Prosecution Insights
Last updated: November 04, 2026
Application No. 19/176,100

TECHNIQUES FOR AN ARTIFICIAL INTELLIGENCE ASSISTANT FOR FAMILY HISTORY AND STORYTELLING

Non-Final OA §102§103
Filed
Apr 10, 2025
Priority
Apr 10, 2024 — provisional 63/632,457 +1 more
Examiner
GEORGANDELLIS, ANDREW C
Art Unit
2459
Tech Center
2400 — Computer Networks
Assignee
World Archives Holdings LLC
OA Round
1 (Non-Final)
56%
Grant Probability
Moderate
1-2
OA Rounds
2y 5m
Est. Remaining
97%
With Interview

Examiner Intelligence

Grants 56% of resolved cases
56%
Career Allowance Rate
282 granted / 499 resolved
-1.5% vs TC avg
Strong +40% interview lift
Without
With
+40.2%
Interview Lift
resolved cases with interview
Typical timeline
4y 0m
Avg Prosecution
12 currently pending
Career history
521
Total Applications
across all art units

Statute-Specific Performance

§101
8.3%
-31.7% vs TC avg
§103
52.8%
+12.8% vs TC avg
§102
18.3%
-21.7% vs TC avg
§112
18.7%
-21.3% vs TC avg
Black line = Tech Center average estimate • Based on career data from 499 resolved cases

