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 .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 05/11/2026 has been entered.
Response to Amendment
Claims 1, 8, and 10 are currently amended.
Claims 1-18 are currently pending and addressed below.
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 1-18 are rejected under 35 U.S.C. § 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1:
Claims 1-18 is/are directed towards a statutory category (i.e., a process, machine, manufacture, or composition of matter) (Step 1, Yes).
Step 2A Prong One:
Claim 1 recites (additional elements underlined):
A system for developing, using artificial intelligence, optimized advertisements comprising:
at least one non-transitory memory; and
at least one processing device, the memory containing software code configured to cause the processing device to:
gather data from a plurality of data sources;
extract, using a machine learning algorithm, at least one of a plurality of customer behavior features or a plurality of financial institution behavior features based on the gathered data;
process the customer behavior features and financial institution behavior features using one or more trained foundation models, wherein the trained foundation models are selected based on a plurality of foundation model selection variables, and wherein the trained foundation models output one or more foundation model outputs related to an advertisement instruction;
input the foundation model outputs and a plurality of goal inputs into a trained campaign execution model, wherein:
the goal inputs comprise one or more of a plurality of financial institution products, a plurality of financial institution product variants, a plurality of financial institution parameters, a plurality of financial institution regions, a plurality of financial institution growth strategies, a product response, a market response, a plurality of a campaign duration, a campaign budget, or an available campaign channel;
the trained campaign execution model is trained based on the foundation model outputs and the goal inputs, and wherein the trained campaign execution model is configured with feedback loops that refine the trained campaign execution model at different levels based on key performance indicators monitored at predetermined times and based on labeled performance data collected from executed campaigns based on customer behavior;
output, from the trained campaign execution model, a natural-language advertisement response, wherein the advertisement response is based on the goal inputs.
Under the broadest reasonable interpretation, the limitations outlined above that describe or set forth the abstract idea, cover performance of the limitations in the mind but for the recitation of generic computer(s) and/or generic computer component(s). That is, other than reciting the additional elements identified below, nothing in the claim precludes the limitations from practically being performed in the mind. These limitations are considered a mental process because the limitations include an observation, evaluation, judgment, and/or opinion. These limitations are also similar to “collecting information, analyzing it, and displaying certain results of the collection and analysis” and/or “collecting and comparing known information” which were determined to be mental processes in MPEP 2106.04(a)(2)(III)(A). The Examiner notes that “[c]laims can recite a mental process even if they are claimed as being performed on a computer” (see MPEP 2106.04(a)(2)(III)(C)). The mere nominal recitation of the additional elements identified above do not take the claims out of the mental process grouping. Therefore, the claim recite a mental process (Step 2A Prong One, Yes).
The limitations outlined above also describe or set forth an advertising/marketing activity. Advertising/marketing fall within the certain method of organizing human activity enumerated grouping of abstract ideas. The limitations outlined above also describe or set forth a fundamental economic principle or practice because advertising/marketing is related to commerce and economy. The limitations outlined above also describe or set forth a commercial interaction (e.g., advertising, marketing or sales activities or behaviors, business relations) and managing personal behavior or relationships or interactions between people (e.g., following rules or instructions). Therefore, the claim recites a certain method of organizing human activity (Step 2A Prong One, Yes).
Step 2A Prong Two:
In Step 2A Prong Two, the additional element(s) outlined above are recited at a high level of generality, and under the broadest reasonable interpretation, are generic computer(s) and/or generic computer component(s) that perform generic computer functions. The additional element(s) are merely used as tools, in their ordinary capacity, to perform the abstract idea. The additional element(s) amount adding the words “apply it” with the judicial exception. Merely implementing an abstract idea on generic computer(s) and/or generic computer component(s) does not integrate the judicial exception similar to how the recitation of the computer in the claim in Alice amounted to mere instructions to apply the abstract idea of intermediated settlement on a generic computer. The additional elements also amount to generally linking the use of the abstract idea to a particular technological environment or field of use (e.g., in a computer environment). The courts have found that simply limiting the use of the abstract idea to a particular environment does not integrate the judicial exception into a practical application. Viewing the limitations as an ordered combination does not add anything further than looking at the limitations individually. There is no indication that the combination of elements improves the functioning of a computer, improves any other technology or technical field, applies or uses the judicial exception to effect a particular treatment or prophylaxis for disease or medical condition, applies the judicial exception with, or by use of a particular machine, effects a transformation or reduction of a particular article to a different state or thing, or applies or uses the judicial exception in some other meaningful way beyond generally linking the use of the judicial exception to a particular technological environment, such that the claims as a whole is more than a drafting effort designed to monopolize the exception. Their collective functions merely provide generic computer implementation (Step 2A Prong Two, No).
