Business process management using adversial agents

Abstract

Systems and methods for improving business processes. In some embodiments, the method includes receiving business process data; generating, based on the business process data, a process map representing a business process using a first machine learning model deployed at the cloud-based cluster; generating, based on the process map, at least one alternative process map using a second machine learning model; evaluating the generated process maps using at least one business objective; selecting an improved process map from the generated process maps based on the evaluation; presenting, at an interactive user interface, at least one of the generated process maps, the at least one generated process map including the improved process map; receiving, at the interactive user interface, a user selection indicating a preferred process map from the at least one generated process map; and implementing the preferred process map.

Description

PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PCBUSINESS PROCESS MANAGEMENT USING AD VERSIAL AGENTSCROSS REFERENCE TO RELATED APPLICATIONS

[0001] The present application claims the benefit of and priority to co-pending United States provisional application no. 63 / 689,054, filed on August 30, 2024, the content of which is hereby incorporated by reference as if set forth in its entirety herein.TECHNICAL FIELD

[0002] Embodiments described herein generally relate to systems and methods for business process management and, more particularly but not exclusively, to systems and methods for improving business processes using generative adversarial process agents.BACKGROUND

[0003] Various businesses and other entities may use business processes to produce services and / or products for their customers. Such businesses often seek to improve their business processes to meet dynamic objectives, improve resource utilization, and maintain a competitive edge in dynamic, rapidly evolving business environments. Improving business processes may be challenging, however, since businesses processes often involve numerous interconnected events, decisions, and / or activities that vary across organizational sub-entities. The complexity of managing these processes may be further amplified by the diverse objectives and constraints that different stakeholders impose.

[0004] Conventional approaches to business process management typically rely on static workflows, manual analysis, and / or generic optimization strategies. For example, organizations may manually create visual representations of workflows, manually gather limited data from a subset of stakeholders, and manually analyze underlying causes of inefficiencies. Such approaches face numerous shortcomings, such as lack of adaptability to changing environments, limited data utilization and fragmented evaluation metrics. There remains a need, therefore, for business process management systems and methods that can adapt to current business environments and improve business processes effectively. Providing such systems and methods may effectively allow businesses to gain more valuable insight as to how to tackle business inefficiencies and challenges.PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PCSUMMARY

[0005] This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description section. This summary is not intended to identify or exclude key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.

[0006] According to one aspect, embodiments relate to a method for improving business processes. The method includes receiving, at a cloud-based processing cluster, business process data; generating, at the cloud-based processing cluster based on the business process data, a process map representing a business process using a first machine learning model deployed at the cloud-based cluster; generating, based on the process map, at least one alternative process map using a second machine learning model deployed in parallel with the first machine learning model at the cloud-based processing cluster; evaluating the generated process maps using at least one business objective; selecting an improved process map from the generated process maps based on the evaluation; presenting, at an interactive user interface, at least one of the generated process maps, the at least one generated process map including the improved process map; receiving, at the interactive user interface, a user selection indicating a preferred process map from the at least one generated process map, and implementing the preferred process map.

[0007] In some embodiments, the method further comprises querying at least one data lake using the generated process maps and associated queries, storing the generated process maps and the associated queries in a query storage resource, and evaluating the generated process maps and the associated queries based on user requirements.

[0008] In some embodiments, the evaluating of the generated process maps and the associated queries is based on a similarity search.

[0009] In some embodiments, the at least one alternative process map is generated from a generative adversarial network.

[0010] In some embodiments, the generative adversarial network uses a multi-objective optimization function.PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC

[0011] In some embodiments, the machine learning model comprises a generative adversarial network, and wherein the process map comprises a connectivity graph generated using the generative adversarial network.

[0012] In some embodiments, the machine learning model is configured to use spectral graph theory to improve a layout of the connectivity graph.

[0013] In some embodiments, the evaluating of the generated process maps is based on a decision function expressed as a classification problem.

[0014] In some embodiments, the method further comprising generating at least one summary for the improved process map, the at least one summary including a rationale for selecting the improved process map.

[0015] In some embodiments, the presenting of at least one of the generated process maps comprises presenting a visual comparison of the improved process map and at least one additional generated process map.

[0016] In some embodiments, the method further comprises receiving, at the interactive user interface, a second user selection indicating at least one implementation parameter for the preferred process map, and wherein the implementing the preferred process map is based on the at least one implementation parameter.

[0017] According to another aspect, embodiments relate to an apparatus for improving business processes. The apparatus includes an interactive user interface; a memory storing instructions; at least one processor executing the instructions stored on the memory to perform the steps of: receiving business process data, generating, based on the business process data, a process map representing a business process using a machine learning model deployed at a cloud-based processing cluster, generating, based on the process map, at least one alternative process map, evaluating the generated process maps using at least one business objective, selecting an improved process map from the generated process maps based on the evaluation, presenting, using the interactive user interface, at least one of the generated process maps, the at least one generated process map including the improved process map, receiving, using the interactive user interface, a user selection indicating a preferred process map from the at least one generated process map, and implementing the preferred process map.PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC

[0018] In some embodiments, the at least one processor is further configured to perform the steps of: querying at least one data lake using the generated process maps and associated queries, storing the generated process maps and the associated queries in a query storage resource, and evaluating the generated process maps and the associated queries based on user requirements received from the user interface.

[0019] In some embodiments, the evaluating of the generated process maps and the associated queries is based on a similarity search.

[0020] In some embodiments, the at least one alternative process map is generated from a generative adversarial network.

[0021] In some embodiments, the generative adversarial network uses a multi-objective optimization function.

[0022] In some embodiments, the machine learning model comprises a generative adversarial network, and wherein the process map comprises a connectivity graph generated using the generative adversarial network.

[0023] In some embodiments, the machine learning model is configured to use spectral graph theory to improve a layout of the connectivity graph.

[0024] According to another aspect, embodiments relate to a system for improving business processes. The system includes a process mapper module for generating a process map representing a business process using a machine learning model deployed at a cloud-based processing cluster; a process innovator module for generating at least one alternative process map; an evaluator module for evaluating the generated process maps using at least one business objective; a decision-making module for selecting an improved process map from the generated process map; and an interactive user interface for: presenting at least one of the generated process maps, the at least one generated process map including the improved process map, and receiving a user selection indicating a preferred process map from the at least one generated process map.

