solution activity when considering evidence in view of Berkheimer v. HP, Inc., 881 F.3d 1360, 1368, 125 USPQ2d 1649, 1654 (Fed. Cir. 2018), see USPTO Berkheimer Memorandum (April 2018)).
Examiner uses Berkheimer: Option 2, a citation to one or more of the court decisions discussed in MPEP 2106.05(d)(II) as noting well-understood, routine, and conventional nature of the additional elements:
Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information); TLI Communications LLC v. AV Auto. LLC, 823 F.3d 607, 610, 118 USPQ2d 1744, 1745 (Fed. Cir. 2016) (using a telephone for image transmission); OIP Techs., Inc., v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014) (computer receives and sends information over a network). See MPEP 2106.05(d)(II).
Further, limitation (VI) merely recites steps that apply generic computer components to perform an abstract idea, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Similarly, limitation (VIII) merely recites steps that apply a generic LLM, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Considering additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible.
Regarding claim 15, it is dependent upon claim 14 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 15 recites provide to the LLM each prompt from the plurality of prompts serially until the task description is satisfied and until a plurality of structured values associated with the list of structured parameters is received from the LLM. Under the broadest reasonable interpretation, the limitation recites determining to supply inputs one at a time until a criterion is satisfied which is a step of evaluation and judgement which can be performed mentally or with pen and paper. The steps of evaluation and judgement are mental processes thus, claim 15 does not solve the deficiencies of claim 14.
Regarding claim 16, it is dependent upon claim 14 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 16 recites send the plurality of prompts to the LLM to cause the LLM to send to a user compute device a first plurality of messages associated with the plurality of prompts. Under the broadest reasonable interpretation, merely recite steps that apply a generic LLM to generate a plurality of responses, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Claim 12 further recites and to receive from the user compute device a second plurality of messages in response to the first plurality of messages, and the orchestration engine is configured to receive the second plurality of messages from the LLM,. Under the broadest reasonable interpretation, the limitation merely recites steps of mere data gathering, which has been recognized by the courts as being well-understood, routine, and conventional functions (MPEP 2106.05(d)). Therefore, claim 12 does not solve the deficiencies of claim 8. Claim 16 also recites and update the state information based on the second plurality of messages. Under the broadest reasonable interpretation, the limitation recites updating stored information which is a step of evaluation and judgement which can be performed mentally or with pen and paper. The steps of evaluation and judgement are mental processes thus, claim 16 does not solve the deficiencies of claim 14.
Regarding claim 17, it is dependent upon claim 14 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 17 recites for each prompt from the plurality of prompts, (1) define that prompt based on the task description, the list of structured parameters and the state information at a first time,. Under the broadest reasonable interpretation, the limitation recites creating multiple prompts which is a step of evaluation and judgement which can be performed mentally or with pen and paper. Claim 17 also recites (2) send that prompt to the LLM and receive a response from the LLM based on that prompt and at a second time after the first time,. Under the broadest reasonable interpretation, merely recite steps that apply a generic LLM to generate a plurality of responses, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Claim 17 further recites and (3) update the state information at a third time after the second time based on the response for that prompt. Under the broadest reasonable interpretation, the limitation recites updating stored information which is a step of evaluation and judgement which can be performed mentally or with pen and paper. The steps of evaluation and judgement are mental processes thus, claim 17 does not solve the deficiencies of claim 14.
Regarding claim 18, it is dependent upon claim 14 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 18 recites iteratively repeat the following until the task description is satisfied by the LLM:(1) define a prompt from the plurality of based on the task description, any applicable prior response from the LLM, the list of structured parameters and the state information at that time,. Under the broadest reasonable interpretation, the limitation recites creating multiple prompts which is a step of evaluation and judgement which can be performed mentally or with pen and paper. Claim 18 also recites (2) send that prompt to the LLM and receive a response from the LLM based on that prompt,. Under the broadest reasonable interpretation, merely recite steps that apply a generic LLM to generate a plurality of responses, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Claim 18 further recites (3) update the state information at that time based on the response for that prompt,. Under the broadest reasonable interpretation, the limitation recites updating stored information which is a step of evaluation and judgement which can be performed mentally or with pen and paper. Claim 18 further recites (4) in response to the response satisfying the task description, ending the iteration and sending to the structured-and-unstructured-logic module a plurality of structured values associated with the list of structured parameters and based on the response and the any applicable prior responses from the LLM. Under the broadest reasonable interpretation, the limitation recites sending feedback data when an answer is provided which is a step of evaluation and judgement which can be performed mentally or with pen and paper. The steps of evaluation and judgement are mental processes thus, claim 18 does not solve the deficiencies of claim 14.
Regarding claim 19, it is dependent upon claim 14 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 19 recites memory further storing a no-code user interface, the no-code user interface configured to. Under the broadest reasonable interpretation, merely recite steps that apply a generic user interface, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Claim 19 further recites receive a plurality of workflow portion indicators associated with the plurality of workflow portions before execution of the structured-and- unstructured-logic module and the orchestration engine with respect to the plurality of workflow portions. Under the broadest reasonable interpretation, the limitation merely recites steps of mere data gathering, which has been recognized by the courts as being well-understood, routine, and conventional functions (MPEP 2106.05(d)). Therefore, claim 19 does not solve the deficiencies of claim 14.
Regarding claim 20, it is dependent upon claim 14 and fails to resolve the deficiencies identified above by integrating the judicial exception into a practical application, or introducing significantly more than the judicial exception. For example, claim 20 recites memory further storing a no-code user interface, the no-code user interface configured to. Under the broadest reasonable interpretation, merely recite steps that apply a generic user interface, which represents merely adding the words “apply it”, or an equivalent, which are not indicative of an inventive concept (MPEP 2106.05(f)). Claim 19 further recites receive (1) a plurality of structured workflow portion indicators associated with the plurality of workflow portions, (2) an indicator of at least one unstructured workflow portion associated with the plurality of workflow portions, and (3) a plurality of connection indicators associated with the plurality of structured workflow portion indicators and the indicator of the at least one unstructured workflow portion,. Under the broadest reasonable interpretation, the limitation merely recites steps of mere data gathering, which has been recognized by the courts as being well-understood, routine, and conventional functions (MPEP 2106.05(d)). Claim 20 further recites …define the plurality of workflow portions based on the plurality of structured workflow portion indicators, the indicator of the at least one unstructured workflow portion, and the plurality of connection indicators,…define the plurality of workflow portions before execution of the structured-and-unstructured-logic module and the orchestration engine with respect to the plurality of workflow portions. Under the broadest reasonable interpretation, the limitation recites determining whether a portion is formatted or unformatted before execution which is a step of evaluation and judgement which can be performed mentally or with pen and paper. The steps of evaluation and judgement are mental processes thus, claim 20 does not solve the deficiencies of claim 14.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-4 and 6-7 are rejected under 35 U.S.C. 103 as being unpatentable over Young, et al., Non-Patent Literature “Fusing Task-Oriented and Open-Domain Dialogues in Conversational Agents” (“Young”) in view of Qin, US Pre-Grant Publication 2024/0346256A1 (“Qin”).