Step 2B:
In Step 2B, the additional elements also do not amount to significantly more for the same reasons set forth with respect to Step 2A Prong Two. The Examiner notes that revised Step 2A Prong Two overlaps with Step 2B, and thus, many of the considerations need not be reevaluated in Step 2B because the answer will be the same. Viewing the limitations as an ordered combination does not add anything further than looking at the limitations individually. Their collective functions merely provide generic computer implementation (Step 2B, No).
Claims 2-9 recite further limitations that also fall within the same abstract ideas identified above with respect to claim 1 (i.e., certain methods of organizing human activities and/or mental processes).
Claim 2 recites the additional elements of “foundation”, “the processing device is further configured to”, and “machine learning”. Claim 3 recites the additional elements of “trained”, “one or more of a logistic regression, a random forest, a gradient boosting, a clustering algorithm, or a deep learning model”, “the processing device is further configured to”, “trained”, and “train the trained”. The Examiner notes that “one or more of logistic regression, a random forest, a gradient boosting, and a clustering algorithm” can also be considered to be mental processes and mathematical concepts. Claim 4 recites the additional elements of “wherein the processing device is further configured to”. Claim 5 recites the additional elements of “trained” and “hybrid recommendation filtering”. Claim 6 recites the additional elements of “wherein the processing device is further configured to” and “trained”. Claim 7 recites the additional elements of “the processing device is further configured to”, “trained”, and “train the trained”. Claim 8 recites the additional elements “wherein the system further comprises a user interface configured to”, “provide the user interface to a user device”, “device on one or more elements of the user interface”, “machine learning”, “trained”, and “foundation”. However, these additional elements also do not integrate the judicial exception into a practical application or amount to significantly more because they amount to adding the words “apply it” with the judicial exception, mere instructions to implement the idea on a computer, merely using a computer as a tool to perform an abstract idea, and generally linking the use of the judicial exception to a particular technological environment or field of use.
Claims 9 does not recite any other additional elements. Therefore, for the same reasons explained above with respect to claim 1, claim 9 also does not integrate the judicial exception into a practical application or amount to significantly more.
Claim 10 recites substantially similar limitations as claim 1. Therefore, for the same reasons explained above with respect to claim 1, claim 10 also recites an abstract idea in Step 2A Prong One (i.e., certain method of organizing human activities, and mental processes). Claim 10 recites the additional elements of “using artificial intelligence”, “using a machine learning algorithm”, “trained”, “foundation”, and “natural-language”. However, for the same reasons explained above with respect to claim 1, these additional elements also do not integrate the judicial exception into a practical application or amount to significantly more.
Claims 11-18 recite further limitations that also fall within the same abstract ideas identified above with respect to claim 10 (i.e., certain methods of organizing human activities and/or mental processes).
