Methods and systems for intelligently managing and using data in a supply chain environment

Abstract

The present disclosure provides systems and methods for querying information related to a supply chain company. The method comprises receiving, via a graphical user interface (GUI), a user prompt in natural language inquiring information related to a supply chain company, generating, using a thought-based query generator, a database query based at least in part on the user prompt, executing the database query in a database to retrieve data and transmitting the data to a large language model (LLM), and generating an output by the LLM and displaying the output in the GUI.

Description

METHODS AND SYSTEMS FOR INTELLIGENTLY MANAGING AND USING DATA IN A SUPPLY CHAIN ENVIRONMENT CROSS-REFERENCE

[0001] This application claims the benefit of U.S. Provisional Application No. 63 / 654,153, filed May 31, 2024, which is incorporated herein by reference in its entirety.BACKGROUND

[0002] The food supply chain industry has lagged behind others in digital transformation, with many companies still relying on outdated systems and processes. For instance, much of the industry uses old software running on closed systems and on-premise servers, resulting in valuable data being trapped. However, in the food supply chain industry, access to real-time and up-to-date information is critical. Buying and selling commodities is akin to trading stocks on Wall Street. When employees can access the right information at the moment they need it, they can secure more deals on more favorable terms. This access to information enables companies to remain competitive and succeed. Many food supply chain users utilize internal naming conventions, leading to custom product names, commodity designations, customer and vendor codes, warehouse identifiers, and other key terms stored within their Enterprise Resource Planning (ERP) systems. Currently, Al-powered tools have struggled to effectively match conversational user prompts to these terms in the ERP database.SUMMARY

[0003] The present disclosure addresses the above needs by providing methods and systems for a supply chain and intelligence analysis platform built on artificial intelligence (Al) and machine learning (ML) models. In particular, the present disclosure provides platform-independent intelligence systems and methods employing Al or ML models allowing for improved information retrieval, advanced data insights, and predictive analytics in a user-friendly, unrestricted, conversational format that is not currently available in the food supply chain industry.

[0004] In an aspect, the present disclosure provides methods and systems for querying information related to a supply chain. In some embodiments, the methods and systems comprise receiving, via a graphical user interface (GUI), a user prompt in natural language inquiring about information related to a supply chain. In some embodiments, the methods and systems comprise generating, using a thought-based query generator, a database query based at least in part on the user prompt. In some embodiments, the methods and systems comprise executing the database query in a database to retrieve data and transmitting the data to a large language model (LLM).In some embodiments, the methods and systems comprise generating an output by the LLM and displaying the output in the GUI.

[0005] In some embodiments, the thought-based query generator comprises a classifier model trained to generate a plurality of thoughts by processing the user prompt and contextual data. In some embodiments, the contextual data comprises a previous conversion inquiring information related to the supply chain. In some embodiments, the thought-based query generator is configured to run the plurality of thoughts concurrently. In some embodiments, the thoughtbased query generator is configured to backtrack to explore different thoughts. In some embodiments, a thought of the plurality of thoughts is a step, equation, a transformation, a model prediction, a rule application, a function call, or a sub-query generation. In some embodiments, a thought of the plurality of thoughts generates a refined input or a candidate lookup query. In some embodiments, the thought-based query generator is configured to evaluate a plurality of refined inputs or a plurality of candidate lookup queries generated by the plurality of thoughts to identify a final input and the database query.

[0006] In some embodiments, evaluating the plurality of candidate lookup queries comprises assigning a score to each candidate lookup query, wherein the score is assigned based at least in part on a relevancy of the retrieved data with respect to the user prompt, or a confidence metric, and wherein the thought-based query generator is configured to refine the inputs until the score is above a predetermined threshold. In some embodiments, the final input is identified based at least in part on performing one or more lookahead trials for each refined input or candidate lookup query. In some embodiments, the database comprises an enterprise resource planning (ERP), a third-party database, a public database, or a custom database. In some embodiments, the user prompt comprises an instruction to generate a visual illustration of the information. In some embodiments, the output comprises the visual illustration and wherein a user is permitted to interact with the visual illustration via the GUI. In some embodiments, the context data further comprises data retrieved from a dictionary comprising one or more entries. In some embodiments, each entry comprises a name in the user prompt and a standard name. In some embodiments, the dictionary is built by a user via the GUI.