Office Action

§102 §103
DETAILED ACTION Status of the Claims Claims 1-20 are pending. No claims have been canceled. No claims have been withdrawn. No claims have been objected to. No claims have been amended. No claims are allowable. Claims 1, 2, 7-10, 13, 15, 16, and 20 are rejected under 35 U.S.C. § 102; claims 3-6, 11, 12, 14, and 17-19 are rejected under 35 U.S.C. § 103. As an alternate ground, claims 1, 7, 9, 13, 15, and 20 are additionally rejected under 35 U.S.C. § 102, and as an alternate ground, claims 2-6, 8, 10-12, 14, and 16-19 are additionally rejected under 35 U.S.C. § 103. Other Prior Art Sejpal et al. (US 11,514,894 B2) discloses an artificial intelligence engine that receives a customer utterance in a conversation, converts the utterance to text, determines a user model and conversation state, selects a response using a reinforcement-learning-based module, and provides the response to the customer (col. 31, lines 25-54). Koritala (US 2021/0117488 A1) discloses a Google Dialogflow AI bot trained with a predefined set of biographical interview questions that accepts user answers by text or voice and compiles the answers into a biography ([0026]-[0030]). Claim Rejections - 35 U.S.C. § 102 The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action: A person shall be entitled to a patent unless - (a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale or otherwise available to the public before the effective filing date of the claimed invention. (a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention. Claims 1, 2, 7-10, 13, 15, 16, and 20 are rejected under 35 U.S.C. § 102(a)(2) as being anticipated by Gustman et al. (US 2025/0063239 A1; “Gustman”). Regarding claims 1, 15, and 20, Gustman teaches an apparatus, comprising: at least one memory; and at least one processor coupled with the at least one memory and configured to cause the apparatus to: present an initial prompt to a user. Gustman discloses that the interviewer and interviewee may be the same person ([0044]). Generated questions are presented in a question timeline that the interviewer uses to determine which question to ask next ([0059]-[0061]; fig. 5B), and the recording GUI reproduces the question being asked in text while the interviewee responds ([0067]; fig. 5C). receive a response to the prompt from the user. After the question is presented, the interactive video generation engine records the interviewee’s answer as a video segment ([0065]). generate one or more additional prompts based on the response to create a conversation with the user using an artificial intelligence engine. In response to the interviewee’s answer, the interactive video generation engine includes the response transcription in the prompt provided to the LLM, obtains one or more follow-up questions based on that answer, and provides the generated follow-up questions in the GUI ([0072]-[0073]). generate contextual information based on responses to the one or more additional prompts using the artificial intelligence engine. The interactive video generation engine generates LLM or AI feedback from the interviewee’s response transcript, audio, and video, providing response-derived contextual content ([0073]-[0074]). Regarding claims 2 and 16, Gustman teaches the apparatus of claim 1, and further teaches wherein the response is an audio response. The answer may be submitted verbally via a microphone ([0071]). the at least one processor is configured to cause the apparatus to transcribe the audio response prior to providing it to the artificial intelligence engine. The system transcribes the recording to text using automatic speech recognition ([0071]), after which the transcription is included in the prompt provided to the LLM ([0072]). Regarding claim 7, Gustman teaches the apparatus of claim 1, and further teaches wherein the contextual information comprises a story, a summary, an outline, a title, commentary, or a combination thereof that the artificial intelligence engine generates based on responses to the initial prompt and the one or more additional prompts. Gustman’s question timeline presents interview questions, and the interactive video generation engine records the interviewee’s answers as video segments ([0059]-[0065]); follow-up questions are generated from an answer and presented in the GUI ([0072]-[0073]). From the questions answered during the interview, the interactive video generation engine generates an index mapping recorded answer segments to the questions answered and uses that index to generate a table of contents for the interactive video ([0092]). The table of contents is an outline generated from the interview responses. Regarding claim 8, Gustman teaches the apparatus of claim 1, and further teaches wherein the at least one processor is configured to identify people, places, events, dates, or a combination thereof associated with the responses. Gustman discloses that when the interviewee states that they were a veteran in World War 2, the system uses World War 2 as the event that drives a related follow-up question; Gustman also applies named entity disambiguation to interview-answer transcriptions ([0058], [0079]). Identifying World War 2 in the response as the event used to generate the follow-up identifies an event associated with the response. Regarding claim 9, Gustman teaches the apparatus of claim 1, and further teaches wherein the at least one processor is configured to cause the apparatus to generate the one or more additional prompts based on a conversation theme and to guide the conversation toward the conversation theme. Gustman discloses clustering questions by topic and generating questions for a topic so as to ensure the interviewee has answered at least one question for every topic ([0085]), thereby generating the prompts based on a theme and guiding the interview toward it. Regarding claim 10, Gustman teaches the apparatus of claim 1, and further teaches wherein the at least one processor is configured to cause the apparatus to receive an image and generate one or more prompts related to the image using the artificial intelligence engine. Gustman discloses that the system may incorporate information from additional digital content such as images, videos, and other multimedia files