Claim 11 recites the additional elements of “foundation”, “the processing device is further configured to”, and “machine learning”. Claim 12 recites the additional elements of “trained”, “one or more of a logistic regression, a random forest, a gradient boosting, a clustering algorithm, or a deep learning model”, “the processing device is further configured to”, “trained”, and “train the trained”. The Examiner notes that “one or more of logistic regression, a random forest, a gradient boosting, and a clustering algorithm” can also be considered to be mental processes and mathematical concepts. Claim 14 recites the additional elements of “trained” and “hybrid recommendation filtering”. Claim 15 recites the additional element of “trained”. Claim 16 recites the additional elements of “trained”, and “train the trained”. Claim 17 recites the additional elements “provide a user interface to a user device”, “device on one or more elements of the user interface”, “machine learning”, “trained”, and “foundation”. However, these additional elements also do not integrate the judicial exception into a practical application or amount to significantly more because they amount to adding the words “apply it” with the judicial exception, mere instructions to implement the idea on a computer, merely using a computer as a tool to perform an abstract idea, and generally linking the use of the judicial exception to a particular technological environment or field of use.
Claims 13 and 18 do not recite any other additional elements. Therefore, for the same reasons explained above with respect to claim 1, claims 13 and 18 also does not integrate the judicial exception into a practical application or amount to significantly more.
Prior Art
The Examiner notes that after conducting a thorough and complete search of the prior art, the claims are found to recite novel and non-obvious subject matter. The closest prior art found to date are the following:
Domeniconi (US 2025/0061351 A1) discloses methods, systems, and computer program products for training a transactional event data analysis model and fine-tuning a pre-trained model to predict future transactions. Domeniconi also discloses gathering and extracting data, backpropagation, transfer learning, customer behavior features, financial institution behavior features, large language models, and the concept of the outputs of the foundation model being input into a final machine learning model (¶¶ 151-152). However, the final machine learning model of Domeniconi does not appear to provide sufficient detail regarding the final output. Domeniconi does not appear to input the plurality of goal inputs into a trained campaign execution model and outputting, from the trained campaign execution model, a natural language advertisement response, wherein the advertisement response is based on the goal inputs as claimed.
Jonnalagadda et al. (US 2020/0143247 A1) disclose systems and methods for improved automated conversations with intent and action response generation. Jonnalagadda et al. also discloses the use of multiple machine learning models, responses, goals, budgets, financial products, training, feedback, and customer data. However, the models in Jonnalagadda et al. operate in parallel. Jonnalagadda et al. does not appear to input the foundation model outputs and a plurality of goal inputs into a trained campaign execution model in order to output, from the trained campaign execution model, a natural language advertisement response as claimed.
Bhatt et al. (US 2022/0155926 A1) discloses the use of a machine learning model that is pre-trained based on historical input data and historical output data. However, Bhatt et al. does not use the foundation model and trained campaign execution model as claimed.
Abdelrahman et al. (US 2025/0037107 A1) discloses systems and methods for contextual transaction data collection using large language processing. Abdelrahman et al. also discloses gathering, extracting, and processing of customer behavior features and financial institution behavior features using one or more trained foundation models. However, Abdelrahman et al. does not appear to explicitly input the foundation model output and plurality of goal inputs into a trained campaign execution model in order to output a natural language advertisement response from the trained campaign execution model as claimed.
Eidelman et al. (US 2022/0191155 A1) discloses the concept of machine learning models to create messages for advocacy campaigns. Eidelman et al. also disclose the use of logistic regression, pre-trained models, advocacy responses, demographic data, surveys, and key performance indicators. However, Eidelman et al. does not appear to use the foundation model and the trained campaign execution model as claimed.
Response to Arguments
Applicant's arguments filed 05/11/2026 have been fully considered but they are not persuasive.
Argument: “The amended claims recite limitations that cannot practically be performed in the human mind. Amended claim 1, for example, recites that "the trained campaign execution model is configured with feedback loops that refine the trained campaign execution model at different levels based on a plurality of key performance indicators monitored at predetermined times and based on labeled performance data collected from executed campaigns based on customer behavior." Amended claim 1 further recites "extract, using a machine learning algorithm, at least one of a plurality of customer behavior features or a plurality of financial institution behavior features based on the gathered data" and "process the customer behavior features and financial institution behavior features using one or more trained foundation models, wherein the trained foundation models are selected based on a plurality of foundation model selection variables." At least these claim elements are not processes that can practically be performed in the human mind. A human mind cannot configure a trained machine learning model with feedback loops that operate at different levels based on monitored key performance indicators. A human mind cannot collect labeled advertisement performance data from executed advertisement campaigns and use that labeled deployment data to refine a trained machine learning model. A human mind cannot select among trained foundation models based on a plurality of foundation model selection variables, nor can it process behavior features through trained foundation models to produce foundation model outputs. These are inherently computational operations that require trained machine learning model infrastructure, feedback loops operating at different levels, labeled data collected from executed advertisement campaigns, and iterative refinement beyond human cognitive capability. See USPTO Memo at 2 ("Claim limitations that encompass Al in a way that cannot be practically performed in the human mind do not fall within this grouping.").”
