DETAILED ACTION
This action is in reply to the submission filed on 6/24/2026.
Status of Claims
Applicant’s cancellation of claim 24, amendments to claims 1-2, 7-14 and 16-23, and addition of claim 25 are acknowledged.
Claims 1-2, 7-14 and 16-23 are currently pending and have been examined.
Response to Remarks
Applicant's remarks filed 6/24/2026 have been fully considered and have been found not persuasive in full. After further search and consideration, prior art “Arel” is cited to teach feature vectors, nodes and dependencies for temporal network analysis. While Broyda teaches the functions and basic embodiment of a long short-term (LSTM) neural network performing temporal analysis for duplicate receipt detection of reimbursable expense eligibility as seen in present application’s para. 87 of the specification, it does not explicitly teach the newly added limitations of feature vectors, nodes and dependencies. Arel is relied upon to teach these limitations. Arel teaches temporal analysis and feature vector generation using multiple nodes and dependencies to input financial data to a neural network. See para. 40 of Arel. Arel also teaches generating confidence scores of the analysis. See para. 183 of Arel.
The added limitations serve to detail the workings of a machine learning model. Using the feature vectors, dependencies and nodes as claimed is seen as a ML model in its ordinary capacity, not an improvement to the functioning of the technology. Further, the core inventive concept is directed to improvements in automated financial data reporting, not to improvements to a technology using a technological solution to a technological problem. See para. 6 of present specification. Further, para. 87 of the specification teaches alternate types of models for said functionality. Therefore, 103 and 101 rejections are presented below in accordance with this response to remarks.
Regarding page 12 of remarks, the claimed technical technique is seen as using machine learning technology in its ordinary capacity to perform the abstract idea(s) of organizing human activity and mental processes.
Regarding page 13 of remarks, newly cited art is relied upon to teach the majority of the added claim limitations in the amendments filed 6/24/2026. The 103 rejection below posits it would have been obvious to combine the machine learning reimbursement analysis of Broyda with the reimbursement reporting system of Li, the reimbursement analysis of Bender, and the machine learning teachings of Arel to arrive at the claimed invention. Arel is relied upon to add to the machine learning teachings of Broyda, at least to provide nomenclature to the model in Broyda. Li and Bender are relied upon to teach certain report generation and analysis being claimed, such as the automatic report generation and location analysis.
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 1-2, 7-14 and 16-24 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1: the claims fall under statutory categories of processes and/or machines.
Step 2A Prong 1: the claims recite: obtaining training data that describes historical expenses and labels indicating reimbursable, monitoring data sources to obtain target employee activity indicated an expense at a business trip location; determining if the data satisfies an expense trigger by determining if other employees expense reports incurred at said location; responsive to said determination, obtain a confidence level of reimbursable for said data; determining if said level meets a threshold; generating a report if so, and submitted said report. These limitations, as drafted, is a process that, under its broadest reasonable interpretation, covers certain methods of organizing human activity, specifically commercial or legal interactions, including business relations. Some of these limitations also covers mental processes, including observation, evaluation, judgement and opinion.
Step 2A Prong 2: Said judicial exception is not integrated into a practical application because the claims as a whole, looking at the additional elements: non-transitory computer readable media, instructions executed by processors, training a machine learning model to predict reimbursable expenses by generating confidence levels, applying data to the machine learning model to obtain a first confidence level, generating and submitting a report without user approval by said computer embodiment, wherein the machine learning model is configured to process feature vectors representing the respective expenses associated with the target activities; wherein the machine learning model is a neural network model comprising a plurality of nodes that capture dependencies based on an order in which the feature vectors are supplied to the machine learning model, wherein applying the first data corresponding to the target activity of the employee to the machine learning model comprises (a) generating one or more feature vectors corresponding to the first data and (b) supplying the one or more feature vectors to the machine learning model, wherein generating the one or more feature vectors corresponding to the first data comprises generating a first feature vector corresponding to a first expense incurred at a first time and a second feature vector corresponding to a second expense incurred at a second time; wherein supplying the one or more feature vectors to the machine learning model is performed in a first temporal order consistent with the first time and the second time; wherein, because the plurality of nodes of the machine learning model captures dependencies based on the order in which the feature vectors are supplied to the machine learning model, the first confidence level is different than if supplying the one or more feature vectors were performed in a second temporal order inconsistent with the first time and the second time individually and in combination, merely use a computer or other machinery as a tool to perform the abstract idea (see MPEP 2106.05f.) The claims use these machines in their ordinary capacity for the purpose of applying the abstract idea(s). See SAP America, Inc. v. InvestPic LLC, 898 F.3d 1168 (Fed. Cir. 2018), “Even if a process of collecting and analyzing information is limited to “particular content” or a particular “source”, that limitation does not make the collection and analysis other than abstract.” Further, added limitations concerning feature vectors, nodes, and dependencies are directed to implementing core tenets of LSTM machine learning models to the abstract idea(s). See Applicant’s specification paragraph 87, in disclosing the use of LSTM models, or other models such as fuzzy logic models to perform the claimed functionality. Therefore, these limitations are invoking computers or other machinery merely as a tool to perform an existing process, such that it amounts to no more than mere instructions to apply the exception. Then, these additional elements do not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea, and the claim is directed to an abstract idea.