[0025] In some embodiments, the at least one alternative process map is generated from a generative adversarial network.PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PCBRIEF DESCRIPTION OF DRAWINGS

[0026] Non-limiting and non-exhaustive embodiments of the invention are described with reference to the following figures, wherein like reference numerals refer to like parts throughout the various views unless otherwise specified:

[0027] FIG. 1 depicts a system for improving business processes using adversarial agents in accordance with one embodiment;

[0028] FIG. 2 depicts a flowchart of a method for improving business processes using adversarial agents in accordance with one embodiment;

[0029] FIG. 3 depicts a flowchart of a workflow for improving business processes using adversarial agents in accordance with another embodiment;

[0030] FIG. 4 depicts a connectivity graph for a business process in accordance with one embodiment; and

[0031] FIG. 5 depicts an implementation of a user interface in accordance with one embodiment.

[0032] FIG. 6 depicts a flowchart of a method for improving business processes using a unified model in accordance with one embodiment.DETAILED DESCRIPTION

[0033] Various embodiments are described more fully below with reference to the accompanying drawings, which form a part hereof, and which show specific exemplary embodiments. However, the concepts of the present disclosure may be implemented in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided as part of a thorough and complete disclosure, to fully convey the scope of the concepts, techniques and implementations of the present disclosure to those skilled in the art. Embodiments may be practiced as methods, systems or devices. Accordingly, embodiments may take the form of a hardware implementation, an entirely software implementation or an implementation combining software and hardware aspects. The following detailed description is, therefore, not to be taken in a limiting sense.PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC

[0034] Reference in the specification to “one embodiment” or to “an embodiment” means that a particular feature, structure, or characteristic described in connection with the embodiments is included in at least one example implementation or technique in accordance with the present disclosure. The appearances of the phrase “in one embodiment” in various places in the specification are not necessarily all referring to the same embodiment. The appearances of the phrase “in some embodiments” in various places in the specification are not necessarily all referring to the same embodiments.

[0035] Some portions of the description that follow are presented in terms of symbolic representations of operations on non-transient signals stored within a computer memory. These descriptions and representations are used by those skilled in the data processing arts to most effectively convey the substance of their work to others skilled in the art. Such operations typically require physical manipulations of physical quantities. Usually, though not necessarily, these quantities take the form of electrical, magnetic or optical signals capable of being stored, transferred, combined, compared and otherwise manipulated. It is convenient at times, principally for reasons of common usage, to refer to these signals as bits, values, elements, symbols, characters, terms, numbers, or the like. Furthermore, it is also convenient at times, to refer to certain arrangements of steps requiring physical manipulations of physical quantities as modules or code devices, without loss of generality.

[0036] However, all of these and similar terms are to be associated with the appropriate physical quantities and are merely convenient labels applied to these quantities. Unless specifically stated otherwise as apparent from the following discussion, it is appreciated that throughout the description, discussions utilizing terms such as “processing” or “computing” or “calculating” or “determining” or “displaying” or the like, refer to the action and processes of a computer system, or similar electronic computing device, that manipulates and transforms data represented as physical (electronic) quantities within the computer system memories or registers or other such information storage, transmission or display devices. Portions of the present disclosure include processes and instructions that may be embodied in software, firmware or hardware, and when embodied in software, may be downloaded to reside on and be operated from different platforms used by a variety of operating systems.PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC

[0037] The present disclosure also relates to an apparatus for performing the operations herein. This apparatus may be specially constructed for the required purposes, or it may comprise a general-purpose computer selectively activated or reconfigured by a computer program stored in the computer. Such a computer program may be stored in a computer readable storage medium, such as, but is not limited to, any type of disk including floppy disks, optical disks, CD-ROMs, magnetic-optical disks, read-only memories (ROMs), random access memories (RAMs), EPROMs, EEPROMs, magnetic or optical cards, application specific integrated circuits (ASICs), or any type of media suitable for storing electronic instructions, and each may be coupled to a computer system bus. Furthermore, the computers referred to in the specification may include a single processor or may be architectures employing multiple processor designs for increased computing capability.

[0038] The processes and displays presented herein are not inherently related to any particular computer or other apparatus. Various general-purpose systems may also be used with programs in accordance with the teachings herein, or it may prove convenient to construct more specialized apparatus to perform one or more method steps. The structure for a variety of these systems is discussed in the description below. In addition, any particular programming language that is sufficient for achieving the techniques and implementations of the present disclosure may be used. A variety of programming languages may be used to implement the present disclosure as discussed herein.

[0039] In addition, the language used in the specification has been principally selected for readability and instructional purposes and may not have been selected to delineate or circumscribe the disclosed subject matter. Accordingly, the present disclosure is intended to be illustrative, and not limiting, of the scope of the concepts discussed herein.

[0040] As discussed above, currently available techniques for managing business processes are insufficient and ineffective. The embodiments herein provide techniques for improving business processes. These embodiments may include receiving, at a cloud-based processing cluster, business process data; generating, at the cloud-based processing cluster based on the business process data, a process map representing a business process using a first machine learning model deployed at the cloud-based cluster; generating, based on the process map, at least one alternative process map using a second machine learning model deployed in parallel with the firstPCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC machine learning model at the cloud-based processing cluster; evaluating the generated process maps using at least one business objective; selecting an improved process map from the generated process maps based on the evaluation; presenting, at an interactive user interface, at least one of the generated process maps, the at least one generated process map including the improved process map; receiving, at the interactive user interface, a user selection indicating a preferred process map from the at least one generated process map; and implementing the preferred process map.

[0041] As used herein, a “business process” may refer to any set of tasks and / or activities implemented by an organization to achieve business objectives. Such tasks and / or activities may involve various inputs (e.g., resources, information, etc.), processes, decision points, and / or outputs (e.g., products, services, outcomes, etc.).

[0042] As used herein, an “entity” may refer to any individual, group, sub-group, business, and / or organization involved with business processes.

[0043] FIG. 1 illustrates a system 100 for improving business processes using adversarial agents in accordance with one embodiment. The system 100 may include any number of components for performing operations related to improving business processes. As shown, for example, the system 100 may include a user device 102 accessible by a user 104. The system 100 may be implemented within any suitable computing environment, such as a cloud-based computing environment (e.g., a cloud-based processing cluster), a distributed computing environment, and / or a virtualized computing environment.

[0044] The user device 102 may be any suitable device capable of presenting a user interface to the user 104. The user device 102 may be a smartphone, tablet, PC, laptop, smart TV, smartwatch, and / or any other type of device whether available now or invented hereafter.

[0045] The user 104 may include an administrator tasked with managing business processes for a business and / or any other suitable organization. The user 104 may be associated with one or more entities 106 (e.g., entities 106-1 and 106-2).