In response, the Examiner respectfully disagrees. As explained above, the limitations that describe or set forth the abstract idea in Step 2A Prong One, which excludes the additional elements, can be practically performed in the human mind or by a human using pen and paper. These limitations are considered a mental process because the limitations include an observation, evaluation, judgment, and/or opinion. These limitations are also similar to “collecting information, analyzing it, and displaying certain results of the collection and analysis” and/or “collecting and comparing known information” which were determined to be mental processes in MPEP 2106.04(a)(2)(III)(A). The Examiner notes that “[c]laims can recite a mental process even if they are claimed as being performed on a computer” (see MPEP 2106.04(a)(2)(III)(C)). The mere nominal recitation of the additional elements identified above do not take the claims out of the mental process grouping. Therefore, the claim recite a mental process (Step 2A Prong One, Yes).
Argument: “Further, when properly construed in view of the specification, the amended claims do not recite a method of organizing human activity or a fundamental economic principle or practice. The amended claims recite a computer-implemented artificial intelligence system architecture comprising trained foundation models selected based on a plurality of foundation model selection variables, a separately trained campaign execution model trained based on the foundation model outputs and the goal inputs, and feedback loops that refine the trained campaign execution model at different levels based on key performance indicators and labeled advertisement performance data from executed advertisement campaigns. Specification at paragraphs [0163], [0175]-[0177]. These operations are technical steps performed automatically by the claimed processing device over large datasets, cannot be performed mentally or with pen and paper, and do not reduce to the kind of commercial interaction or interpersonal relationship-management that defines the organizing-human-activity grouping. Humans do not configure trained machine learning models with feedback loops, select among foundation models based on computational selection variables, or refine machine learning models based on labeled deployment data collected from executed campaigns. The amended claims define how multiple trained machine learning models interact within a computing system to generate adaptive, data-driven advertisement responses, not the abstract idea of advertising.”
In response, the Examiner respectfully disagrees. As explained above, the limitations outlined above that describe or set forth the abstract idea in Step 2A Prong One, which excludes the additional elements, also describe or set forth an advertising/marketing activity at least because the claims explicitly use the word “advertisement”. Advertising/marketing fall within the certain method of organizing human activity enumerated grouping of abstract ideas. The limitations outlined above also describe or set forth a fundamental economic principle or practice because advertising/marketing is related to commerce and economy. The limitations outlined above also describe or set forth a commercial interaction (e.g., advertising, marketing or sales activities or behaviors, business relations) and managing personal behavior or relationships or interactions between people (e.g., following rules or instructions). Therefore, the claim recites a certain method of organizing human activity (Step 2A Prong One, Yes).
Argument: “Notwithstanding the resolution of Prong One of the 2019 PEG, Applicant submits that the claims are also patent eligible at Prong Two because "the claim[s] as a whole integrat[e] the recited judicial exception into a practical application of that exception." 2019 PEG at 54.