Regarding claim 1, Young discloses:
repeat the following during a session with a user compute device: receive a message from the user compute device; (Young, pg. 11625 col. 2, “A multi-turn dialogue system generates a response R based on a multi-turn context C. In inter-mode dialogues [repeat the following during a session], C is composed of both TOD and ODD turns [with a user compute device: receive a message from the user compute device;]. In the FusedChat setting, R can be in either TOD mode or ODD mode, but has to be in only one of the two.”).
determine whether a workflow portion from a plurality of workflow portions is structured or unstructured; execute the workflow portion as structured based on the message when the workflow portion is determined to be structured, to produce a structured response; (Young, pg. 11623 col. 1, “We develop and evaluate two baseline models for this new setting: (1) The classification-based model. Two response generation models Mtod and Modd are independently trained on the turns of the respective modes. They generate the response of their respective mode given a conversational context; task-oriented dialogue, or TOD, is interpreted as a structured workflow portion (i.e. execute the workflow portion as structured based on the message when the workflow portion is determined to be structured, to produce a structured response;). A separate mode classification model C is trained and used to determine which mode to invoke given the context; TOD is interpreted as structured and open-domain dialogue, or ODD, is interpreted as unstructured (i.e. determine whether a workflow portion from a plurality of workflow portions is structured or unstructured;).”).
and execute the workflow portion as unstructured based on the message…when the workflow portion is determined to be unstructured, to produce an unstructured response via a generative artificial intelligence model. (Young, pg. 11625 col. 2, “(1) The classification-based model. Two response generation models Modd and Mtod are independently trained to handle each conversation mode. A separate classification model C is trained and used to determine which mode of model to invoke [and execute the workflow portion as unstructured based on the message…when the workflow portion is determined to be unstructured,] given an inter-mode context. Note that all 3 models above take inter-mode context as input. • For Modd, we follow (Shuster et al. 2019) and experiment with DialoGPT (Zhang et al. 2019) as the pretrained model [to produce an unstructured response via a generative artificial intelligence model.], fine-tuned on all ODD turns in FusedChat.”).
While Young teaches determining whether workflow portions are either structured or unstructured, Young does not explicitly teach:
A processor-readable non-transitory medium storing code representing instructions to be executed by a processor, the code comprising code to cause the processor to:
a task description and a list of structured parameters to be captured
Qin teaches:
A processor-readable non-transitory medium storing code representing instructions to be executed by a processor, the code comprising code to cause the processor to: (Qin, ⁋69, “may be each implemented as computer program code/instructions configured to be executed in one or more processors and stored in a computer readable storage medium [A processor-readable non-transitory medium storing code representing instructions to be executed by a processor, the code comprising code to cause the processor to:].”).
a task description and a list of structured parameters to be captured (Qin, ⁋19, “In on implementation, a query [a task description] may be used to retrieve pieces of augmentation information that may be included in a prompt to the LLM…In embodiments, the augmented prompt may include the original query, contextual information for answering the query, the retrieved augmentation information, and/or a request to answer the original query based on the contextual information and/or the retrieved augmentation information [and a list of structured parameters to be captured].”).
Young and Qin are both in the same field of endeavor (i.e. generative AI). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Young and Qin to teach the above limitation(s). The motivation for doing so is that incorporating augmentation information into LLMs improves the generated responses (cf. Qin, ⁋18, “Embodiments are disclosed herein that improve the scope and accuracy of responses generated by an LLM. For instance, in embodiments, an LLM may be augmented with augmentation information”).
Regarding claim 2, Young in view of Qin teaches the processor-readable non-transitory medium of claim 1. Young further teaches wherein the code further comprises code to cause the processor to: maintain state while executing the workflow portion as structured and while executing the workflow portion as unstructured. (Young, pg. 11626 col. 2, “We consider two context options: using only the latest user turn as the context (single-turn) or using the whole history containing multiple turns [maintain state while executing the workflow portion as structured and while executing the workflow portion as unstructured.] as the context (multi-turn).”).
Regarding claim 3, Young in view of Qin teaches the processor-readable non-transitory medium of claim 1. Young further teaches wherein: each workflow portion from the plurality of workflow portions is associated with at least one required parameter. (Young, pg. 11625 col. 2, “A separate classification model C is trained and used to determine which mode of model to invoke [wherein: each workflow portion from the plurality of workflow portions] given an inter-mode context. Note that all 3 models above take inter-mode context [is associated with at least one required parameter.] as input.”).
Regarding claim 4, Young in view of Qin teaches the processor-readable non-transitory medium of claim 3. Young further teaches wherein: each workflow portion from the plurality of workflow portions is further associated with at least one optional parameter. (Young, pg. 11626 col. 2, “We consider two context options: using only the latest user turn as the context (single-turn) or using the whole history containing multiple turns as the context (multi-turn) [is further associated with at least one optional parameter.]. Results show that the accuracy is quite high in both cases, with “multiturn” marginally outperforming “single-turn”.”).