[0046] The user 104 may be interested in viewing and / or evaluating generated process maps associated with a business process. For example, the user 104 may be an administrator or stakeholder interested in knowing the efficacy of a process map, how well an implementation of the process map may perform in current and / or future business environments, the cost(s) associatedPCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC with implementing the process map, the benefits resulting from implementing the process map, and / or the like. Accordingly, the user device 102 may present the user 104 with summaries and / or reports for a selected process map at the user interface.

[0047] The user device 102 may allow the user 104 to provide input related to managing business processes (e.g., at the user interface). For example, the user device 102 may allow the user 104 to select a preferred process map from a presented set of process maps and initiate an implementation of the preferred process map.

[0048] The user device 102 may allow the user 104 to monitor implementations of a selected process map. The user device 102 may present metrics related to real-time implementation of the process map and allow the user 104 to make adjustments to the implementation, such as by adjusting business objectives and / or process parameters during unforeseen challenges and / or deviations.

[0049] The user device 102 may be in communication with one or more processors 108 including an interface executing instructions on memory 110. The processor(s) 108 may be any hardware device capable of executing instructions stored on memory 110 to provide various components or modules, discussed below. The processor(s) 108 may include a microprocessor, a field programmable gate array (FPGA), an application-specific integrated circuit (ASIC), or other similar devices.

[0050] In some embodiments, such as those relying on one or more ASICs, the functionality described as being provided in part via software may instead be configured into the design of the ASICs and, as such, the associated software may be omitted. The processor(s) 108 may be configured as part of the user device 102 (e.g., a laptop) or located at some remote location.

[0051] The memory 110 may be LI, L2, L3 cache, or RAM memory configurations. The memory 110 may include non-volatile memory such as flash memory, EPROM, EEPROM, ROM, and PROM, or volatile memory such as static or dynamic RAM, as discussed above. The exact configuration / type of memory 110 may vary as long as instructions for performing the various steps described herein can be performed by the system 100.

[0052] The processor(s) 108 may be in operable connectivity with the entities 106 over one or more networks 112. The network(s) 112 may link the various components with various types ofPCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC network connections. The network(s) 112 may be comprised of, or may interface to, any one or more of the Internet, an intranet, a Personal Area Network (PAN), a Local Area Network (LAN), a Wide Area Network (WAN), a Metropolitan Area Network (MAN), a storage area network (SAN), a frame relay connection, an Advanced Intelligent Network (AIN) connection, a synchronous optical network (SONET) connection, a digital Tl, T3, El, or E3 line, a Digital Data Service (DDS) connection, a Digital Subscriber Line (DSL) connection, an Ethernet connection, an Integrated Services Digital Network (ISDN) line, a dial-up port such as a V.90, a V.34, or a V.34bis analog modem connection, a cable modem, an Asynchronous Transfer Mode (ATM) connection, a Fiber Distributed Data Interface (FDDI) connection, a Copper Distributed Data Interface (CDDI) connection, or an optical / DWDM network.

[0053] The network(s) 112 may also comprise, include, or interface to any one or more of a Wireless Application Protocol (WAP) link, a Wi-Fi link, a microwave link, a General Packet Radio Service (GPRS) link, a Global System for Mobile Communication (GSM) link, a Code Division Multiple Access (CDMA) link, or a Time Division Multiple access (TDMA) link such as a cellular phone channel, a Global Positioning System (GPS) link, a cellular digital packet data (CDPD) link, a Research in Motion, Limited (RIM) duplex paging type device, a Bluetooth radio link, or an IEEE 802.11 -based link.

[0054] The entities 106 may communicate data and / or commands related to business processes. For example, the entities 106 may be entire companies and / or organizations with overall business objectives. In some examples, the entities 106 may include various sub-entities or groups. The sub-entities may include any suitable sub-groups within a larger entity. In some embodiments, the sub-entities may include external suppliers and / or partners. The entities 106 may provide data such as survey data and / or quantitative data for generating process maps.

[0055] The processor(s) 108 may include and / or otherwise execute various modules, such as a process mapper 114, a process innovator 116, an evaluator 118, a decision maker 120, and an implementor 122.

[0056] The process mapper 114 may be configured to generate process maps representing workflows and business operations for business processes. For example, the process mapper 114 may generate a process map corresponding to a business process, construct matrices (e.g., structural and / or process matrices) defining relationships between sub-entities and processes,PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC implement hierarchical embedding for process components (e.g., activities, decision points, etc.), and / or compute context weight matrices.

[0057] The process innovator 116 may be configured to generate process maps that are proposed as alternatives to process maps generated by the process mapper 114. The process innovator 116 may aim to challenge the status quo of business processes by proposing innovative and adaptable alternative process maps. For example, the process innovator 116 may generate at least one alternative process map, perform eigenvector decomposition (e.g., on connectivity matrices), calculate process deviations (e.g., to assess alternative processes), and / or apply stochastic optimization to process maps. The process innovator 116 and the process mapper 114 may communicate with each other to perform their operations, such as by sharing learned features with each other.

[0058] The evaluator 118 may be configured to evaluate process maps, such as process maps generated by the process mapper 114 and / or the process innovator 116. For example, the evaluator 118 may assess performance metrics, analyze resource usage levels, and / or measure business impacts of implementing process maps. The evaluator 118 may evaluate process maps using at least one business objective and / or user preference.

[0059] The decision maker 120 may be configured to select an improved process map from a set of generated process maps. The decision maker 120 may select the improved process map using evaluations from the evaluator 118. For example, the decision maker 120 may select the improved process map based on evaluated performance metrics, implementation feasibility, and / or alignment with business objectives.

[0060] The implementor 122 may be configured to implement a selected process map, such as the improved process map selected by the decision maker 120. For example, the implementor 122 may generate deployment plans, automate process updates, reallocate implementation resources, and / or monitor results of implementing process maps.

[0061] In some embodiments, the processor(s) 108 may communicate with a data lake 124. The data lake 124 may operate as a repository for an organization’s data to flow to, where such data may be in a raw format and / or in a structured format. For example, the processor(s) 108 may store generated process maps and / or business process data at the data lake 124. Metadata tagging may be implemented to facilitate queries of data elements in the data lake (e.g., process maps),PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC particularly in embodiments where the data lake contains multiple stores of data and / or data in formats not easily accessible or readable (e.g., unstructured data, semi -structured data, structured data). The queries may be inputted by the user 104 (e.g., at the user device 102) and / or automatically generated by the system 100.