Step 2B: Said claims recite additional elements as listed above, which are not sufficient to amount to significantly more than the judicial exception because, as mentioned in Step 2A Prong 2, they use computers or other machinery to perform an abstract idea in such a way that amounts to no more than mere instructions to apply the exception using computers or other machinery. Mere instructions to apply an exception using computers or other machinery cannot provide an inventive concept. Therefore, the claim is not patent eligible.
Claim 2 clarifies the employee did not prove an expense description for the first data prior to the report and expense generation. This does not constitute an improvement to the functioning of the claimed technology. Claim 7 recites detecting movement of employee, and generating a second expense description associated with said movement from first to second location. This is seen as part of said abstract idea(s). Claim 8 recites the particulars of the first data including location activity of the employee. Claim 9 recites the data comprising expense description submission by a co-traveler. Claim 10 recites the data comprises a credit card charge of the target activity. Claim 11 recites the charge is temporally correlated with the target activity. Claim 12 recites the charge is geographically correlated with the activity. Claim 13 recites the data sources comprise applications operating independent of claimed system. Claim 14 recites said applications comprise a ride sharing application or food ordering application. Detailing information about the data being analyzed is seen as part of said abstract idea.
Claim 16 recites determining an employee has not submitted an expected expense according to a template. Claim 17 recites detecting a user has submitted a second expense item typically associated with a first expense. Claim 18 recites detecting employee qualifies for a recurring expense. This is seen as a mental process.
Claim 21 recites monitoring data sources to obtain second data; applying the second data to obtain a second confidence level of a second expense; determining a threshold level of confidence level; and requesting user approval of second expense. This is seen as a mental process performed by a computer running a machine learning model, similar to independent claims.
Claim 22 recites the training data comprises data that does not describe any specific expense and is temporally correlated with at least one expense in the historical expense data. Claim 23 recites the training data is geographically correlated with at least one historical expense. Claim 25 recites temporally-based vectors and confidence levels. This is not a meaningful limitation to the abstract idea by amounting to more than a computer or technology in its ordinary capacity applying the exception, in light of at least, SAP America, Inc. v. InvestPic LLC, 898 F.3d 1168 (Fed. Cir. 2018), “Even if a process of collecting and analyzing information is limited to “particular content” or a particular “source”, that limitation does not make the collection and analysis other than abstract.”
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.
The factual inquiries set forth in Graham v. John Deere Co., 383 U.S. 1, 148 USPQ 459 (1966), that are applied for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-2, 7-14 and 16-25 are rejected under 35 U.S.C. 103 as being unpatentable over Li et al. (Pub. No. US 2020/0058078 A1) in view of Broyda et al. (Pub. No. US 2021/0004949 A1), and in further view of Bender (US 2020/0065912) and Arel (US 2019/0279105).
Regarding Claims 1, 19 and 20, Li discloses one or more non-transitory machine-readable media storing instructions which, when executed by one or more processors, cause the one or more processors to perform operations comprising: (Li ¶0070; Processing system 1803 may comprise a microprocessor and other circuitry that retrieves and executes software 1806 from memory device)
monitoring, by an expense report generation system, one or more data sources to obtain data corresponding to target activity of an employee, (Li ¶0002; to provide an expense calculation based on event data.) wherein the target activity of the employee indicates an expense incurred by the employee at a particular location during a particular business trip; (Paragraphs 3 and 5 showing business trips including location data)
determining, by the expense report generation system, if the first data corresponding to target activity of the employee satisfies an expense trigger; (Li para. 41 showing qualifying event type is determined based on location.)