Regarding claim 6, Young in view of Qin teaches the processor-readable non-transitory medium of claim 1. Qin further teaches:
wherein: the session is a first session, (Qin, ⁋49, “Flowchart 300 starts at step 302. In Step 302, a query [wherein: the session is a first session,] is received. For instance, pre-processor 202 of response generator 110 may receive query 216. As described above, In embodiments, query 216 may include a question in the form of a text string or voice data.”).
the generative artificial intelligence model is a large language model (LLM), the unstructured response produced via the LLM includes a plurality of structured values associated with the list of structured parameters (Qin, ⁋47, “In embodiments, LLM 214 [the generative artificial intelligence model is a large language model (LLM),] receives augmented prompt 236 from prompt generator 212 and generates response 238 [the unstructured response produced via the LLM]. For example, LLM 214 may process prompt augmented 236 to generate a response 238 based on contextual information 215 using augmentation information 232 [includes a plurality of structured values associated with the list of structured parameters].”).
and during a second session between an entity and the LLM, the second session being shorter than the first session. (Qin, ⁋46, “For example, prompt generator 212 may employ natural language processing (NLP) techniques to generate an augmented prompt 236 that requests LLM 214 to respond to query 216 based on contextual information using augmentation information 232. In embodiments, augmented prompt 236 may include, identify and/or link to augmentation information 232; the first session is interpreted as the query/question and providing a response is interpreted as the second session which is shorter than the first session as the processing the query includes the augmented information retrieval steps (i.e. and during a second session between an entity and the LLM, the second session being shorter than the first session.).”).
Regarding claim 7, Young in view of Qin teaches the processor-readable non-transitory medium of claim 1. Young further teaches and update state information associated with that workflow portion (Young, pg. 11626 col. 2, “We consider two context options: using only the latest user turn as the context (single-turn) or using the whole history containing multiple turns; the mention of saving the historical turns of the conversation is interpreted as updating the state information as the current session will be saved for future use (i.e. and update state information associated with that workflow portion) as the context (multi-turn).”).
Qin further teaches wherein: the generative artificial intelligence model is a large language model (LLM), and the code to execute the workflow portion as unstructured includes code to: send a plurality of prompts to the LLM based on the task description and the list of structured parameters associated with the workflow portion that is unstructured and that is from the plurality of workflow portions,…and output from the LLM based on the plurality of prompts. (Qin, ⁋18, “For instance, in embodiments, an LLM [wherein: the generative artificial intelligence model is a large language model (LLM),] may be augmented with augmentation information (e.g., domain-specific information; entity-specific information; product-specific information; recent information unavailable at generation of the large language model; or information changed after generation of the large language model) [based on the task description and the list of structured parameters associated with the workflow portion that is unstructured and that is from the plurality of workflow portions,]. A retrieval augmented generation (RAG) approach is disclosed herein that adds an information retrieval component to create augmented prompts [and the code to execute the workflow portion as unstructured includes code to: send a plurality of prompts to the LLM] to feed into the generative language model for generating the final answer/prediction [and output from the LLM based on the plurality of prompts.].”).
Claims 5 and 8-13 are rejected under 35 U.S.C. 103 as being unpatentable over Young, et al., Non-Patent Literature “Fusing Task-Oriented and Open-Domain Dialogues in Conversational Agents” (“Young”) in view of Qin, US Pre-Grant Publication 2024/0346256A1 (“Qin”) and further in view of Cai, et al., Non-Patent Literature “Low-code LLM: Graphical User Interface over Large Language Models” (“Cai”).
Regarding claim 5, Young in view of Qin teaches the processor-readable non-transitory medium of claim 1. While Young in view of Qin teaches determining structured and unstructured workflow portions using augmented data, the combination does not explicitly teach wherein the plurality of workflow portions is defined in a no-code user interface.
Cai teaches wherein the plurality of workflow portions is defined in a no-code user interface. (Cai, Abstract, “This paper introduces a novel human-LLM interaction framework, Low-code LLM. It incorporates six types of simple low-code visual programming interactions to achieve more controllable and stable responses; low-code interface is interpreted as a no-code interface (i.e. wherein the plurality of workflow portions is defined in a no-code user interface.).”).
Young, in view of Qin, and Cai are both in the same field of endeavor (i.e. language models). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Young, in view of Qin, and Cai to teach the above limitation(s). The motivation for doing so is that utilizing a low code user interface allows a user friendly way to edit workflows (cf. Cai, pg. 2 col. 1, “The visible workflow provides users with a clear understanding of how LLMs execute tasks, and enable users to easily edit it through a graphical user interface.”).
Regarding claim 8, Young discloses:
receive,…, a plurality of structured workflow portion indicators associated with a plurality of structured workflow portions, receive,…, an indicator of at least one unstructured workflow portion indicator, (Young, pg. 11625 col. 2, “(1) The classification-based model. Two response generation models Modd and Mtod are independently trained to handle each conversation mode. A separate classification model C is trained and used to determine which mode of model to invoke given an inter-mode context; inter-mode context is interpreted as the portion type indicator (i.e. receive,…, a plurality of structured workflow portion indicators associated with a plurality of structured workflow portions, receive,…, an indicator of at least one unstructured workflow portion indicator,). Note that all 3 models above take inter-mode context as input.”).
and define a plurality of workflow portions of a workflow based on the plurality of structured workflow portion indicators, the at least one unstructured workflow portion indicator, (Young, pg. 11625 col. 2, “(1) The classification-based model. Two response generation models Modd and Mtod are independently trained to handle each conversation mode. A separate classification model C is trained and used to determine which mode of model to invoke given an inter-mode context; classifying which mode/type of portion given the inter-mode context is interpreted as defining the portions based on indicators (i.e. and define a plurality of workflow portions of a workflow based on the plurality of structured workflow portion indicators, the at least one unstructured workflow portion indicator,). Note that all 3 models above take inter-mode context as input.”).
While Young teaches determining whether workflow portions are either structured or unstructured, Young does not explicitly teach:
A processor-readable non-transitory medium storing code representing instructions to be executed by a processor, the code comprising code to cause the processor to:
the at least one unstructured workflow portion indicator configured to receive a task description and a list of structured parameters that is to be identified by a generative artificial intelligence model,
through a no-code user interface
receive, through the no-code user interface, a plurality of connection indicators, each connection indicator from the plurality of connection indicators identifying a workflow order between at least two collectively of (1) at least one structured workflow portion from the plurality of structured workflow portions or (2) the at least one unstructured workflow portion, and the plurality of connection indicators.
Qin teaches:
A processor-readable non-transitory medium storing code representing instructions to be executed by a processor, the code comprising code to cause the processor to: (Qin, ⁋69, “may be each implemented as computer program code/instructions configured to be executed in one or more processors and stored in a computer readable storage medium [A processor-readable non-transitory medium storing code representing instructions to be executed by a processor, the code comprising code to cause the processor to:].”).
the at least one unstructured workflow portion indicator configured to receive a task description and a list of structured parameters that is to be identified by a generative artificial intelligence model, (Qin, ⁋19, “In on implementation, a query [the at least one unstructured workflow portion indicator configured to receive a task description] may be used to retrieve pieces of augmentation information that may be included in a prompt to the LLM…In embodiments, the augmented prompt may include the original query, contextual information for answering the query, the retrieved augmentation information, and/or a request to answer the original query based on the contextual information and/or the retrieved augmentation information [and a list of structured parameters that is to be identified by a generative artificial intelligence model,].”).