[0062] FIG. 2 depicts a flowchart of a method 200 for improving business processes using adversarial agents in accordance with one embodiment. While FIG. 2 shows illustrative operations according to one embodiment, other embodiments may omit, add to, reorder, and / or modify any of the operations shown in FIG. 2. As described, the method 200 may be implemented by one or more components of the system 100 as illustrated in FIG. 1. However, it is to be understood that the method 200 may be implemented by any suitable component and / or system.

[0063] Operation 202 may include receiving business process data. The business process data may include any suitable data related to business processes. For example, the business process data may include entity structure data (e.g., data indicating relationships between sub-entities of an entity), resource data (e.g., resources available to and / or in use by sub-entities), process execution data (e.g., activity logs of business activities, transaction data, workflow histories, etc.), operational metrics (e.g., error rates of business operations, cycle times of business operations, etc ), financial data (e.g., costs, revenues, profits, budgets, etc.), document data (e.g., communication records, forms, etc.), and / or performance indicator data (e.g., performance metrics, forecasts, etc.). The business process data may be associated with any suitable business sector and / or domain, such as manufacturing and / or healthcare management. In some embodiments, the business process data may include business objectives and / or user preferences associated with a business process (e.g., objectives and / or preferences inputted by a user and / or administrator). In this manner, the business process data may include data originating from a variety of organizational sources.

[0064] In some embodiments, the business process data may be represented by at least one matrix and / or other data structures. For example, an entity may be represented by an entity matrix E. Assuming the entity is composed of n sub-entities, E may be written as:where each etrepresents a sub-entity.PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC

[0065] In some examples, the entity may be represented by an entity structure matrix S that represents relationships between sub-entities. S may be written as:

[0066] where each S[j represents the interaction strength between sub-entities etand ej.

[0067] In some examples, the business objectives of an entity may be represented by an objective vector O, which may be written as:O = olto2, o3, ... , on].In some examples, the business objectives for a sub-entity e{may be represented by a vector Obwhich may be written as: O(- = [o(1, oi2, oi3>.... Oim].

[0068] In some examples, the resources available to sub-entities may be represented by a vector R, which may be written as:R = [ri, r2, r3, ... , rn], where each element r(- represents the resources allocated to sub-entity eb

[0069] In some examples, the processes executed by a sub-entity etto meet its objectives may be represented by a process matrix Pbwhich may be written asPill Pil2 ■■■ PilmP > Pi21 Pi22 ■■■ Pi2mPim2 Pim2 ■■■ Pimm where each element ptjkrepresents a process carried out within the sub-entity eb

[0070] Operation 204 may include generating a process map representing a business process. The system 100 (e.g., the process mapper 114) may use a first machine learning model trained to generate process maps based on business process data. The first machine learning model may generate the process map based on the business process data. The process map may include any suitable representation that outlines various items of a business process, such as steps, activities, decision points, workflows, inputs, and / or outputs of the business process. For example, the business process data may indicate steps, workflows, and / or resources regarding an orderPCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC fulfillment process, such as an “order placed” step and an “order confirmation” step. The corresponding process map may visually represent a workflow flowing from the “order placed” step to the “order confirmation” step. In some examples, the process map may include a connectivity graph where nodes represent process steps and edges represent the flow of information, materials, and / or decisions.

[0071] The first machine learning model may include any number of suitable machine learning models, such as at least one of decision tree learning algorithms, association rule learning algorithms, artificial neural network learning algorithms, deep learning algorithms, bitmap algorithms, graph transformers, and / or any other suitable data analysis technique as may serve a particular implementation. In some embodiments, the model may be implemented by one or more neural networks, such as one or more deep convolutional neural networks (CNN) using internal memories of its respective kernels (filters), graph neural networks (GNN), recurrent neural networks (RNN), and / or long / short term memory neural networks (LSTM). The model may be multi-layer. For example, the model may be implemented by a neural network that includes an input layer, one or more hidden layers, and an output layer. The model may be trained on historical business process data, historical business process maps, simulated processes, and / or optimization outcomes to identify various process patterns and / or relationships. The model may be trained using reinforcement learning, supervised learning, and / or other suitable training methods.

[0072] In some embodiments, the first machine learning model may be implemented as a generative adversarial agent (GAN). In general, a GAN may include a generative machine learning model (e.g., a neural network) trained to generate process maps based on input data and an evaluative machine learning model (e.g., a second neural network) trained to distinguish between process maps derived from actual data (i.e., real process maps) and process maps generated by the generative machine learning model. The generative machine learning model and the evaluative machine learning model may compete in a min-max optimization game where the objective is to generate realistic process maps while ensuring the evaluative model can distinguish between real and generated process maps effectively.

[0073] In some embodiments, the first machine learning model may generate the process map using an objective function. For example, the objective function used by the first machine learning model may be expressed as the following equation:PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC minGmaxDV D, G) = Ex PdQta(x)[logD(x)] + Ez~Pz(z)[log (1 - £>(G(z)))] , where G is the generative machine learning model, D is the evaluative machine learning model, x represents the business process data, and z represents the input noise to the generative machine learning model.

[0074] The first machine learning model may be deployed at any suitable system and / or processing resource, such as any of the systems and / or resources described herein. In some embodiments, the first machine learning model may be deployed at a processing cluster (e.g., a cloud-based processing cluster) that includes a plurality of processing nodes.

[0075] As described above, the process map may include a connectivity graph. In some embodiments, the system 100 may apply spectral graph techniques to optimize the connectivity graph. The system 100 may identify properties of a graph’s connectivity, such as clusters and / or partitions in the graph that correspond to functional sub-processes and / or potential bottlenecks for the business process. The system 100 may optimize graph partitioning (e.g., minimizing interconnections between partitions), minimize edge crossings (e.g., repositioning nodes), and / or maximize visual clarity (e.g., through force-directed layouts).

[0076] In some embodiments, the spectral graph processes may include applying eigenvalue decomposition by examining the eigenvalues of the graph Laplacian. This may be expressed mathematically as the following equation:L = D -A, where L is the graph Laplacian, D is the degree matrix, and A is the adjacency matrix of the connectivity graph. This eigenvalue decomposition may transform the graph Laplacian into the spectral domain.

[0077] In some embodiments, the system 100 may use sparse matrix operations, dimensionality reduction techniques, and / or other suitable operations to increase the efficiency of the spectral graph process.

[0078] In some embodiments, the system 100 may use hierarchical embedding techniques to represent process components as embeddings in multiple layers. The system 100 may decompose the business process into sub-processes and represent the sub-processes using hierarchicalPCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC embeddings. As described above, the system 100 may use spectral graph theory to identify the sub-processes. The hierarchical embedding may include any suitable embeddings associated with the business process, such as activity-level embedding, sub-process-level embeddings, processlevel embeddings, and / or organizational-level embeddings.