responsive to determining that the first data corresponding to target activity of the employee satisfies the expense trigger: applying data (para. 44 showing next operation of determination based on previous determined qualifying event)
generating, by the expense report generation system without user approval of the first expense description, an expense report comprising the first expense description; (Li ¶0048; mileage tracking report may be generated by application 102 and used to calculate a tax deduction and/or reimbursement) If the application generates the report, then the user is not involved with the report generation.
submitting the expense report for reimbursement without user approval of the expense report. (Para. 48, …which may be automatically submitted to a third-party entity)
Li does not, but Broyda teaches:
obtaining training data (paragraph 50, obtain data) that (a) describes a plurality of historical expenses and (b) comprises labels indicating, for each particular historical expense was reimbursable, (Para. 50, Broyda, historical audit of reimbursable expenses) at least by training the machine learning model to generate confidence levels corresponding to whether or not respective expenses associated with target activities of one or more employees are reimbursable; (Para 109, ML model configured for confidence score determination regarding reimbursable expenses)
training a machine learning model, using the training data, to predict whether expenses are reimbursable; (Para. 59, ML model for answering audit questions)
responsive to determining that the first data corresponding to target activity of the employee satisfies an expense trigger: (para. 106 showing determination that conflict is generated) applying, by the expense report generation system, the first data corresponding to target activity of the employee to the machine learning model, (para. 107, ML model processing applied data) to obtain from the machine learning model a first confidence level corresponding to whether or not a first set of one or more expenses associated with the target activity of the employee is reimbursable; (Para. 109, confidence value for reimbursable expenses is determined by ML)
determining, by the expense report generation system, whether or not the first confidence level satisfies a predetermined threshold; (Para. 110 teaching threshold decisions)
responsive to determining that the confidence level satisfies the predetermined threshold:
generating, by the expense report generation system, a first expense description for the first set of one or more expenses; (para. 118 determines result of analysis and associates result with transaction)
wherein the machine learning model is configured to process the respective expenses associated with the target activities, and first data corresponding to the target activity of the employee (Para. 59, ML model for answering audit questions) (Para 109, ML model configured for confidence score determination regarding reimbursable expenses)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the system of expense report generation using machine learning in Li with the known technique of training a machine learning model and using threshold-based logic in Broyda because applying the known technique would have yielded predictable results and resulted in an improved system by allowing automation of data categorization. (Paragraph 118 of Broyda showing automated decision making for reimbursability.)
Li as modified by Broyda does not, but Bender teaches: determining, by the expense report generation system, if the first data corresponding to target activity of the employee satisfies an expense trigger, at least by determining if one or more other employees prepared respective expense descriptions for expenses incurred at the particular location during the particular business trip. (para. 83 showing other employees’ expense descriptions’ locations compared to target data location for reimbursement eligibility determination; see also paragraphs 79 and 80)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the system of expense report generation using machine learning in Li as modified by Broyda with the known technique of location verification in Bender, because applying the known technique would have yielded predictable results and resulted in an improved system by allowing for greater confidence levels of reimbursable labeling. (See para. 80 of Bender showing confidence level determination using location verification.)
Li as modified by Broyda and Bender does not teach the following limitations. Broyda teaches multiple ML inputs, which are usually formatted as feature vectors, but it does not teach explicitly feature vectors. Broyda also teaches a long short-term memory neural network. (para. 210 of Broyda). Examiner notes the network in present application’s specification paragraph 87 teaches the same type of network, and paragraph 87 is used as the subject matter material for these newly added claim limitations. Additionally, Broyda teaches the duplicate receipt detection example in paragraph 87 that is used as an embodiment for the newly added claim limitations of temporal inputs and confidence scores. (para. 51 of Broyda, Figures 6-9, para. 68, and para. 210 of Broyda). Then, Broyda teaches the functionality of the claimed subject matter in present application. However, Broyda does not explicitly teach feature vectors, nodes or dependencies. These limitations are seen as elements of certain machine learning models. Arel teaches these:
wherein the machine learning model is configured to process feature vectors; wherein the machine learning model is a neural network model comprising a plurality of nodes that capture dependencies based on an order in which the feature vectors are supplied to the machine learning model, wherein applying the first data to the machine learning model comprises (a) generating one or more feature vectors corresponding to the first data and (b) supplying the one or more feature vectors to the machine learning model, (para. 40 of Arel showing multiple nodes, feature vectors for temporal analysis of data; para. 43 showing feature vector provisioning to nodes)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the system of expense report generation using machine learning in Li as modified by Broyda with the known technique of using feature vectors, nodes, and dependencies in machine learning models, because applying the known technique would have yielded predictable results and resulted in an improved system by allowing for data analysis of temporal nature. See para. 40 of Arel.