Young and Qin are both in the same field of endeavor (i.e. generative AI). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Young and Qin to teach the above limitation(s). The motivation for doing so is that incorporating augmentation information into LLMs improves the generated responses (cf. Qin, ⁋18, “Embodiments are disclosed herein that improve the scope and accuracy of responses generated by an LLM. For instance, in embodiments, an LLM may be augmented with augmentation information”).
While Young in view of Qin teaches determining structured and unstructured workflow portions using augmented data, the combination does not explicitly teach:
through a no-code user interface
receive, through the no-code user interface, a plurality of connection indicators, each connection indicator from the plurality of connection indicators identifying a workflow order between at least two collectively of (1) at least one structured workflow portion from the plurality of structured workflow portions or (2) the at least one unstructured workflow portion, and the plurality of connection indicators.
Cai teaches:
through a no-code user interface (Cai, Abstract, “This paper introduces a novel human-LLM interaction framework, Low-code LLM. It incorporates six types of simple low-code visual programming interactions to achieve more controllable and stable responses; low-code interface is interpreted as a no-code interface (i.e. wherein the plurality of workflow portions is defined in a no-code user interface.).”).
receive, through the no-code user interface, a plurality of connection indicators, each connection indicator from the plurality of connection indicators identifying a workflow order (Cai, pg. 2 col. 2, “A structured planning workflow is designed by the Planning LLM based on user input task prompt. Generally, the workflow consists of multiple steps and jump logic between steps…Planning LLM is instructed to produce structured workflows, as shown in Table 1, with every step consisting of two parts: (1) Step: including step name and step description that users can directly revise; (2) Jump logic; jump logic is interpreted as connection indicators as it determines the order of the workflow (i.e. receive, through the no-code user interface, a plurality of connection indicators, each connection indicator from the plurality of connection indicators identifying a workflow order).”).
between at least two collectively of (1) at least one structured workflow portion from the plurality of structured workflow portions or (2) the at least one unstructured workflow portion, (Cai, pg. 2 col. 2, “To facilitate the transformation from a workflow in natural language to an intuitive graphical flowchart, Planning LLM is instructed to produce structured workflows [between at least two collectively of (1) at least one structured workflow portion from the plurality of structured workflow portions], as shown in Table 1, with every step consisting of two parts: (1) Step: including step name and step description that users can directly revise; (2) Jump logic. Additionally, users can extend every step of the workflow into a sub-workflow with more details according to their preferences; sub-workflows are interpreted as unstructured workflow portions as they are created based on user preferences and thus open-ended (i.e. or (2) the at least one unstructured workflow portion,), and keep extending until reaching their desired level of detail.”).
and the plurality of connection indicators. (Cai, pg. 2 col. 2, “, Planning LLM is instructed to produce structured workflows, as shown in Table 1, with every step consisting of two parts: (1) Step: including step name and step description that users can directly revise; (2) Jump logic [and the plurality of connection indicators.].”).
Young, in view of Qin, and Cai are both in the same field of endeavor (i.e. language models). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Young, in view of Qin, and Cai to teach the above limitation(s). The motivation for doing so is that utilizing a low code user interface allows a user friendly way to edit workflows (cf. Cai, pg. 2 col. 1, “The visible workflow provides users with a clear understanding of how LLMs execute tasks, and enable users to easily edit it through a graphical user interface.”).
Regarding claim 9, Young in view of Qin and Cai teaches the processor-readable non-transitory medium of claim 8. Cai further teaches wherein: the code to receive the plurality of structured workflow portion indicators includes code to receive the plurality of structured workflow portion indicators through a drag-and-drop function of the no-code user interface, the code to receive the indicator of the at least one unstructured workflow portion includes code to receive the indicator of the at least one unstructured workflow portion through the drag-and-drop function of the no-code user interface, and the code to receive the plurality of connection indicators includes the code to receive the plurality of connection indicators through the drop-and-drop function of the no-code user interface. (Cai, pg. 2 col. 1, “As shown in Figure 1, human-LLM interaction can be completed through the following steps: (1) A Planning LLM generates a highly structured workflow for complex tasks. (2) Users edit the workflow using predefined low-code operations, which are all supported by clicking, dragging [wherein: the code to receive the plurality of structured workflow portion indicators includes code to receive the plurality of structured workflow portion indicators through a drag-and-drop function of the no-code user interface, the code to receive the indicator of the at least one unstructured workflow portion includes code to receive the indicator of the at least one unstructured workflow portion through the drag-and-drop function of the no-code user interface, and the code to receive the plurality of connection indicators includes the code to receive the plurality of connection indicators through the drop-and-drop function of the no-code user interface.], or text editing.”).
Regarding claim 10, Young in view of Qin and Cai teaches the processor-readable non-transitory medium of claim 8. Qin further teaches wherein the code further comprises code to cause the processor to: receive a plurality of structured values based on the task description and the list of structured parameters, each structured value from the plurality of structured values being associated with a structured parameter from the plurality of structured parameters. (Qin, ⁋19, “In on implementation, a query [receive a plurality of structured values based on the task description] may be used to retrieve pieces of augmentation information that may be included in a prompt to the LLM…In embodiments, the augmented prompt may include the original query, contextual information for answering the query, the retrieved augmentation information, and/or a request to answer the original query based on the contextual information and/or the retrieved augmentation information [and the list of structured parameters, each structured value from the plurality of structured values being associated with a structured parameter from the plurality of structured parameters.].”).