to generate interview questions, where the digital content is tokenized into sections and provided as part of the LLM prompt so that questions are generated in view of that content ([0053]-[0054]). Regarding claim 13, Gustman teaches the apparatus of claim 1, and further teaches wherein the artificial intelligence engine comprises a generative artificial intelligence engine. Gustman discloses that the engine uses a large language model, which it describes as a type of generative artificial intelligence model that learns patterns and connections between words and, once trained, can be prompted to generate content such as questions ([0046]). Alternatively, claims 1, 7, 9, 13, 15, and 20 are rejected under 35 U.S.C. § 102(a)(1) as being anticipated by de Mazancourt et al. (US 11,095,579 B1; “de Mazancourt”). Regarding claim 1, de Mazancourt teaches an apparatus, comprising: at least one memory; and at least one processor coupled with the at least one memory and configured to cause the apparatus to: present an initial prompt to a user. The dialog manager selects an initial dialog module, and the NLP engine uses natural language generation to generate a natural-language expression for presentation to the user (col. 9, lines 8-15; fig. 5). receive a response to the prompt from the user. The user answers the chatbot question, and the response information is transmitted to the NLP engine for natural-language-understanding processing (col. 9, lines 15-20; fig. 5). generate one or more additional prompts based on the response to create a conversation with the user using an artificial intelligence engine. The NLP engine processes the user’s response, determines when information required for the current dialog module is missing, and uses natural language generation to generate a further request for the missing information during the ongoing chatbot interaction (col. 9, lines 30-50; fig. 5). generate contextual information based on responses to the one or more additional prompts using the artificial intelligence engine. After the user supplies the information requested by the follow-up, the NLP engine processes that response and the interaction summary is updated from the NLP-processed information (col. 9, lines 50-60; fig. 5). Regarding claim 7, de Mazancourt teaches the apparatus of claim 1, and further teaches wherein the contextual information comprises a story, a summary, an outline, a title, commentary, or a combination thereof that the artificial intelligence engine generates based on responses to the initial prompt and the one or more additional prompts. The interaction summary is updated based on NLP processing of successive user inputs, including first and second inputs, and the fig. 5 process updates the summary after later requested information is supplied (col. 2, lines 1-7; col. 9, lines 50-60). This limitation is recited in the alternative (“or”); because de Mazancourt teaches the “summary” alternative, the limitation is satisfied. Regarding claim 9, de Mazancourt teaches the apparatus of claim 1, and further teaches wherein the at least one processor is configured to cause the apparatus to generate the one or more additional prompts based on a conversation theme and to guide the conversation toward the conversation theme. Context knowledge concerns the particular project or transaction associated with the chatbot dialog; project-specific dialog modules obtain project, financing, and loan-parameter information; and rules transition among those modules based on information received from the user (col. 5, lines 20-35, 45-65; col. 6, lines 20-35; figs. 2 and 5). The project or transaction is the conversation theme toward which the follow-up questions are directed. Regarding claim 13, de Mazancourt teaches the apparatus of claim 1, and further teaches wherein the artificial intelligence engine comprises a generative artificial intelligence engine. The NLP engine may use one or more language models trained using machine-learning techniques to transform dialog-module output into natural chatbot dialog text and may perform natural-language generation using semantic objects (col. 6, lines 3-15). Regarding claims 15 and 20, de Mazancourt teaches a method, comprising: presenting an initial prompt to a user. The dialog manager selects an initial dialog module, and the NLP engine uses natural language generation to generate a natural-language expression for presentation to the user (col. 9, lines 8-15; fig. 5). receiving a response to the prompt from the user. The user answers the chatbot question, and the response information is transmitted to the NLP engine for natural-language-understanding processing (col. 9, lines 15-20; fig. 5). generating one or more additional prompts based on the response to create a conversation with the user using an artificial intelligence engine. The NLP engine processes the user’s response, determines when information required for the current dialog module is missing, and uses natural language generation to generate a further request for the missing information during the ongoing chatbot interaction (col. 9, lines 30-50; fig. 5). generating contextual information based on responses to the one or more additional prompts using the artificial intelligence engine. After the user supplies the information requested by the follow-up, the NLP engine processes that response and the interaction summary is updated from the NLP-processed information (col. 9, lines 50-60; fig. 5). Claim Rejections - 35 U.S.C. § 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 3 and 17 are rejected under 35 U.S.C. § 103 as being unpatentable over Gustman, as applied to claims 1 and 15 above, in further view of Junod (US 2022/0197931 A1; “Junod”). Regarding claims 3 and 17, Gustman teaches or suggests the apparatus of claim 1, but does not teach or suggest wherein the at least one processor is configured to cause the apparatus to present a plurality of predefined prompts and receive a selection of the initial prompt from the plurality of predefined prompts. Nonetheless, Junod teaches or suggests a question database storing questions for presentation to an interviewee; the interviewee reviews the questions and selects the questions the interviewee desires to answer; the interviewee may organize the selected questions in a preferred order of presentation; and, after the interviewee initiates the interview, the series of selected questions is presented to that interviewee, including an identified first question ([0054], [0057]-[0059]; figs. 1 and 10). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of Gustman so that, in its self-interview embodiment, the user is presented with stored interview questions and selects and orders the questions so that a selected question is presented first, as taught by Junod, because doing so allows the user to begin Gustman’s adaptive interview with a predefined subject the user wishes to address while preserving Gustman’s response-based generation of subsequent follow-up questions. Claims 4, 5, and 18 are rejected under 35 U.S.C. § 103 as being unpatentable over Gustman, as applied to claims 1 and 15 above, in further view of Gallagher (WO 2023/244986 A1; “Gallagher”). Regarding claim 4, Gustman teaches or suggests the apparatus of claim 1, but does not teach or suggest wherein the initial prompt is a user-generated prompt. Nonetheless, Gallagher teaches or suggests that the user may type in additional prompts, add questions, and order the prompts in a preferred sequence; when the user provides a desired order, the server uses that order; and, after the list has been revised by the user, the user device outputs the questions in the determined order one at a time in a self-interview mode, such that a user-added prompt placed first in the user-determined order is presented as the initial prompt ([0098]-[0102]). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of Gustman so that, in its self-interview embodiment, the user may add a question to the question script and place that question first, as taught by Gallagher, because doing so allows the interview to begin with subject matter the user specifically wishes to address while preserving Gustman’s response-based generation of subsequent follow-up questions. Regarding claims 5 and 18, the combination of Gustman and Gallagher teaches or suggests the apparatus of claim 4, and Gallagher further teaches or suggests wherein the user-generated prompt is based on information from the user or from a different user associated with the user, or a combination thereof. The user may provide information identifying subject matter for the user’s story, including by stating “I’m having a baby,” from which the platform identifies a pregnancy theme ([0095]), and may type additional prompts related to the identified theme ([0098]). This limitation is recited in the alternative (“or”); because the user-supplied-information alternative is taught, the limitation is satisfied. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to further modify the system of Gustman and Gallagher so that the user-added question placed first in the question sequence is related to the pregnancy theme identified from information supplied by the user, because doing so focuses the opening question on subject matter already identified by the user as relevant while preserving Gustman’s response-based generation of subsequent follow-up questions. Claims 6 and 19 are rejected under 35 U.S.C. § 103 as being unpatentable over Gustman, as applied to claims 1 and 15 above, in further view of Mitjans et al. (US 2023/0395204 A1; “Mitjans”). Regarding claims 6 and 19, Gustman teaches or suggests the apparatus of claim 1, but does not teach or suggest wherein the initial prompt is generated by the artificial intelligence engine based on previous responses. Nonetheless, the combination of Gustman and Mitjans teaches or suggests using Gustman’s generated questions to initialize the pre-interview question script ([0050]) and Mitjans’s machine-learning prediction model to determine or generate relevant questions based on previous interactions, including previous answers to previous questions ([0032]). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of Gustman so that the generated question used to initialize the pre-interview question script is generated based on previous answers, as taught by Mitjans, because doing so focuses the interview on relevant questions and reduces unnecessary questioning. The combination of Gustman and Mitjans further teaches or suggests wherein the initial prompt is generated by the artificial intelligence engine based on … contextual information associated with the user. Gustman’s generated questions may initialize the pre-interview question script ([0050]), while Mitjans’s machine-learning prediction model determines or generates relevant questions based on patient-specific features including personal preferences and social-post history, and its generative model generates questions from longitudinal patient records including health and historical patient data ([0032]-[0033]). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to further modify the system of Gustman and Mitjans so that the generated question used to initialize the pre-interview question script is also generated based on patient-specific features including personal preferences or historical patient data, because doing so further tailors the opening question to the individual and reduces irrelevant questioning. Claim 11 is rejected under 35 U.S.C. § 103 as being unpatentable over Gustman, as applied to claim 1 above, in further view of Kamat et al. (US 2018/0316636 A1; “Kamat”). Regarding claim 11, Gustman teaches or suggests the apparatus of claim 1, and Gustman further teaches or suggests wherein the at least one processor is configured to identify third-party contextual information from one or more third party sources. Additional digital content such as books, websites, and reference documents is tokenized and encoded, and relevant text is identified by a similarity search against the encoded response transcript ([0053]-[0054]). Gustman further teaches or suggests wherein the at least one processor is configured to use the third-party contextual information to generate the one or more additional prompts. The interactive video generation engine adds matched document text identified from reference documents to the question-generation prompt to produce in-depth follow-up questions ([0054]). However, Gustman does not teach or suggest wherein the at least one processor is configured to integrate the third-party contextual information into the contextual information. Nonetheless, Kamat teaches or suggests determining context for a query using the contents of the current chat conversation as one source and information from the Internet as another, and expressly combining recent chat messages with recent social media postings to determine that context ([0036]-[0038]). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of Gustman so that its response-derived contextual content is combined with matched document text to determine context, as taught by Kamat, because doing so supplies Gustman’s question-generation prompt with both interviewee-derived context and relevant reference-document information, allowing the engine to generate in-depth follow-up questions from a fuller interview context. Claim 12 is rejected under 35 U.S.C. § 103 as being unpatentable over Gustman, as applied to claim 1 above, in further view of Libow et al. (US 2019/0306268 A1; “Libow”). Regarding claim 12, Gustman teaches or suggests the apparatus of claim 1, but does not teach or suggest wherein the at least one processor is configured to extract contextual information from one or more other users associated with the user. Nonetheless, Libow teaches or suggests that one of the other users may send an IM to another one of the other users and the IM may be about the target user, and the extracting circuit extracts communicated information available about the target user from the electronic communications ([0049]-[0050]). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of Gustman so that its interview apparatus extracts communicated information available about the interviewee from electronic communications between other users that are about the interviewee, as taught by Libow, because doing so supplies additional interviewee information that may not have been provided directly by the interviewee and provides additional context for the adaptive interview. Claim 14 is rejected under 35 U.S.C. § 103 as being unpatentable over Gustman, as applied to claim 1 above, in further view of Menon et al. (US 11,855,934 B2; “Menon”). Regarding claim 14, Gustman teaches or suggests the apparatus of claim 1, but does not teach or suggest wherein the at least one processor is configured to continuously retrain the artificial intelligence engine based on the responses and contextual information. Nonetheless, Menon teaches or suggests that a reinforcement learning engine uses new contextual information to retrain ML/RL models for continuous improvement (col. 12, lines 30-37), and that the models are trained based on the user’s answers (col. 4, line 65; col. 5, line 1). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of Gustman so that Menon’s reinforcement learning engine continuously retrains the model used for Gustman’s question generation based on the user’s answers and new contextual information, because doing so incorporates interview-specific information into subsequent model updates and improves the relevance of later generated questions while preserving Gustman’s adaptive interview process. Alternatively, claims 2-4, 6, 8, 11, 16, 17, and 19 are rejected under 35 U.S.C. § 103 as being unpatentable over de Mazancourt, as applied to claims 1 and 15 above, in further view of Gustman. Regarding claims 2 and 16, de Mazancourt teaches or suggests the apparatus of claim 1, and de Mazancourt further teaches or suggests wherein the response is an audio response. The chatbot interface may interact with the user using voice, text, or both, and information obtained from the user is transmitted for processing by the NLP engine (col. 7, lines 1-9). However, de Mazancourt does not teach or suggest the at least one processor is configured to cause the apparatus to transcribe the audio response prior to providing it to the artificial intelligence engine. Nonetheless, Gustman teaches or suggests that the interactive video generation engine transcribes a recorded verbal answer to text using automatic speech recognition and includes the transcription in the prompt provided to the LLM ([0071]-[0072]). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of de Mazancourt so that the chatbot interface transcribes the user’s voice response to text using automatic speech recognition before the text is processed by the NLP engine, as taught by Gustman, because doing so enables de Mazancourt’s NLP engine to process voice input in textual form while preserving the existing voice interaction. Regarding claims 3 and 17, de Mazancourt teaches or suggests the apparatus of claim 1, but does not teach or suggest wherein the at least one processor is configured to cause the apparatus to present a plurality of predefined prompts and receive a selection of the initial prompt from the plurality of predefined prompts. Nonetheless, the combination of de Mazancourt and Gustman teaches or suggests the limitation by using, as de Mazancourt’s initial dialog question, a question selected by the self-interviewing user from Gustman’s library of preexisting questions and topics; de Mazancourt supplies the initial dialog-question position, while Gustman permits the same user to select script questions from the library and use the question timeline to choose which question to ask next (de Mazancourt, col. 9, lines 8-15; Gustman, [0042], [0044], and [0059]-[0061]). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of de Mazancourt so that its chatbot interface presents a library of preexisting questions and topics from which the user selects the initial question, as taught by Gustman, because doing so lets the user begin the project or transaction dialog at a subject the user wants to address and avoids unnecessary preliminary dialog. Regarding claim 4, de Mazancourt teaches or suggests the apparatus of claim 1, but does not teach or suggest wherein the initial prompt is a user-generated prompt. Nonetheless, the combination of de Mazancourt and Gustman teaches or suggests the limitation by using, as de Mazancourt’s initial dialog question, a question specified by the self-interviewing user; de Mazancourt supplies the initial dialog-question position, while Gustman permits the interviewer and interviewee to be the same person, permits the interviewer to specify or add their own questions before the interview, and lets the interviewer choose which question to ask next (de Mazancourt, col. 9, lines 8-15; Gustman, [0042], [0044], [0056], and [0059]-[0061]). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of de Mazancourt so that the chatbot interaction uses, as the initial prompt, a question specified by the user, as taught by Gustman, because doing so allows the project or transaction dialog to start with subject matter the user knows is relevant rather than only a system-selected opening. Regarding claims 6 and 19, de Mazancourt teaches or suggests the apparatus of claim 1, but does not teach or suggest wherein the initial prompt is generated by the artificial intelligence engine based on previous responses. Nonetheless, the combination of de Mazancourt and Gustman teaches or suggests that de Mazancourt’s NLP engine updates the interaction summary from NLP-processed user response information after requested information is supplied (de Mazancourt, col. 9, lines 50-60), while Gustman’s LLM question-generation process may use the transcript of the most recent question response and related question transcripts to guide question generation, and generated questions may initialize the question script prior to the interview (Gustman, [0050]). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of de Mazancourt so that information from its interaction summary, updated from user response information, is included in Gustman’s LLM prompt to generate a question that initializes the question script prior to the interview, because doing so allows a later project or transaction interaction to begin from information already supplied by the user and avoids repeating previously covered subject matter. The combination of de Mazancourt and Gustman further teaches or suggests wherein the initial prompt is generated by the artificial intelligence engine based on … contextual information associated with the user. de Mazancourt’s interaction summary is updated from NLP-processed user response information (de Mazancourt, col. 9, lines 50-60), and Gustman’s LLM prompt is generated based on contextual information including the interviewee’s name, biography, and topics of interest, with generated questions used to initialize the question script prior to the interview (Gustman, [0047], [0050]). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to further modify the system of de Mazancourt and Gustman so that Gustman’s LLM prompt is generated based on de Mazancourt’s interaction summary as contextual information and the generated question initializes the question script prior to the interview, because doing so aligns the opening question with information already available for that user and reduces irrelevant questioning. Regarding claim 8, de Mazancourt teaches or suggests the apparatus of claim 1, but does not teach or suggest wherein the at least one processor is configured to identify people, places, events, dates, or a combination thereof associated with the responses. Nonetheless, Gustman teaches or suggests applying named entity disambiguation to interviewee-answer transcriptions and gives an example in which an interviewee states that they were a veteran in World War 2 and the transcript is used to generate a related follow-up question ([0058], [0079]). Identifying World War 2 in the response as the event that drives the follow-up identifies an event associated with the response. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of de Mazancourt so that its NLP engine applies Gustman’s named entity disambiguation to user response information to identify an event associated with the response, as taught by Gustman, because doing so improves de Mazancourt’s interpretation of the response and supplies event-specific context for targeted follow-up requests and the interaction summary. Regarding claim 11, de Mazancourt teaches or suggests the apparatus of claim 1, but does not teach or suggest wherein the at least one processor is configured to identify third-party contextual information from one or more third party sources. Nonetheless, Gustman teaches or suggests incorporating additional digital content such as text documents and websites and identifying relevant text in that digital content by a similarity search between the encoded response transcript and encoded digital-content sections ([0053]-[0054]). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of de Mazancourt so that its chatbot searches additional digital content for relevant text using Gustman’s similarity search between the user response and digital-content sections, as taught by Gustman, because doing so supplements de Mazancourt’s user and application information with external information relevant to the project or transaction. Gustman further teaches or suggests wherein the at least one processor is configured to use the third-party contextual information to generate the one or more additional prompts. The interactive video generation engine adds matched document text identified from reference documents to the question-generation prompt to produce in-depth follow-up questions ([0054]). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to further modify the system of de Mazancourt and Gustman so that matched document text is added to the prompt used to generate follow-up questions, because doing so produces more informed follow-up requests when de Mazancourt’s user response and stored transaction context do not contain all relevant information. The combination of de Mazancourt and Gustman further teaches or suggests wherein the at least one processor is configured to integrate the third-party contextual information into the contextual information. de Mazancourt’s NLP engine updates the interaction summary from NLP-processed user response information (de Mazancourt, col. 9, lines 50-60), while Gustman’s interactive video engagement engine applies retrieval augmented generation techniques to documents associated with the interactive video, including news articles, biographies, and websites, and provides a text-based summary determined from those documents (Gustman, [0105]). Including Gustman’s text-based summary in de Mazancourt’s interaction summary integrates the third-party information into the contextual information. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to further modify the system of de Mazancourt and Gustman so that de Mazancourt’s interaction summary, which is updated from the user’s responses, includes Gustman’s text-based summary determined from documents associated with the interactive video, because doing so produces a more complete interaction summary containing both information supplied by the user and information available from those documents. Alternatively, claims 5 and 18 are rejected under 35 U.S.C. § 103 as being unpatentable over de Mazancourt and Gustman, as applied to claims 4 and 15 above, in further view of Larson et al. (WO 2020/236651 A1; “Larson”). Regarding claims 5 and 18, the combination of de Mazancourt and Gustman teaches or suggests the apparatus of claim 4, but does not teach or suggest wherein the user-generated prompt is based on information from the user or from a different user associated with the user, or a combination thereof. Nonetheless, the combination of Gustman and Larson teaches or suggests the limitation: Gustman’s self-interview embodiment permits the same user to specify a question used to initiate the dialog, while Larson uses biographic information supplied by the enrollee through the client application to generate interview questions (Gustman, [0042], [0044], [0056], and [0059]-[0061]; Larson, [17], [89], and [94]). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of de Mazancourt and Gustman so that, in Gustman’s self-interview embodiment, the question specified by the user and used to initiate the dialog is based on biographic information supplied by the user, as taught by Larson, because doing so focuses the opening question on information already supplied by the user and avoids irrelevant introductory questioning. Alternatively, claim 10 is rejected under 35 U.S.C. § 103 as being unpatentable over de Mazancourt, as applied to claim 1 above, in further view of Jones et al. (US 12,067,657 B2; “Jones”). Regarding claim 10, de Mazancourt teaches or suggests the apparatus of claim 1, but does not teach or suggest wherein the at least one processor is configured to cause the apparatus to receive an image and generate one or more prompts related to the image using the artificial intelligence engine. Nonetheless, Jones teaches or suggests that, after a particular image has been added to the system, machine learning models identify features within the image and generate question prompts, and that a user can upload one or more images and media files (col. 13, lines 4-17, 35-38). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of de Mazancourt so that its chatbot receives an image uploaded by the user and applies Jones’s machine learning models to identify features within the image and generate question prompts, as taught by Jones, because doing so provides de Mazancourt’s project or transaction chatbot with image-derived information from which to request specific details that may not be captured in text alone. Alternatively, claim 12 is rejected under 35 U.S.C. § 103 as being unpatentable over de Mazancourt, as applied to claim 1 above, in further view of Gilmour (US 6,405,197 B2; “Gilmour”). Regarding claim 12, de Mazancourt teaches or suggests the apparatus of claim 1, but does not teach or suggest wherein the at least one processor is configured to extract contextual information from one or more other users associated with the user. Nonetheless, Gilmour teaches or suggests that profile terms are extracted from an electronic document submitted by a third party, the source of a term may be recorded as another user, and a first member of an organization may submit documents that provide profile information regarding a second member who is an interactive user (col. 29, lines 10-31, 46-53). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of de Mazancourt so that its NLP engine extracts profile information regarding a second member from documents submitted by a first member of an organization, as taught by Gilmour, because doing so supplements de Mazancourt’s interaction summary with information supplied by another user associated with the user and provides additional information for later chatbot requests. Alternatively, claim 14 is rejected under 35 U.S.C. § 103 as being unpatentable over de Mazancourt, as applied to claim 1 above, in further view of Menon. Regarding claim 14, de Mazancourt teaches or suggests the apparatus of claim 1, but does not teach or suggest wherein the at least one processor is configured to continuously retrain the artificial intelligence engine based on the responses and contextual information. Nonetheless, Menon teaches or suggests that a reinforcement learning engine uses new contextual information to retrain ML/RL models for continuous improvement (col. 12, lines 30-37), and that the models are trained based on the user’s answers (col. 4, line 65; col. 5, line 1). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the system of de Mazancourt so that Menon’s reinforcement learning engine continuously retrains the machine-learning-trained language models used by de Mazancourt’s NLP engine based on user answers and new contextual information, as taught by Menon, because doing so incorporates transaction-specific interaction information into subsequent model updates and improves later generated chatbot text and follow-up requests. Conclusion Any inquiry concerning this communication or earlier communications from the examiner should be directed to Andrew Georgandellis whose telephone number is 571-270-3991. The examiner can normally be reached on Monday through Friday, 7:30-5:00 PM EST. If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Tonia Dollinger, can be reached on 571-272-4170. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300. Information regarding application status may be obtained from Patent Center; unpublished application information is available there to authorized users. Questions about access to the USPTO patent electronic filing system may be directed to the Electronic Business Center (EBC) at 866-217-9197 (toll-free). /ANDREW C GEORGANDELLIS/Primary Examiner, Art Unit 2459
Read full office action