[0079] In some embodiments, the hierarchical embedding techniques may include representing each sub-process as an adjacency matrix, applying graph embedding techniques to create vector representations of the sub-processes, and / or combining the vector representations using weighted sums (e.g., based on process precedence). The hierarchical embedding techniques may use any suitable graph embedding technique, such as Node2vec, GraphSAGE, Metapath2vec, and / or Deep Walk.

[0080] In some embodiments, the hierarchical embedding process may include enhancing embeddings by incorporating metadata into the embeddings to include contextual information, such as process duration, process duration variability, process frequency, task precedence levels, resource usage levels, timestamps, and / or performance metrics (e.g., error rates, success rates).

[0081] In some embodiments, the system 100 may process each sub-process individually and then integrate each sub-process into an overall process map. For example, the system 100 may optimize connectivity graphs for each sub-process individually before generating the overall process map. The system 100 may process the sub-processes in a distributed manner (e.g., using cloud-based computing clusters), thus reducing computational processing time for optimization. In some embodiments, the system 100 may process hierarchical embeddings individually and then using the embeddings to generate the overall process map. Operation 206 may include generating, based on the process map, at least one alternative process map. The system 100 (e.g., the process innovator 116) may use a second machine learning model to generate the alternative process map. The second machine learning model may include any of the exemplary models described herein. For example, the first and second machine learning model may each include a trained neural network. In some examples, the second machine learning model may be implemented as a GAN, such as a conditional GAN. In some examples, the second machine learning model may be implemented as a graph variational autoencoder. In this manner, the first and second machine learning model may be implemented as competing GANs in a larger generative adversarial framework.PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC

[0082] The second machine learning model may be deployed at any suitable system and / or processing resource, such as any of the systems and / or resources described herein. In some embodiments, the second machine learning model may be deployed in parallel with the first machine learning model (e.g., at a second node of a cloud-based processing cluster deploying the first machine learning model). The second machine learning model may generate the alternative process map using any of the techniques described herein, such as the techniques described for the first machine learning model.

[0083] The second machine learning model may generate the alternative process map as a proposed alternative to the process map generated by the first machine learning model. The second machine learning model may extrapolate from the process map to generate the alternative process map. That is, the second machine learning model may alter various process parameters and monitor the resulting changes to the workflow.

[0084] In some embodiments, the second machine learning model may be trained to optimize process maps according to various business objectives, such as resource and / or time efficiency. The second machine learning model may use an objective optimization function to optimize process maps. For example, the objective optimization function may be a multi -objective function that accounts for various business objectives metrics, resource utilization parameters, and / or process efficiency metrics. In some examples, the multi-objective function may use multi - objective genetic algorithms that evaluate and select process maps based on a fitness function. For example, the fitness function may be expressed as the following equation:Fitness M) = S?=IP£(0£(M)), where M represents a particular process map, ptis a weighting function that balances the importance of each objective Oi, and n is the number of business objectives considered.

[0085] In some embodiments, the optimization may use a Pareto front approach to identify Pareto-optimal solutions for competing business objectives. The optimization may calculate tradeoff metrics between the competing objectives. The Pareto-based optimization may be expressed as the following expression:PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC where x is the solution vector and F represents a feasible set satisfying all constraints. This may ensure that certain objectives (e.g., process resilience and / or adaptability) are not unduly burdened for the sake of optimizing for competing objectives (e.g., process efficiency).

[0086] In some embodiments, the second machine learning model may be trained to calculate process alignment scores for the generated alternative process maps. Each process alignment score may indicate the level of alignment or deviation between an alternative process map and another process map, such as the process map generated by the first machine learning model. The second machine learning model may calculate the process alignment scores by generating connectivity matrices, performing eigenvector decomposition on the matrices, comparing eigenvectors, and then normalizing the comparison to produce a score between 0 and 1.

[0087] In some embodiments, the first and second machine learning models may use graph regularization to ensure that learned features do not deviate excessively from each other within the generated process maps. For example, applying regularization may ensure that closely connected nodes of a connectivity graph do not exhibit large disparities in attributes and / or flow characteristics. This may prevent generation of unrealistic and / or impractical process maps.

[0088] In some embodiments, the models may communicate with each other to apply graph regularization to the generated process maps. For example, the models may share learned features with each other. The communication may be formalized as a feature exchange where features from the first machine learning model are combined with features from the second machine learning model.

[0089] Operation 208 may include evaluating the generated process maps using at least one business objective.

[0090] The system 100 (e.g., the evaluator 118) may evaluate the generated process maps using an evaluative machine learning model configured to classify the generated process maps. The evaluative model may include any of the models described herein. The objective of the evaluative model may be expressed as a classification problem over the generated process maps, with an objective to correctly classify process maps as real (e.g., based on actual data) or generated (e.g., generated by the first and second machine learning models). The classification problem may incorporate various performance metrics, such as resource utilization levels, business objective achievement levels, and / or stakeholder satisfaction scores.PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC

[0091] The classification problem may be expressed as a decision function that aims to maximize the likelihood of a correct classification. For example, the decision function may be expressed as the following expression: maxDEx~Pdata(x)[log / )(%)] + Ex~^(x)[log(l - £>(%)] , where p5(x) is the distribution of the generated processes.

[0092] In some embodiments, the evaluative model may evaluate the generated process maps by extracting and comparing features from the process maps. For example, the evaluative model may compute the minimum number of changes needed to transform one process map into another and / or measure overlaps in features between two process maps. In some examples, the evaluative model may use similarity searches using graph embedding comparisons.

[0093] In some embodiments, the evaluative model may calculate alignment scores that indicates how well a generated process map aligns with a reference process map. For example, the evaluative model may use eigenvector decompositions on connectivity matrices for a generated process map and a reference process map and compare the eigenvectors to assess process deviation. In some embodiments, the evaluative model may use sensitivity metrics for the alignment scores.

[0094] In some embodiments, the evaluative model may evaluate the generated process maps using contribution scores corresponding to sub-entities (e.g., sub-processes, actors, steps, etc.) of the generated process maps. Each contribution score may quantify the impact of a particular subentity in achieving business objectives and / or user preferences. The contribution scores may be influenced by multiple factors, such as sub-entity importance, resource utilization, performance impact, and / or alignment with business objectives and / or user preferences.