Regarding claim 11, Young in view of Qin and Cai teaches the processor-readable non-transitory medium of claim 8. Qin further teaches:
wherein: the generative artificial intelligence model is a large language model (LLM), and the task description indicates a task for completion by the LLM during execution of the at least one unstructured workflow portion, (Qin, ⁋19, “In on implementation, a query [and the task description indicates a task for completion by the LLM during execution of the at least one unstructured workflow portion,] may be used to retrieve pieces of augmentation information that may be included in a prompt to the LLM [the generative artificial intelligence model is a large language model (LLM),].”).
the task description associated with at least one structured value from a plurality of structured values that is obtained during a session by the LLM with an entity and that is associated with the plurality of structured parameters. (Qin, ⁋19, “In on implementation, a query [the task description] may be used to retrieve pieces of augmentation information [associated with at least one structured value from a plurality of structured values] that may be included in a prompt to the LLM. For instance, a query string may be encoded into a first feature vector that is compared to a plurality of second feature vectors to determine a subset of the second feature vectors that satisfy a predetermined condition (e.g., threshold similarity). Augmentation information corresponding to the determined subset of second feature vectors may be retrieved and included in an augmented prompt to the LLM [that is obtained during a session by the LLM with an entity and that is associated with the plurality of structured parameters.].”).
Regarding claim 12, Young in view of Qin and Cai teaches the processor-readable non-transitory medium of claim 8. Qin further teaches wherein: the generative artificial intelligence model is a large language model (LLM), the task description listing a plurality of tasks for completion by the LLM during execution of the at least one unstructured workflow portion, each task from the plurality of tasks associated with at least one structured value from a plurality of structured values obtained during a session by the LLM with an entity, the code further comprises code to cause the processor to: receive, from the LLM, the plurality of structured values in response to the session by the LLM. (Qin, ⁋18, “For instance, in embodiments, an LLM [wherein: the generative artificial intelligence model is a large language model (LLM),] may be augmented with augmentation information (e.g., domain-specific information; entity-specific information; product-specific information; recent information unavailable at generation of the large language model; or information changed after generation of the large language model) [each task from the plurality of tasks associated with at least one structured value from a plurality of structured values obtained during a session by the LLM with an entity,]. A retrieval augmented generation (RAG) approach is disclosed herein that adds an information retrieval component to create augmented prompts [the task description listing a plurality of tasks for completion by the LLM during execution of the at least one unstructured workflow portion,] to feed into the generative language model for generating the final answer/prediction [the code further comprises code to cause the processor to: receive, from the LLM, the plurality of structured values in response to the session by the LLM.].”).
Regarding claim 13, Young in view of Qin and Cai teaches the processor-readable non-transitory medium of claim 8. Cai further teaches:
wherein: a connection indicator from the plurality of connection indicators identifies a first workflow order between a structured workflow portion from the plurality of structured workflow portions and an unstructured workflow portion from the at least one unstructured workflow portion, (Cai, pg. 2 col. 2, “Planning LLM is instructed to produce structured workflows [between a structured workflow portion from the plurality of structured workflow portions], as shown in Table 1, with every step consisting of two parts: (1) Step: including step name and step description that users can directly revise; (2) Jump logic; jump logic is interpreted as the connection indicators as it tells a portion the next portion is feeds to (i.e. wherein: a connection indicator from the plurality of connection indicators identifies a first workflow order). Additionally, users can extend every step of the workflow into a sub-workflow [and an unstructured workflow portion from the at least one unstructured workflow portion,] with more details according to their preferences, and keep extending until reaching their desired level of detail.”).
the first workflow order being the structured workflow portion, then the unstructured workflow portion, and then returning to the structured workflow portion. (Cai, pg. 3 see Figure 2, In Figure 2, the extend sub-flowchart example shows an outline portion feeding into multiple write sub-portions and then the write sub-portions feeding into a proofread portion. The outline and proofread portions are interpreted as structured portions as these are generated without user preferences. The write sub-portions are interpreted as the unstructured portions as these are open-ended and generated based on user defined preferences. (i.e. the first workflow order being the structured workflow portion, then the unstructured workflow portion, and then returning to the structured workflow portion.)).
Claims 14-18 are rejected under 35 U.S.C. 103 as being unpatentable over Young, et al., Non-Patent Literature “Fusing Task-Oriented and Open-Domain Dialogues in Conversational Agents” (“Young”) in view of Qin, US Pre-Grant Publication 2024/0346256A1 (“Qin”) and further in view of Baeuml, et al., US Pre-Grant Publication 2025/0037711A1 (“Baeuml”).
Regarding claim 14, Young discloses:
a structured-and-unstructured-logic module and an orchestration engine, the structured-and-unstructured-logic module configured to determine, for each workflow portion from the plurality of workflow portions, whether the workflow portion is structured or unstructured, (Young, pg. 11625 col. 2, “(1) The classification-based model. Two response generation models Modd and Mtod are independently trained to handle each conversation mode; the two response models are interpreted as the orchestration engine (i.e. and an orchestration engine,). A separate classification model C [a structured-and-unstructured-logic module] is trained and used to determine which mode of model to invoke; TOD is interpreted as structured and ODD is interpreted as unstructured (i.e. the structured-and-unstructured-logic module configured to determine, for each workflow portion from the plurality of workflow portions, whether the workflow portion is structured or unstructured,) given an inter-mode context. Note that all 3 models above take inter-mode context as input.”).
in response to determining that a workflow portion from the plurality of workflow portions is structured, the structured-and-unstructured-logic module configured to send to the orchestration engine an indication that that workflow portion is structured, (Young, pg. 11625 col. 2, “(1) The classification-based model. Two response generation models Modd and Mtod are independently trained to handle each conversation mode. A separate classification model C is trained and used to determine which mode of model to invoke; if the model C determines that input is TOD, or structured, then it uses the TOD model (i.e. in response to determining that a workflow portion from the plurality of workflow portions is structured, the structured-and-unstructured-logic module configured to send to the orchestration engine an indication that that workflow portion is structured,) given an inter-mode context. Note that all 3 models above take inter-mode context as input.”).
in response to determining that a workflow portion from the plurality of workflow portions is unstructured, the structured-and-unstructured-logic module configured to send to the orchestration engine…, (Young, pg. 11625 col. 2, “(1) The classification-based model. Two response generation models Modd and Mtod are independently trained to handle each conversation mode. A separate classification model C is trained and used to determine which mode of model to invoke; if the model C determines that input is ODD, or unstructured, then it uses the ODD model (i.e. in response to determining that a workflow portion from the plurality of workflow portions is unstructured, the structured-and-unstructured-logic module configured to send to the orchestration engine…,) given an inter-mode context. Note that all 3 models above take inter-mode context as input.”).
in response to receiving the indication that a workflow portion from the plurality of workflow portions is structured, the orchestration engine configured to execute that workflow portion as structured to produce a structured response, (Young, pg. 11625 col. 2, “(1) The classification-based model that is composed of a mode classification model [in response to receiving the indication that a workflow portion from the plurality of workflow portions is structured,] and two response generation models for TOD [the orchestration engine configured to execute that workflow portion as structured to produce a structured response,] and ODD”).