Prosecution Timeline

Apr 10, 2025
Application Filed
Sep 15, 2026
Non-Final Rejection mailed — §102, §103 (current)

Precedent Cases

Applications granted by this same examiner with similar technology

Patent 12786130
SYSTEMS AND METHODS FOR CLASSIFYING TRAFFIC IN A HIERARCHICAL SD-WAN NETWORK
4y 3m to grant Granted Nov 03, 2026
Patent 12786157
Domain Name Query Method and Related Device
4y 5m to grant Granted Nov 03, 2026
Patent 12771761
SYSTEMS AND METHODS FOR MODIFYING SESSIONS IN ACCORDANCE WITH A USER PLANE FUNCTION SELECTION BASED ON LATENCY
3y 0m to grant Granted Oct 20, 2026
Patent 12756639
AUTOMATED SERVICE LEVEL OBJECTIVES FOR BATCH DATA
3y 0m to grant Granted Oct 06, 2026
Patent 12756757
FEEDBACK INDICATION METHOD AND APPARATUS
2y 9m to grant Granted Oct 06, 2026
Study what changed to get past this examiner. Based on 5 most recent grants.

Strategy Recommendation AI-generated — please review before filing

Get a prosecution strategy drawn from examiner precedents, rejection analysis, and claim mapping.
Typically takes 5-10 seconds — AI-generated, attorney review required before filing

Prosecution Projections

1-2
Expected OA Rounds
56%
Grant Probability
97%
With Interview (+40.2%)
4y 0m (~2y 5m remaining)
Median Time to Grant
Low
PTA Risk
Based on 499 resolved cases by this examiner. Grant probability derived from career allowance rate.

Sign in with your work email

Enter your email to receive a magic link. No password needed.

Personal email addresses (Gmail, Yahoo, etc.) are not accepted.

Free tier: 3 strategy analyses per month