[0007] Another aspect of the present disclosure provides a non-transitory computer readable medium comprising machine executable code that, upon execution by one or more computer processors, implements any of the methods above or elsewhere herein.

[0008] Another aspect of the present disclosure provides a system comprising one or more computer processors and computer memory coupled thereto. The computer memory comprisesmachine executable code that, upon execution by the one or more computer processors, implements any of the methods above or elsewhere herein.

[0009] Additional aspects and advantages of the present disclosure will become readily apparent to those skilled in this art from the following detailed description, wherein only illustrative embodiments of the present disclosure are shown and described. As will be realized, the present disclosure is capable of other and different embodiments, and its several details are capable of modifications in various obvious respects, all without departing from the disclosure.Accordingly, the drawings and description are to be regarded as illustrative in nature, and not as restrictive.INCORPORATION BY REFERENCE

[0010] All publications, patents, and patent applications mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent, or patent application was specifically and individually indicated to be incorporated by reference. To the extent publications and patents or patent applications incorporated by reference contradict the disclosure contained in the specification, the specification is intended to supersede and / or take precedence over any such contradictory material.BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The novel features of the invention are set forth with particularity in the appended claims. Abetter understanding of the features and advantages of the present invention will be obtained by reference to the following detailed description that sets forth illustrative embodiments, in which the principles of the invention are utilized, and the accompanying drawings (also “Figure” and “FIG.” herein), of which:

[0012] FIG. 1 schematically shows an example workflow of the system, in accordance with some embodiments of the present disclosure.

[0013] FIG. 2 schematically shows an example of an evolving lookup query algorithm implemented by the smart retriever of the present disclosure.

[0014] FIGs. 3-19 show examples graphical user interface (GUI) and data visualization.

[0015] FIGs. 20-22 show examples of GUIs and dictionary configuration.

[0016] FIG. 23 shows a computer system that is programmed or otherwise configured to process, analyze, represent, or any combination thereof, in accordance with any of the systems or methods provided herein.

[0017] FIG. 24 shows an example of GUI for user feedback.DETAILED DESCRIPTION

[0018] While various embodiments of the invention have been shown and described herein, it will be obvious to those skilled in the art that such embodiments are provided by way of example only. Numerous variations, changes, and substitutions may occur to those skilled in the art without departing from the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed.

[0019] In an aspect, the systems and methods herein may be applied for improving information retrieval in a supply chain environment. The systems and methods herein may provide an intelligence analysis platform built on artificial intelligence (Al) and machine learning (ML) models. This configuration may offer users information retrieval, advanced data insights, predictive analytics, or any combination thereof, in a user-friendly, unrestricted, conversational format that has not been previously available in the supply chain environment.

[0020] In some embodiments, Al and ML models may be employed to analyze vast volumes of datasets from multiple sources within a supply chain environment. The platform may comprise unique algorithm leveraging the Al and ML models to uncover patterns, trends, anomalies, or any combination thereof, that may inform strategic decision-making.

[0021] In some embodiments, algorithms may process historical data, real-time inputs, contextual, or any combination thereof to generate actionable insights tailored to the specific needs of users. In some embodiments, the advanced data insights generated by the platform may encompass a wide range of metrics relevant to supply chain operations, including inventory levels, supplier performance, demand forecasting, logistical efficiency or any combination thereof. The insights may be presented in visually intuitive formats, such as charts and graphs, facilitating easy interpretation and analysis by decision-makers.

[0022] While preferred embodiments of the present subject matter have been shown and described herein within the context of supply chain or food supply chain, it will be obvious to those skilled in the art that such embodiments are provided by way of example only and can be applied to various other industries that are not limited to supply chain. Numerous variations, changes, and substitutions may occur to those skilled in the art without departing from the present subject matter. It should be understood that various alternatives to the embodiments of the present subject matter described herein may be employed in practicing the present subject matter.