While Young teaches determining whether workflow portions are either structured or unstructured, Young does not explicitly teach:
An apparatus, comprising: a processor; and a memory coupled to the processor, the memory storing
a task description and a list of structured parameters associated with that workflow portion
in response to receiving the task description and the list of structured parameters associated with a workflow portion that is from the plurality of workflow portions and this is unstructured, the orchestration engine configured to (1) send a plurality of prompts to a large language model (LLM) based on the task description and the list of structured parameters associated with that workflow portion
and (2) update state information associated with that workflow portion and output from the LLM based on the plurality of prompts.
Qin teaches:
An apparatus, comprising: a processor; and a memory coupled to the processor, the memory storing (Qin, ⁋69, “may be each implemented as computer program code/instructions configured to be executed in one or more processors and stored in a computer readable storage medium [An apparatus, comprising: a processor; and a memory coupled to the processor, the memory storing].”).
a task description and a list of structured parameters associated with that workflow portion (Qin, ⁋19, “In on implementation, a query [a task description] may be used to retrieve pieces of augmentation information that may be included in a prompt to the LLM…In embodiments, the augmented prompt may include the original query, contextual information for answering the query, the retrieved augmentation information, and/or a request to answer the original query based on the contextual information and/or the retrieved augmentation information [and a list of structured parameters associated with that workflow portion].”).
in response to receiving the task description and the list of structured parameters associated with a workflow portion that is from the plurality of workflow portions and this is unstructured, the orchestration engine configured to (1) send a plurality of prompts to a large language model (LLM) based on the task description and the list of structured parameters associated with that workflow portion (Qin, ⁋18, “For instance, in embodiments, an LLM may be augmented with augmentation information (e.g., domain-specific information; entity-specific information; product-specific information; recent information unavailable at generation of the large language model; or information changed after generation of the large language model) [in response to receiving the task description and the list of structured parameters associated with a workflow portion that is from the plurality of workflow portions and this is unstructured,]. A retrieval augmented generation (RAG) approach is disclosed herein that adds an information retrieval component to create augmented prompts to feed into the generative language model for generating the final answer/prediction [the orchestration engine configured to (1) send a plurality of prompts to a large language model (LLM) based on the task description and the list of structured parameters associated with that workflow portion].”).
Young and Qin are both in the same field of endeavor (i.e. generative AI). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Young and Qin to teach the above limitation(s). The motivation for doing so is that incorporating augmentation information into LLMs improves the generated responses (cf. Qin, ⁋18, “Embodiments are disclosed herein that improve the scope and accuracy of responses generated by an LLM. For instance, in embodiments, an LLM may be augmented with augmentation information”).
While Young in view of Qin teaches determining structured and unstructured workflow portions using augmented data, the combination does not explicitly teach:
and (2) update state information associated with that workflow portion and output from the LLM based on the plurality of prompts.
Baeuml teaches and (2) update state information associated with that workflow portion and output from the LLM based on the plurality of prompts. (Baeuml, ⁋51, “In additional or alternative versions of those implementations, the context 202 of the dialog session can be determined based on…dialog history of the ongoing dialog session; having dialog history of an ongoing dialog session is interpreted as updating state information associated with a workflow because the state information is constantly changing since the session is not complete (i.e. and (2) update state information associated with that workflow portion) between the user and the automated assistant 115 and/or historical dialog history of one or more prior dialog sessions between the user and the automated assistant 115 [and output from the LLM based on the plurality of prompts.]”).
Young, in view of Qin, and Baeuml are both in the same field of endeavor (i.e. conversational AI). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Young, in view of Qin, and Baeuml to teach the above limitation(s). The motivation for doing so is that keeping track of the previous dialog of an ongoing conversation improves the automated assistant’s ability to engage in natural feeling conversations with users (cf. Baeuml, ⁋20, “the techniques described herein enable the automated assistant to engage in natural conversations with a user during a dialog session. For instance, the automated assistant can generate modified assistant outputs using one or more LLM outputs that are more conversational in nature.”).
Regarding claim 15, Young in view of Qin and Baeuml teach the apparatus of claim 14. Qin further teaches wherein the orchestration engine is configured to provide to the LLM each prompt from the plurality of prompts serially until the task description is satisfied and until a plurality of structured values associated with the list of structured parameters is received from the LLM. (Qin, ⁋47, “In embodiments, LLM 214 receives augmented prompt 236 from prompt generator 212 and generates response 238; processing a single response for each prompt is interpreted as processing the prompts serially (i.e. wherein the orchestration engine is configured to provide to the LLM each prompt from the plurality of prompts serially). For example, LLM 214 may process prompt augmented 236 to generate a response 238 based on contextual information 215 using augmentation information 232 [and until a plurality of structured values associated with the list of structured parameters is received from the LLM.]…For instance, LLM 214 may respond by indicating that it does not know the answer to the query, by generating a response based on the information in its training data, and/or by asking the user to clarify their query; asking the user to clarify the query is interpreted as until satisfying the task description as the LLM needs further processing to get the answer (i.e. until the task description is satisfied).”).
Regarding claim 16, Young in view of Qin and Baeuml teach the apparatus of claim 14. Baeuml further teaches:
wherein: the orchestration engine is configured to send the plurality of prompts to the LLM to cause the LLM to send to a user compute device a first plurality of messages associated with the plurality of prompts (Baeuml, ⁋60, “At block 354, the system processes, using one or more LLMs, a given assistant query, of the plurality of assistant queries [send the plurality of prompts to the LLM], to generate one or more corresponding LLM outputs, where each of the one or more corresponding LLM outputs being predicted to be responsive to the given assistant query [to cause the LLM to send to a user compute device a first plurality of messages associated with the plurality of prompts].”).
and to receive from the user compute device a second plurality of messages in response to the first plurality of messages, and the orchestration engine is configured to receive the second plurality of messages from the LLM, and update the state information based on the second plurality of messages. (Baeuml, ⁋67, “In some implementations, and as indicated at block 358, the system may optionally receive user input to review and/or modify one or more of the corresponding LLM outputs [and to receive from the user compute device a second plurality of messages in response to the first plurality of messages,]. For example, a human reviewer can analyze the one or more corresponding LLM outputs generated using the one or more LLM models and modify one or more of the corresponding LLM outputs by changing one or more of the terms and/or phrases included in the one or more corresponding LLM outputs; changing the outputs of the LLM is interpreted as updating the state/historical information of a portion as a generated response is now changed (i.e. and the orchestration engine is configured to receive the second plurality of messages from the LLM, and update the state information based on the second plurality of messages.).”).