[0095] In some embodiments, each contribution score C may be expressed as the following equation:C = W * R, where W i s the context weight matrix and R is the resource vector. The context weight matrix may represent the importance of each process in achieving objectives for a sub-entity. W may be expressed as the following:PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PCW = P * 0, where P is the process matrix and 0 is the objective vector for a sub -entity.

[0096] In some embodiments, the evaluative model may use sensitivity metrics for the contribution scores. The sensitivity of the contribution score may be expressed as: dC / dP and dC / dR, where P is the process matrix and R is the resource vector.

[0097] In some embodiments, the system 100 may evaluate generated process maps that are stored at a data lake, such as the data lake 124. The system 100 may query the data lake to retrieve process maps that satisfy certain user preferences and / or business objectives. The system 100 may then evaluate the retrieved process maps. The system 100 may store the generated process maps and associated queries at the data lake 124 and / or any other suitable storage resource.

[0098] Operation 210 may include selecting an improved process map from the generated process maps based on the evaluation. The system 100 (e.g., the decision maker 120) may select the improved process map based on any of the features determined by the evaluator, such as alignment scores, evaluated performance metrics, classification outcomes, implementation feasibility, and / or contribution scores.

[0099] Operation 212 may include presenting, at an interactive user interface, at least one of the generated process maps, the at least one generated process map including the improved process map. As described in more detail herein, the interactive user interface may present any number of indicators related to the generated process maps, such as performance metrics, process map comparisons, impact projections, customization options, and / or simulation capabilities. For example, the interactive user interface may present side-by-side visualizations of the improved process map and alternative process maps.

[0100] Operation 214 may include receiving, at the interactive user interface, a user selection indicating a preferred process map from the at least one generated process map.

[0101] In some embodiments, the user selection may indicate at least one user preference for implementing the preferred process map. For example, the user interface may indicate at least one preferred process parameter, such as implementation constraints and / or resource allocationPCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC preferences. For example, the user selection may indicate a requirement to use a particular resource (e.g., a packaging machine) for implementing the preferred process map.

[0102] Operation 216 may include implementing the preferred process map. The system 100 (e.g., the implementor 122) may implement the preferred process map in a manner that fulfills any related user preferences, business objectives, and / or constraints.

[0103] In some embodiments, implementing the preferred process map may include automatically generating an implementation plan for the preferred process map. The implementation plan may ensure that the implementation meets business obj ectives and aligns with user preferences. For example, the system 100 may use the implementation plan to identify any required resource adjustments, generate training requirements, establish monitoring protocols, and / or create a deployment schedule for the preferred process map.

[0104] In some embodiments, implementing the preferred processing map may include deploying the preferred process map. The system 100 may perform any aspect of the deployment of the preferred process map. For example, the system 100 may modify existing workflows to align with the preferred process map, reallocate resources based on the implementation plan, reconfigure systems (e.g., databases, integrations with other systems, software configurations, etc.) to support the modified workflows, and / or perform stakeholder notification protocols (e.g., to inform stakeholders about the deployment).

[0105] While the method 200 is shown as including operations 202-16, it is to be appreciated that method 200 may include any number of additional and / or alternative operations. For example, in some embodiments, the method 200 may include monitoring the implementation of the preferred process map. The system 100 may determine one or more performance metrics of the implementation, such as throughput rates, cycle times, resources utilization rates, process compliance metrics (e.g., error rates, adherence to regulatory standards, etc.), business objective achievement scores, and / or business impact measurements (e.g., customer satisfaction scores, revenue growth rates, etc.). The performance metrics may indicate how well the implementation aligns with business objectives and / or user preferences. The system 100 may display the performance metrics and / or other suitable indicators at the interactive user interface.

[0106] FIG. 3 depicts a flowchart of a workflow 300 for improving business processes using adversarial agents in accordance with one embodiment. While FIG. 3 shows illustrative operationsPCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC of the workflow 300 according to one embodiment, other embodiments may omit, add to, reorder, and / or modify any of the operations of the workflow 300. As illustrated, the workflow 300 may be implemented by one or more components of the system 100 as illustrated in FIG. 1. However, it is to be understood that the workflow 300 may be implemented by any suitable component and / or system.

[0107] Operation 302 may include accessing business process data at the data lake 124. It is to be appreciated, however, that the business process data may be accessed from any suitable source, such as the entities 106 as shown in FIG. 1.

[0108] Operation 304 may include inputting the business process data at the process mapper 114.

[0109] Operation 306 may include generating a process map at the process mapper 114. The process mapper 114 may generate the process map using a first machine learning model trained to generate process maps.

[0110] Operation 308 may include inputting the process map at the process innovator 116.

[0111] Operation 310 may include generating at least one alternative process map at the process innovator 116. The process mapper 114 may generate the process map using a second machine learning model trained to generate alternative process maps.

[0112] Operation 312 may include inputting the generated process map at the evaluator 118.

[0113] Operation 314 may include evaluating the generated alternative process maps at the evaluator 118.

[0114] Operation 316 may include selecting a preferred process map at the decision maker 120.

[0115] Operation 318 may include implementing the preferred process map at the implementor 122.

[0116] FIG. 4 depicts a connectivity graph 400 for a business process in accordance with one embodiment. While FIG. 4 shows illustrative components of the connectivity graph 400 according to one embodiment, other embodiments may omit, add to, reorder, and / or modify any aspect of the connectivity graph 400. As described above, the connectivity graph 400 may be generated byPCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC one or more components of the system 100 as illustrated in FIG. 1. However, it is to be understood that the connectivity graph 400 may be generated by any suitable component and / or system.

[0117] As described above, the connectivity graph 400 may correspond to a suitable process map for a business process. As shown, the connectivity graph 400 may include at least one node 402 (e.g., nodes 402-1 to 402-5) and at least one edge 404 (e.g., edges 404-1 to 404-5). The nodes 402 may represent process steps (e.g., activities, events, decisions, etc.) and edges 404 may represent the flow of information, materials, and / or decisions. For example, the node 402-1 may represent receipt of a customer order and the edge 404-1 may represent the flow of data related to the customer order, such as an identifier for an ordered good.

[0118] In some embodiments, the nodes 402 may include at least one decision node, such as node 402-2. The decision node may indicate a conditional branch in the corresponding business process. For example, the node 402-2 may represent the condition “Is the ordered good in stock?” Accordingly, the edges 404-2 and 404-3 may correspond to the possible outcomes of the condition, such as “no” and “yes.”

[0119] FIG. 5 depicts an implementation of a user interface 500 in accordance with one embodiment. The user interface 500 may be similar to any of the interfaces described herein. The user interface 500 may be interactively presented to a user (e.g., the user 104), such as an administrator tasked with managing business processes for a business and / or any other suitable organization. The user interface 500 may be rendered by any suitable service and / or application (e.g., a web browser) deployed at a suitable device, such as the user device 102.