The amended claims solve a technical problem identified in the specification. The specification explains that existing computer systems for recommending products and services and for developing advertisements "fail to accurately capture all data, and fail to utilize machine learning techniques to forecast customer decisions," that "current systems only use descriptive analytics instead of predictive or prescriptive recommendations," and that traditional segmentation and product targeting approaches are "flawed, as [they] lack[] granularity, nuance, and customization." Specification at paragraphs [0003]-[0006]. The amended claims address these technical deficiencies by reciting a specific computer-implemented artificial intelligence architecture comprising trained foundation models, foundation model selection variables, a separately trained campaign execution model, and feedback loops configured to refine the campaign execution model at different levels based on key performance indicators monitored at predetermined times and based on labeled advertisement performance data collected from executed advertisement campaigns. The specification expressly describes the technical improvements provided by the amended claim language. The specification explains that "feedback loops may fine- tune campaign execution model 812 at different levels (e.g., the household level)" and that "the system 800 may include one or more interconnected feedback loops" that may be "parallel feedback loops or nested feedback loops." Specification at paragraph [0175]. The specification further describes the integration of labeled data collected from executed advertisement campaigns into the feedback architecture: "[a]s the campaign is executed, labeled market data may be collected (indicating e.g., if the campaign is successful, if an entity should attribute new deposits, loans, client acquisition)," and "[c]ollected data may be fed back into one or more of the campaign execution model 812 ... [where] the underlying output models ... may be refined or retrained based on market and client behavior." Specification at paragraph [0176]. And the specification specifies that "KPI feedback 816 may be inputted into campaign execution model 812 in real-time or at predetermined times" and that "KPI feedback 816 is used to optimize response 814 including advertising creatives, messages, channels, and efficiency." Specification at paragraph [0177]. The specification thus discloses, in detail, how the claimed multi-model architecture and feedback loops technically improve the operation of the artificial intelligence system itself: by enabling continual, multi-level refinement of a trained machine learning model based on labeled data collected from real-world campaign deployment, the claimed architecture maintains model accuracy and responsiveness as customer-engagement data and advertising performance evolve over time.”
In response, the Examiner respectfully disagrees. First, the alleged technical problems of failing to accurately capture all data to forecast customer decisions, only using descriptive analytics instead of predictive or prescriptive recommendations, and traditional segmentation and product targeting approaches being flawed, are business problems, not technical problems. Unlike in DDR in which the claimed invention solved the business challenge of retaining website visitors that is particular to the Internet, here the claimed invention amounts to merely reciting the performance of a business practice along with the requirement to perform it on the Internet. Similar to SAP America, Inc. v. InvestPic, LLC (Fed. Cir. 2018), the advance here lies entirely in the realm of the abstract idea, with no plausible alleged innovation int eh non-abstract application realm.
Argument: “The Patent Trial and Appeal Board's precedential decision in Ex parte
Desjardins, Appeal No. 2024-000567 (PTAB Sept. 26, 2025) supports patent eligibility of the amended claims under Step 2A, Prong Two. In Desjardins, the Appeal Review Panel determined that claims directed to an improved way of training a machine learning model integrate a mathematical concept into a practical application by improving the functioning of the Al technology itself. The Panel credited specification disclosure of technical improvements including reduced storage, reduced system complexity, and preservation of performance attributes during subsequent computational tasks, particularly addressing the problem of "catastrophic forgetting" in sequential learning. The instant case fits the same pattern. Just as the Desjardins specification identified improvements to how the machine learning model itself operates, the present specification identifies improvements to how the trained campaign execution model operates: specifically, refinement of the model at different levels through feedback loops at different levels and supervised retraining of the model using labeled data collected from executed campaigns. See Specification at paragraphs [0175]- [0176]. These are improvements to the technical operation of the artificial intelligence system, not merely to the abstract idea of advertising. Accordingly, the amended claims integrate any alleged abstract idea into a practical application, consistent with the precedential guidance in Desjardins.”
In Ex Parte Desjardins, the Appeal Review Panel (APR) credited benefits including reduced storage, reduced system complexity and streamlining, and preservation of performance attributes associated with earlier tasks during subsequent computational task as technological improvements that were disclosed in the patent application specification. The APR then determined that the specification identified improvements as to how the machine learning model itself operates, including training a machine learning model to learn new tasks while protecting knowledge about previous tasks to overcome the problem of "catastrophic forgetting". Here, the claims and specification are completely silent with regard to such technical improvements. The models here are recited at a high level of generality, and are merely used as tools, in their ordinary capacity, to perform the abstract idea. “Use of a computer or other machinery in its ordinary capacity for economic or other task (e.g., to receive, store, or transmit data) or simply adding a general purpose computer or computer components after the fact to an abstract idea (e.g., a fundamental economic practice or mathematical equation) does not integrate a judicial exception into a practical application or provide significantly more” (MPEP 2106.05(f)(2)).