Regarding claim 17, Young in view of Qin and Baeuml teach the apparatus of claim 14. Baeuml further teaches:
wherein the orchestration engine is configured to, for each prompt from the plurality of prompts, (1) define that prompt based on the task description, the list of structured parameters and the state information at a first time, (Baeuml, ⁋59, “At block 352, the system obtains a plurality of assistant queries [for each prompt from the plurality of prompts, (1) define that prompt based on the task description,] that are directed to an automated assistant and a corresponding context [the list of structured parameters] of a corresponding prior dialog session for each of the plurality of assistant queries. For example, the system can cause the assistant activity engine 171 of the offline output modification engine of FIGS. 1 and 2 to obtain the plurality of assistant queries and the corresponding contexts of the prior dialog sessions [and the state information at a first time,] in which the plurality of assistant queries was received from, for example, assistant activity database 170A depicted in FIG. 1.”).
(2) send that prompt to the LLM and receive a response from the LLM based on that prompt and at a second time after the first time, (Baeuml, ⁋60, “At block 354, the system processes, using one or more LLMs, a given assistant query, of the plurality of assistant queries [(2) send that prompt to the LLM], to generate one or more corresponding LLM outputs, where each of the one or more corresponding LLM outputs being predicted to be responsive to the given assistant query [and receive a response from the LLM based on that prompt and at a second time after the first time,].”).
and (3) update the state information at a third time after the second time based on the response for that prompt. (Baeuml, ⁋67, “In some implementations, and as indicated at block 358, the system may optionally receive user input to review and/or modify one or more of the corresponding LLM outputs; changing the outputs of the LLM is interpreted as updating the state/historical information of a portion as a generated response is now changed (i.e. and (3) update the state information at a third time after the second time based on the response for that prompt.).”).
Regarding claim 18, Young in view of Qin and Baeuml teach the apparatus of claim 14. Baeuml further teaches:
wherein the orchestration engine is configured to iteratively repeat the following until the task description is satisfied by the LLM: (Baeuml, ⁋69, “if, at an iteration of block 360, the system determines that there is no additional assistant query included in the plurality of assistant queries obtained at block 352 that have not been processed using one or more of the LLMs [iteratively repeat the following until the task description is satisfied by the LLM:]”).
(1) define a prompt from the plurality of based on the task description, any applicable prior response from the LLM, the list of structured parameters and the state information at that time, (Baeuml, ⁋59, “At block 352, the system obtains a plurality of assistant queries [(1) define a prompt from the plurality of based on the task description,] that are directed to an automated assistant and a corresponding context [the list of structured parameters] of a corresponding prior dialog session for each of the plurality of assistant queries. For example, the system can cause the assistant activity engine 171 of the offline output modification engine of FIGS. 1 and 2 to obtain the plurality of assistant queries and the corresponding contexts of the prior dialog sessions [any applicable prior response from the LLM…and the state information at a first time,] in which the plurality of assistant queries was received from, for example, assistant activity database 170A depicted in FIG. 1.”).
(2) send that prompt to the LLM and receive a response from the LLM based on that prompt, (Baeuml, ⁋60, “At block 354, the system processes, using one or more LLMs, a given assistant query, of the plurality of assistant queries [(2) send that prompt to the LLM], to generate one or more corresponding LLM outputs, where each of the one or more corresponding LLM outputs being predicted to be responsive to the given assistant query [and receive a response from the LLM based on that prompt,].”).
(3) update the state information at that time based on the response for that prompt, (Baeuml, ⁋67, “In some implementations, and as indicated at block 358, the system may optionally receive user input to review and/or modify one or more of the corresponding LLM outputs; changing the outputs of the LLM is interpreted as updating the state/historical information of a portion as a generated response is now changed (i.e. (3) update the state information at that time based on the response for that prompt,).”).
(4) in response to the response satisfying the task description, ending the iteration… (Baeuml, ⁋74, “At block 366, the system causes the automated assistant to utilize one or more of the corresponding LLM outputs in generating one or more current assistant outputs to be provided for presentation to a user; presenting an output to the user is interpreted as satisfying the task description, or query (i.e. (4) in response to the response satisfying the task description, ending the iteration…) of the client device.”).
…and the any applicable prior responses from the LLM. (Baeuml, ⁋67, “Moreover, any non-discarded, re-indexed and/or curated LLM outputs can be utilized to modify or re-train the LLM in an offline manner […and the any applicable prior responses from the LLM.].”).
Young further teaches …and sending to the structured-and-unstructured-logic module a plurality of structured values associated with the list of structured parameters and based on the response… (Young, pg. 11625-11626, “For Modd, we follow (Shuster et al. 2019) and experiment with DialoGPT (Zhang et al. 2019) as the pretrained model, fine-tuned on all ODD turns in FusedChat. • For Mtod, we follow… Neural Pipeline (Ham et al. 2020) for such a model for Mtod, initialized with GPT2 and fine-tuned on all TOD turns in FusedChat; fine-tuning the TOD and ODD models is interpreted as sending structured values back to the structured-unstructured module based on a response because fine-tuning sends model parameter update values back to the models to update their performance (i.e. …and sending to the structured-and-unstructured-logic module a plurality of structured values associated with the list of structured parameters and based on the response…).”).
Claims 19 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Young, et al., Non-Patent Literature “Fusing Task-Oriented and Open-Domain Dialogues in Conversational Agents” (“Young”) in view of Qin, US Pre-Grant Publication 2024/0346256A1 (“Qin”) and further in view of Baeuml, et al., US Pre-Grant Publication 2025/0037711A1 (“Baeuml”) and Cai, et al., Non-Patent Literature “Low-code LLM: Graphical User Interface over Large Language Models” (“Cai”).