[0120] The user interface 500 may present any suitable content related to managing business processes. For example, the user interface 500 may present a visual comparison of a plurality of process maps. As shown, for example, the user interface 500 may present a side-by-side comparison of process A and process B . The user interface 500 may present contextual information associated with each presented process map. For example, the user interface 500 may present narrative descriptions for the presented process maps. The narrative descriptions may describe various aspects of the presented process maps, such as high-level overviews, key steps and / or events, related actors and / or roles, inputs and / or outputs, decision points, timelines, recommended implementations and / or enhancements, and / or related compliance rules.PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC

[0121] In some embodiments, the user interface 500 may present at least one contextual level 502 (e.g., contextual levels 502-1 to 502-4). The contextual levels 502 may indicate various levels for contextual information associated with the presented process maps. For example, as shown, the contextual levels 502 may include contextual levels 502-1 and 502-2 that indicate a cost savings level for the process map A, and contextual levels 502-3 and 502-4 that indicate a resource usage level for the process map B. As shown, the contextual levels 502 may indicate a level from 1 to 5, with 1 being the lowest level and 5 being the highest level. However, it is to be appreciated that the contextual levels 502 may indicate the relevant levels in any suitable manner, such as a percentage and / or a color-coded alert.

[0122] In some embodiments, the user interface 500 may present at least one selectable option 504 (e.g., selectable options 504-1 to 504-4). The user may select a selectable option 504 to initiate a process associated with a particular process map. The user interface 500 may, in response to the selection, initiate the process. For example, the process may include at least one of presenting additional information, presenting a connectivity graph for a selected process map, implementing a selected process map, selecting a different process map for comparison, providing user feedback for implemented process maps, and / or providing input for generating and / or modifying process maps (e.g., business objectives and / or user preferences). For example, the user interface 500 may, in response to detecting a selection of selectable option 504-1, may initiate an automated implementation of process A. The automated implementation may include the user interface 500 presenting selectable options for customizing implementation parameters, constraints, and / or user preferences.

[0123] In some embodiments, the system 100 (and / or any other suitable system) may improve business processes using a unified model. The system 100 may improve the processes with and / or without using adversarial agents and / or other machine learning models. To illustrate, FIG. 6 depicts a flowchart of a method 600 for improving business processes using a unified model in accordance with one embodiment. While FIG. 6 shows illustrative operations according to one embodiment, other embodiments may omit, add to, reorder, and / or modify any of the operations shown in FIG. 6. As described, the method 600 may be implemented by one or more components of the system 100 as illustrated in FIG. 1. However, it is to be understood that the method 600 may be implemented by any suitable component and / or system.PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC

[0124] Operation 602 may include receiving business process data. The business process data may include any of the data described herein. The business process data may relate to at least one entity, which may be composed of at least one sub-entity.

[0125] In some embodiments, the business process data may be represented using data structures and / or other suitable representations. For example, the business process data include at least one matrix, such as entity matrices, entity structure matrices, objective vectors, and / or resources vectors. In some embodiments, a machine learning model (e.g., a graph neural network or a graph transformer) may encode the business process data, such as by encoding the data as matrices.

[0126] Operation 604 may include building process matrices for the sub-entities. The process matrices may represent the processes carried out within the sub-entities. The processes may include ongoing processes that the entities are currently implementing and / or standard processes that define ideal processes for serving as benchmarks to evaluate the ongoing processes. The standard processes may be generated using a machine learning model, such as a graph neural network.

[0127] In some embodiments, the process matrices may be represented as directed graphs. The nodes may represent individual processes and / or process steps. The edges may represent dependencies between processes and / or process steps.

[0128] In some embodiments, the relationships between sub-entities may be represented by a relations graph, where the edges of the relations graph define how processes and resources in one sub-entity affect other sub-entities.

[0129] Operation 606 may include computing context weight matrices for each sub-entity. The system 100 may compute the context weight matrices based on the business process data and / or process matrices. For example, the context weight matrix may be expressed as the following:Wi = Pi * Oi, where Ptis the process matrix and Otis the objective vector for entity e^.

[0130] Operation 608 may include comparing ongoing and standard processes. The processes may be represented using connectivity matrices. Ctmay represent the connectivity matrix for thePCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC ongoing processes for sub-entity et. Citdmay represent the connectivity matrix for the standard processes for e£.

[0131] In some embodiments, the system 100 may decompose the connectivity matrices using eigenvector analysis. For example, for ongoing process maps:where vtis the eigenvector and is the eigenvalue. A similar decomposition may be performed for the standard process maps using v-tdand Atd.

[0132] In some embodiments, the system 100 may calculate process deviations for the ongoing processes as compared to the standard processes. For Ctthe deviation Av, may be expressed as:

[0133] In some embodiments, the system 100 may calculate process alignment scores, which may be defined as:where a score of 1 indicates perfect alignment and a score of 0 indicates complete misalignment.

[0134] Operation 610 may include calculating alignment measures and contribution scores. The alignment measure A may assess how well the processes of the sub-entities align with the overall objectives of the entity. The alignment measure A may be defined as:A = f (O, O O2, ... , On, PltP2, ... , Pn), where O represents the entity objectives, and 0, and P, represent the objectives and processes of sub-entities e£, respectively.

[0135] In some embodiments, the alignment formula may be modified to consider external factors. Examples of external factors may include seasonal demand fluctuations, geopolitical events, supply chain volatility, changing customer preferences, economic indicators, and / or industry-specific compliance requirements. The system 100 may introduce a context matrix C that represents at least one external factor. The system 100 may calculate a weighted alignment basedPCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC on the context matrix C. For example, during a supply chain disruption of a certain resource, the system may dynamically decrease the alignment measure of an entity that consumes the resource.

[0136] The contribution score may measure the contribution of each sub-entity etto the overall performance of the entity. The contribution score may be defined as:where W(is the context weight matrix and Rtis the resource vector.

[0137] Operation 612 may include performing sensitivity analysis on the alignment measures and contribution scores. The system 100 may calculate alignment sensitivity and / or contribution sensitivity to measure how changes in processes affect the alignment measures and / or contribution scores, respectively.