Additionally, unlike in Enfish in which the claimed invention achieved other benefits over conventional databases such as increased flexibility, faster search times, and smaller memory requirements that provided improvements to the functioning of the computer itself, here looking at the limitations as an ordered combination adds nothing that is not already present when looking at the elements taken individually. There is no indication that the combination of elements improves the functioning of a computer or improve any other technology. Their collective functions merely provide generic computer implementation.
Argument: “The Office's assertion that the additional elements are "recited at a high level of generality" and merely "used as tools, in their ordinary capacity, to perform the abstract idea" overlooks the specific technical configuration recited in the amended claims and described in the specification. Office Action at 4. The amended claim language is not generic. It recites specific, technologically-rooted operations including "process the customer behavior features and financial institution behavior features using one or more trained foundation models, wherein the trained foundation models are selected based on a plurality of foundation model selection variables," and "the trained campaign execution model is configured with feedback loops that refine the trained campaign execution model at different levels based on a plurality of key performance indicators monitored at predetermined times and based on labeled performance data collected from executed campaigns based on customer behavior." These elements are specifically configured to improve the accuracy and responsiveness of the artificial intelligence system, not to apply generic computer components to an abstract idea. The Office's reliance on Recentive Analytics, Inc. v. Fox Corp. (Office Action at 17) is also misplaced as applied to the amended claims. In Recentive Analytics, the Federal Circuit held that "patents that do not more than claim the application of generic machine learning to new data environments, without disclosing improvements to the machine learning models to be applied, are patent ineligible under § 101." Recentive Analytics, Inc. v. Fox Corp., 134 F.4th 1205, 1216 (Fed. Cir. 2025). By its own terms, the Recentive Analytics holding addresses only claims that apply generic machine learning without disclosing improvements to the machine learning models. Claims that do disclose such improvements are not within the scope of the holding. The amended claims, as discussed below, do disclose such improvements. The amended claims do not merely apply generic machine learning to a new data environment. The amended claims recite specific architectural improvements to the trained machine learning model itself. Amended claim 1 recites that "the trained campaign execution model is configured with feedback loops that refine the trained campaign execution model at different levels based on a plurality of key performance indicators monitored at predetermined times and based on labeled performance data collected from executed campaigns based on customer behavior." These architectural properties are disclosed in the specification: feedback loops "at different levels" including "interconnected feedback loops" and "parallel feedback loops or nested feedback loops"; collection of labeled data from executed campaigns that "may be fed back into one or more of the campaign execution model" where "the underlying output models ... may be refined or retrained"; and KPI feedback inputted into the campaign execution model "in real-time or at predetermined times." See Specification at paragraphs [0175]-[0177]. The Federal Circuit's concern in Recentive Analytics was with claims that recite no more than generic iterative training or dynamic adjustment based on real-time data, properties the Court held are "incident to the very nature of machine learning." Recentive Analytics, Inc. v. Fox Corp., p. 12 (Fed. Cir. 2025). The amended claims recite materially more. The multi-level feedback architecture, supervised learning from labeled data collected from executed advertisement campaigns, and KPI-driven refinement at predetermined times are not properties incident to the nature of machine learning generally; they are specific architectural choices disclosed in the specification and recited in the amended claims. Accordingly, Recentive Analytics does not support the rejection of the amended claims.”