Regarding claim 19, Young in view of Qin and Baeuml teach the apparatus of claim 14. Young further teaches receive a plurality of workflow portion indicators associated with the plurality of workflow portions before execution of the structured-and-unstructured-logic module and the orchestration engine with respect to the plurality of workflow portions. (Young, pg. 11625 col. 2, “(1) The classification-based model. Two response generation models Modd and Mtod are independently trained to handle each conversation mode. A separate classification model C is trained and used to determine which mode of model to invoke given an inter-mode context; inter-mode context is interpreted as the portion type indicator and being giving the inter-mode context to model C is interpreted as receiving the indicators before execution of the logic module or the orchestration engine as model C dictates which model, TOD or ODD, to use (i.e. receive a plurality of workflow portion indicators associated with the plurality of workflow portions before execution of the structured-and-unstructured-logic module and the orchestration engine with respect to the plurality of workflow portions.). Note that all 3 models above take inter-mode context as input.”).
While Young in view of Qin and Baeuml teaches identifying structured and unstructured workflow portions, the combination does not explicitly teach wherein the memory further storing a no-code user interface, the no-code user interface configured to
Cai teaches wherein the memory further storing a no-code user interface, the no-code user interface configured to (Cai, Abstract, “This paper introduces a novel human-LLM interaction framework, Low-code LLM. It incorporates six types of simple low-code visual programming interactions to achieve more controllable and stable responses [wherein the memory further storing a no-code user interface, the no-code user interface configured to].”).
Young, in view of Qin and Baeuml, and Cai are both in the same field of endeavor (i.e. language models). It would have been obvious for a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Young, in view of Qin and Baeuml, and Cai to teach the above limitation(s). The motivation for doing so is that utilizing a low code user interface allows a user friendly way to edit workflows (cf. Cai, pg. 2 col. 1, “The visible workflow provides users with a clear understanding of how LLMs execute tasks, and enable users to easily edit it through a graphical user interface.”).
Regarding claim 20, Young in view of Qin and Baeuml teach the apparatus of claim 14. Young further teaches:
to receive (1) a plurality of structured workflow portion indicators associated with the plurality of workflow portions, (2) an indicator of at least one unstructured workflow portion associated with the plurality of workflow portions, (Young, pg. 11625 col. 2, “(1) The classification-based model. Two response generation models Modd and Mtod are independently trained to handle each conversation mode. A separate classification model C is trained and used to determine which mode of model to invoke given an inter-mode context; inter-mode context is interpreted as the portion type indicator (i.e. receive (1) a plurality of structured workflow portion indicators associated with the plurality of workflow portions, (2) an indicator of at least one unstructured workflow portion associated with the plurality of workflow portions,). Note that all 3 models above take inter-mode context as input.”).
to define the plurality of workflow portions based on the plurality of structured workflow portion indicators, the indicator of the at least one unstructured workflow portion, (Young, pg. 11625 col. 2, “(1) The classification-based model. Two response generation models Modd and Mtod are independently trained to handle each conversation mode. A separate classification model C is trained and used to determine which mode of model to invoke given an inter-mode context; classifying which mode/type of portion given the inter-mode context is interpreted as defining the portions based on indicators (i.e. to define a plurality of workflow portions based on the plurality of structured workflow portion indicators, the at least one unstructured workflow portion indicator,). Note that all 3 models above take inter-mode context as input.”).
While Young in view of Qin and Baeuml teaches identifying structured and unstructured workflow portions, the combination does not explicitly teach:
wherein: the memory further storing a no-code user interface, the no-code user interface configured
and (3) a plurality of connection indicators associated with the plurality of structured workflow portion indicators and the indicator of the at least one unstructured workflow portion, the no-code user interface configured…and the plurality of connection indicators, the no-code user interface configured to define the plurality of workflow portions before execution of the structured-and-unstructured-logic module and the orchestration engine with respect to the plurality of workflow portions.
Cai teaches:
wherein: the memory further storing a no-code user interface, the no-code user interface configured (Cai, Abstract, “This paper introduces a novel human-LLM interaction framework, Low-code LLM. It incorporates six types of simple low-code visual programming interactions to achieve more controllable and stable responses; low-code interface is interpreted as a no-code interface (i.e. wherein: the memory further storing a no-code user interface, the no-code user interface configured).”).
and (3) a plurality of connection indicators associated with the plurality of structured workflow portion indicators and the indicator of the at least one unstructured workflow portion, (Cai, pg. 2 col. 2, “A structured planning workflow is designed by the Planning LLM based on user input task prompt. Generally, the workflow consists of multiple steps and jump logic between steps…Planning LLM is instructed to produce structured workflows, as shown in Table 1, with every step consisting of two parts: (1) Step: including step name and step description that users can directly revise; (2) Jump logic; jump logic is interpreted as connection indicators as it determines the order of the workflow (i.e. and (3) a plurality of connection indicators associated with the plurality of structured workflow portion indicators and the indicator of the at least one unstructured workflow portion,).”).
the no-code user interface configured…and the plurality of connection indicators, (Cai, pg. 2 col. 2, “, Planning LLM is instructed to produce structured workflows, as shown in Table 1, with every step consisting of two parts: (1) Step: including step name and step description that users can directly revise; (2) Jump logic [and the plurality of connection indicators.].”).
the no-code user interface configured to define the plurality of workflow portions before execution of the structured-and-unstructured-logic module and the orchestration engine with respect to the plurality of workflow portions. (Cai, pg. 2 col. 1, “Then (1) a Planning LLM will design a workflow for completing the task. The workflow is a kind of structured plan, including execution procedure and jump logic. (2) The user will edit the workflow using six pre-defined low-code visual programming operations. (3) Once the user confirms workflow, it is interpreted into natural language and inputted to the Executing LLM, which will generate a response with the user’s guide; creating the workflow before using the executing LLM is interpreted as the no-code interface defining the workflow portions prior to execution by the of the structured-and-unstructured-logic module and the orchestration engine (i.e. the no-code user interface configured to define the plurality of workflow portions before execution of the structured-and-unstructured-logic module and the orchestration engine with respect to the plurality of workflow portions.).”).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Peter, et al., US20250104017A1 discloses a visual workflow builder as part of a micro-engagement system that allows for creation of workflows without writing any code. The micro-engagement platform is enhanced to support an environment where the content of the micro-engagement can be generated, presented and selected by an enterprise persona and launched to the end-users.
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to NICHOLAS S WU whose telephone number is (571)270-0939. The examiner can normally be reached Monday - Friday 8:00 am - 4:00 pm EST.
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, Michelle Bechtold can be reached on 571-431-0762. 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.
/N.S.W./Examiner, Art Unit 2148 /MICHELLE T BECHTOLD/Supervisory Patent Examiner, Art Unit 2148
Prosecution Insights