Regarding Claim 2, Li as modified by Broyda and Bender teaches the one or more media of claim 1, wherein the employee did not provide any expense description associated with the first set of one or more expenses prior to the expense report generation system generating the first expense description. (Li ¶0027; One or more calendar events are received indicating an event type, a location, a date, and a time associated with each of the one or more calendar events.) Examiner notes the expense description comes from calendar events, not user input.
Regarding Claim 7, Li as modified by Broyda and Bender teaches the one or more media of claim 1, the operations further comprising:
responsive to detecting movement of the employee from a first location to a second location: generating a second expense description associated with movement of the employee from the second location to the first location. (Li ¶0095; wherein providing an estimated mileage calculation based at least in part on the location associated with the qualifying calendar event comprises calculating the estimated mileage calculation based on a distance estimation from a default address associated with the user to an address associated with the qualifying calendar event)
Regarding Claim 8, Li as modified by Broyda and Bender teaches the one or more media of claim 1, wherein the first data corresponding to target activity of the employee comprises data indicating that the employee has visited a particular location. (Li ¶0029; location and time may be further determined based on a GPS coordinates of the user device at a time in which the receipt was received by the user device.)
Regarding Claim 9, Li as modified by Broyda and Bender teaches the one or more media of claim 1, wherein the first data corresponding to target activity of the employee comprises data indicating an expense description submission by a co-traveler of the employee. (Li ¶0029; automatically tracking other attendees present at the meeting…¶0041; event type may be determined based on the location (e.g., office or designated business meeting location), based on additional attendees (e.g., employees and/or clients))
Regarding Claim 10, Li as modified by Broyda and Bender teaches the one or more media of claim 1, wherein the first data corresponding to target activity of the employee comprises data indicating a credit card charge that is associated with target activity of the employee. (Broyda para. 80, ML model categorizes credit card transactions)
Regarding Claim 11, Li as modified by Broyda and Bender teaches the one or more media of claim 10, the operations further comprising:
determining that the credit card charge correlates temporally with the target activity of the employee. (Broyda para. 54, determines date of transaction is supported by expense rules)
Regarding Claim 12, Li as modified by Broyda and Bender teaches the one or more media of claim 10, the operations further comprising:
determining that the credit card charge correlates geographically with the target activity of the employee. (Broyda para. 53, location of transaction supported by expense rules)
Regarding Claim 13, Li as modified by Broyda and Bender teaches the one or more media of claim 1, wherein the one or more data sources comprise one or more applications operating independent of the expense report generation system. (Li ¶0030; the event to be tracked as a business purpose may be a transportation service associated with a transportation application)
Regarding Claim 14, Li as modified by Broyda and Bender teaches the one or more media of claim 13, wherein the one or more applications operating independent of the expense report generation system comprise one or more of a ride sharing application or a food ordering application. (Li ¶0030; One or more businesses associated with a user account may be transportation applications such as Uber or Lyft. A user may associate a transportation business application with the user account to synchronize events that are to be recorded for business purposes.)
Regarding Claim 16, Li as modified by Broyda and Bender teaches the one or more media of claim 15, the operations further comprising:
determining that the employee has not submitted an expected expense according to the expense template. (Li ¶0029; categorizing associated receipts without requiring the user to manually enter and categorize the information in the application)
Regarding Claim 17, Li as modified by Broyda and Bender teaches the one or more media of claim 1, the operations further comprising:
detecting that the user has submitted a second expense item that is typically associated with the first expense item. (Li ¶0029; location and time that the receipt was acquired may further be used to associate the additional expenses with the categorized calendar event)
Regarding Claim 18, Li as modified by Broyda and Bender teaches the one or more media of claim 1:
wherein the first expense description is associated with a recurring expense; the instructions further comprising: (Li ¶0067; if event “a” is a recurring event, then event “a” by be queried in CoreData using a calendar ID.)
detecting that the employee qualifies for the recurring expense. (Li ¶0067; The calendar ID may be set using the calendar ID plus the reoccurrence time, such as the number of the occasions that the event has occurred)
Regarding Claim 21, Li as modified by Broyda and Bender teaches the one or more media of claim 1, the operations further comprising:
monitoring, by the expense report generation system, the one or more data sources to obtain second data corresponding to target activity of the employee; (Li ¶0002; to provide an expense calculation based on event data.)