In response, the Examiner respectfully disagrees. Refining the campaign execution model does not provide a technological improvement to machine learning. The Federal Circuit in Recentive Analytics Inc. v. Fox Corp. stated on page 12 that “[t]he requirement that the machine learning model be ‘iteratively trained’ or dynamically adjusted in the Machine Learning Training patents do not represent a technological improvement… Iteratively training using selected training materials and dynamic adjustments based on real-time changes are incident to the very nature of machine learning.” Therefore, claimed invention does not provide an improvement to machine learning itself because the machine learning is being used as a tool, in its ordinary capacity, to perform the abstract idea. “Use of a computer or other machinery in its ordinary capacity for economic or other task (e.g., to receive, store, or transmit data) or simply adding a general purpose computer or computer components after the fact to an abstract idea (e.g., a fundamental economic practice or mathematical equation) does not integrate a judicial exception into a practical application or provide significantly more” (MPEP 2106.05(f)(2)).
Argument: “The Office further alleges that "the additional elements also amount to generally linking the use of the abstract idea to a particular technological environment or field of use" and that the additional elements amount to "adding the words 'apply it"' with the judicial exception. Office Action at 4-5. Applicant respectfully disagrees. The amended claims do not generally link an abstract idea to a technological environment. The amended claims recite the specific multi-model machine learning architecture described above, including the configured-with-feedback-loops architecture that refines the trained campaign execution model based on labeled data collected from executed advertisement campaigns. This is not generic linkage to a technological field. It is the technological field. Likewise, the amended claims do not amount to "apply it" with a judicial exception. The amended claim language directs the processing device to perform specific computational operations on specific machine learning models, configured in a specific architecture, using specific kinds of data. The amended claims specify how the artificial intelligence system operates.”
In response, the Examiner respectfully disagrees. As explained above, the additional elements are recited at a high level of generality, and are being used as tools, in their ordinary capacity, to perform the abstract idea. There is no indication from the claims or specification that the claimed invention provides an improvement to the functioning of a computer or an improvement to the machine learning itself. Similar to the claimed invention in Recentive Analytics Inc. v. Fox Corp., the claimed invention here merely uses machine learning in an advertising environment. Therefore, since there are no improvements to the functioning of a computer or improvements to any other technology or technical field, the additional elements amount to no more than generally linking the use of the judicial exception to a particular technical environment or field of use. “[P]atents that do not more than claim the application of generic machine learning to new data environments, without disclosing improvements to the machine learning models to be applied, are patent ineligible under § 101” (see. p.18 of Recentive Analytics, Inc. v. Fox Corp. (Fed. Cir. 2025)).
Argument: “The combination of elements recited in the amended claims, including the
recitation that "the trained campaign execution model is configured with feedback loops that refine the trained campaign execution model at different levels based on a plurality of key performance indicators monitored at predetermined times and based on labeled performance data collected from executed campaigns based on customer behavior," together with the recitations of trained foundation models selected based on a plurality of foundation model selection variables and the separately trained campaign execution model trained based on the foundation model outputs and the goal inputs, is not well- understood, routine, or conventional. The Office has not provided any evidence to the contrary. See USPTO Memorandum regarding Berkheimer v. HP, Inc., dated April 19, 2018 (instructing that an examiner should conclude that an element represents well- understood, routine, conventional activity "only when the examiner can readily conclude that the element(s) is widely prevalent or in common use in the relevant industry").”
In response, the Examiner respectfully disagrees. The Office Action does not take the position that any of the additional elements amount to adding insignificant extra-solution activity in Step 2A Prong Two that would warrant an analysis in Step 2B to determine that the additional element also amounts to simply appending well-understood, routine, and conventional activity. The Examiner notes that revised Step 2A Prong Two overlaps with Step 2B, and thus, many of the considerations need not be reevaluated in Step 2B because the answer will be the same. Viewing the limitations as an ordered combination does not add anything further than looking at the limitations individually. Their collective functions merely provide generic computer implementation (Step 2B, No).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SAM REFAI whose telephone number is (313)446-4822. The examiner can normally be reached M-F 9:00am-6:00pm.
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, Waseem Ashraf can be reached at 571-270-3948. 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.
/SAM REFAI/Primary Examiner, Art Unit 3621
Prosecution Insights