[0138] The alignment sensitivity may be defined as: dA dPtjk

[0139] The contribution sensitivity to changes in processes and resources may be defined as: dCt, dCi- — and - — . dPijk 3Ra

[0140] In some embodiments, the alignment sensitivity may be weighted to adjust for the priority and / or impact of certain objectives over others. For example, a weighted alignment sensitivity Awmay be expressed as:

[0141] Operation 614 may include optimizing the performance of the entity. This optimization may be expressed as an optimization problem that aims to maximize or minimize an objective function. For example, the objective function may be: maxM(

[0142] This optimization may be subject to process capacity constraints and / or resource constraints.PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC

[0143] The system 100 may use a variety of optimization algorithms to solve the optimization problem, such as linear programming, quadratic programming, and / or evolutionary methods such as multi -objective genetic algorithms. In some embodiments, the system 100 may use machine learning models to solve the optimization problem.

[0144] In some embodiments, the system 100 may use multi-objective optimization to handle competing goals such as cost reduction, customer satisfaction, throughput, compliance, and / or operational resilience. The system 100 may employ Pareto optimization techniques to identify optimal solutions.

[0145] In some embodiments, the system 100 may use stochastic optimization techniques to allow optimization during uncertain conditions. For example, the system 100 may identify uncertainties in process execution outcomes, resource availability, and / or external factors. The system 100 may model data (e.g., process matrices, resource vectors, etc.) as random variables with associated probability distributions derived from historical variability or forecasted volatility. The system 100 may generate multiple simulated process execution scenarios and compute expected values and / or worst-case bounds for alignment measures and / or contribution scores.

[0146] In some embodiments, the output of optimizing the performance of the entity may be a generated process matrix. The generated matrix may represent a process that optimizes the performance of the entity and is aligned with the overall objectives of the entity. The process matrix may be used by the entity to improve the business performance of the entity. For example, the entity may use the process matrix to adjust ongoing processes, such as by reallocating resources among sub-entities, assessing future mergers and acquisitions, initiating training and / or recruiting efforts, and / or monitoring performance for current employees. The process matrix may be used to identify particular sub-entities that are the most misaligned with entity objectives, so that the identified sub-entities may be targeted for future business adjustments.

[0147] While method 600 is illustrated as including operations 602-614, it is to be appreciated that method 600 may include any number of additional and / or alternative operations. For example, in some embodiments, the system 100 may continuously respond to updated business process data. For example, the system 100 may adjust input matrices such as objective vectors. In some examples, the system 100 may adjust calculated values such as alignment measures and / orPCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC contribution scores. The system 100 may then generate new process matrices for future business adjustments.

Claims

PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PCCLAIMSWhat is claimed is:

1. A method for improving business processes, the method comprising: receiving, at a cloud-based processing cluster, business process data; generating, at the cloud-based processing cluster based on the business process data, a process map representing a business process using a first machine learning model deployed at the cloud-based cluster; generating, based on the process map, at least one alternative process map using a second machine learning model deployed in parallel with the first machine learning model at the cloudbased processing cluster; evaluating the generated process maps using at least one business objective; selecting an improved process map from the generated process maps based on the evaluation; presenting, at an interactive user interface, at least one of the generated process maps, the at least one generated process map including the improved process map; receiving, at the interactive user interface, a user selection indicating a preferred process map from the at least one generated process map; and implementing the preferred process map.

2. The method of claim 1, further comprising: querying at least one data lake using the generated process maps and associated queries; storing the generated process maps and the associated queries in a query storage resource; and evaluating the generated process maps and the associated queries based on user requirements.

3. The method of claim 2, wherein the evaluating of the generated process maps and the associated queries is based on a similarity search.

4. The method of claim 1, wherein the at least one alternative process map is generated from a generative adversarial network.PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC5. The method of claim 4, wherein the generative adversarial network uses a multi -objective optimization function.

6. The method of claim 1, wherein the machine learning model comprises a generative adversarial network, and wherein the process map comprises a connectivity graph generated using the generative adversarial network.

7. The method of claim 6, wherein the machine learning model is configured to use spectral graph theory to improve a layout of the connectivity graph.

8. The method of claim 1, wherein the evaluating of the generated process maps is based on a decision function expressed as a classification problem.

9. The method of claim 1, further comprising generating at least one summary for the improved process map, the at least one summary including a rationale for selecting the improved process map.

10. The method of claim 1, wherein the presenting of at least one of the generated process maps comprises presenting a visual comparison of the improved process map and at least one additional generated process map.

11. The method of claim 1, further comprising receiving, at the interactive user interface, a second user selection indicating at least one implementation parameter for the preferred process map, wherein the implementing the preferred process map is based on the at least one implementation parameter.

12. An apparatus for improving business processes, comprising: an interactive user interface; a memory storing instructions;PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC at least one processor executing the instructions stored on the memory to perform the steps of: receiving business process data; generating, based on the business process data, a process map representing a business process using a machine learning model deployed at a cloud-based processing cluster; generating, based on the process map, at least one alternative process map; evaluating the generated process maps using at least one business objective; selecting an improved process map from the generated process maps based on the evaluation; presenting, using the interactive user interface, at least one of the generated process maps, the at least one generated process map including the improved process map; receiving, using the interactive user interface, a user selection indicating a preferred process map from the at least one generated process map; and implementing the preferred process map.

13. The apparatus of claim 12, wherein the at least one processor is further configured to perform the steps of: querying at least one data lake using the generated process maps and associated queries; storing the generated process maps and the associated queries in a query storage resource; and evaluating the generated process maps and the associated queries based on user requirements received from the user interface.

14. The apparatus of claim 13, wherein the evaluating of the generated process maps and the associated queries is based on a similarity search.

15. The apparatus of claim 12, wherein the at least one alternative process map is generated from a generative adversarial network.PCT / US25 / 44368 30 August 2025 (30.08.2025)DSK-001PC16. The apparatus of claim 15, wherein the generative adversarial network uses a multiobjective optimization function.

17. The apparatus of claim 12, wherein the machine learning model comprises a generative adversarial network, and wherein the process map comprises a connectivity graph generated using the generative adversarial network.

18. The apparatus of claim 17, wherein the machine learning model is configured to use spectral graph theory to improve a layout of the connectivity graph.

19. A system for improving business processes, comprising: a process mapper module for generating a process map representing a business process using a machine learning model deployed at a cloud-based processing cluster; a process innovator module for generating at least one alternative process map; an evaluator module for evaluating the generated process maps using at least one business objective; a decision-making module for selecting an improved process map from the generated process map; and an interactive user interface for: presenting at least one of the generated process maps, the at least one generated process map including the improved process map; and receiving a user selection indicating a preferred process map from the at least one generated process map.

20. The system of claim 19, wherein the at least one alternative process map is generated from a generative adversarial network.

Images