Li does not, but Broyda teaches:
applying, by the expense report generation system, the second data corresponding to target activity of the employee to the machine learning model, (para. 107, ML model processing applied data) to obtain a second confidence level corresponding to whether or not a second set of one or more expenses associated with the target activity of the employee is reimbursable; (Para. 109, confidence value for reimbursable expenses is determined by ML)
determining, by the expense report generation system, whether or not the second confidence level satisfies the predetermined threshold; (Para. 110 teaching threshold decisions)
responsive to determining that the second confidence level does not satisfy the predetermined threshold: requesting user approval of a second expense description for the second set of one or more expenses. (Para. 118, output to a user auditor if threshold is not met)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the system of expense report generation using machine learning in Li with the known technique of training a machine learning model and using threshold based logic in Broyda because applying the known technique would have yielded predictable results and resulted in an improved system by allowing automation of data categorization. (Para. 118 of Broyda showing automation of reimbursability determination.)
Regarding Claim 22, Li as modified by Broyda and Bender teaches the one or more media of claim 1, wherein the training data comprises data that (a) does not describe any specific expense and (b) is temporally correlated with at least one expense in the plurality of historical expenses. (Li ¶0031; times and locations associated with a plurality of qualifying calendar events are tracked in a cloud-based data repository to be ingested by a machine learning system) Examiner notes the data consists of times, not expenses.
Regarding Claim 23, Li as modified by Broyda and Bender teaches the one or more media of claim 1, wherein the training data comprises data that (a) does not describe any specific expense and (b) is geographically correlated with at least one expense in the plurality of historical expenses. (Li ¶0031; times and locations associated with a plurality of qualifying calendar events are tracked in a cloud-based data repository to be ingested by a machine learning system) Examiner notes the data consists of locations, not expenses.
Regarding Claim 24, Li as modified by Broyda and Bender teaches the one or more media of claim 1. Li does not, but Broyda teaches: wherein the machine learning model is a neural network model (Broyda para. 62, model can be neural network model) in which nodes comprise a memory to remember temporal associations between different expenses and activities. (Para. 122, model can take into account transaction times for processing different activities)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the system of expense report generation using machine learning in Li with the known technique of training a machine learning model and using threshold based logic in Broyda because applying the known technique would have yielded predictable results and resulted in an improved system by allowing automation of data categorization. (Paragraph 118 of Broyda showing automated decision making for reimbursability.)
Regarding Claim 25, Li as modified by Broyda and Bender teaches claim 1. It does not, but Arel teaches:
wherein generating the one or more feature vectors corresponding to the first data comprises generating a first feature vector corresponding to a first expense incurred at a first time and a second feature vector corresponding to a second expense incurred at a second time; (para. 40 of Arel showing multiple nodes, feature vectors for temporal analysis of data)
wherein supplying the one or more feature vectors to the machine learning model is performed in a first temporal order consistent with the first time and the second time; (para. 43 showing feature vector provisioning to nodes; para. 172 showing time “t+1” logic of supplying feature vectors)
wherein, because the plurality of nodes of the machine learning model captures dependencies based on the order in which the feature vectors are supplied to the machine learning model, the first confidence level is different than if supplying the one or more feature vectors were performed in a second temporal order inconsistent with the first time and the second time. (paras 183 and 184 showing confidence in output generated from temporal inputs)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the system of expense report generation using machine learning in Li as modified by Broyda with the known technique of using feature vectors, nodes, and dependencies in machine learning models, because applying the known technique would have yielded predictable results and resulted in an improved system by allowing for data analysis of temporal nature. See para. 40 of Arel.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, this action is made final. See MPEP § 706.07(a). 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 extension fee 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 Aaron Tutor, whose telephone number is 571-272-3662. The examiner can normally be reached Monday through Friday, 9 AM to 5 PM.
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, Fahd Obeid, can be reached at 571-270-3324. The fax number for the organization where this application or proceeding is assigned is 571-273-5266.
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.
/AARON TUTOR/Primary Examiner, Art Unit 3627
Prosecution Insights