[0023] Certain definitions

[0024] Unless otherwise defined, all technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0025] Reference throughout this specification to “some embodiments,” or “an embodiment,” means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. Thus, the appearances of the phrase “in some embodiment,” or “in an embodiment,” in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0026] As utilized herein, terms “component,” “system,” “interface,” “unit” and the like are intended to refer to a computer-related entity, hardware, software (e.g., in execution), and / or firmware. For example, a component can be a processor, a process running on a processor, an object, an executable, a program, a storage device, and / or a computer. By way of illustration, an application running on a server and the server can be a component. One or more components can reside within a process, and a component can be localized on one computer and / or distributed between two or more computers.

[0027] Further, these components can execute from various computer readable media having various data structures stored thereon. The components can communicate via local and / or remote processes such as in accordance with a signal having one or more data packets (e.g., data from one component interacting with another component in a local system, distributed system, and / or across a network, e.g., the Internet, a local area network, a wide area network, etc. with other systems via the signal).

[0028] Moreover, the word “exemplary” is used herein to mean serving as an example, instance, or illustration. Any aspect or design described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other aspects or designs. Rather, use of the word exemplary is intended to present concepts in a concrete fashion. As used in this application, the term “or” is intended to mean an inclusive “or” rather than an exclusive “or.” That is, unless specified otherwise, or clear from context, “X employs A or B” is intended to mean any of the natural inclusive permutations. That is, if X employs A; X employs B; or X employs both A and B, then “X employs A or B” is satisfied under any of the foregoing instances. In addition, the articles “a” and “an” as used in this application and the appended claims should generally be construed to mean “one or more” unless specified otherwise or clear from context to be directed to a singular form.

[0029] Whenever the term “at least,” “greater than,” or “greater than or equal to” precedes the first numerical value in a series of two or more numerical values, the term “at least,” “greater than” or “greater than or equal to” applies to each of the numerical values in that series of numerical values. For example, greater than or equal to 1 , 2, or 3 is equivalent to greater than or equal to 1 , greater than or equal to 2, or greater than or equal to 3.

[0030] Whenever the term “no more than,” “less than,” or “less than or equal to” precedes the first numerical value in a series of two or more numerical values, the term “no more than,” “less than,” or “less than or equal to” applies to each of the numerical values in that series of numerical values. For example, less than or equal to 3, 2, or 1 is equivalent to less than or equal to 3, less than or equal to 2, or less than or equal to 1.

[0031] When ranges are present, the ranges include the range endpoints. Additionally, every sub range and value within the range is present as if explicitly written out. The terms “about” and “approximately,” as used herein, when preceding a numerical value indicates the value plus or minus a range of 10%. For example, about 10 maybe reasonably understood to convey 9, 10, or 11, or a range of numerical values spanning from 9 to 11. Whenever “about” or “approximately” precedes the first numerical value in a series of two or more numerical values, the term “about” or “approximately” applies to each of the numerical values in that series of numerical values.

[0032] The present disclosure provides systems, devices, platforms, methods, and techniques for a supply chain and intelligence analysis platform built on artificial intelligence (Al) and machine learning (ML) models. In some embodiments, the systems and methods herein may utilize a cloud-based architecture that may ensure accessibility from various computer devices, including desktops, tablets, smartphones or a combination thereof. This configuration may allow users to access real-time insights anytime and / or anywhere. This accessibility may ensure that users may engage with the systems and methods herein from various devices, enhancing usability.

[0033] In some embodiments, the systems and methods herein may comprise seamless integration capabilities with Enterprise Resource Planning (ERP) and accounting platforms. The systems and methods herein maybe fully compatible with various third-party systems, such as WholesaleWare, Famous, PICS, Thyme, Granite State Software, or QuickBooks. In some embodiments, the integration may allow for real-time data synchronization between the system herein and the aforementioned ERP and accounting platforms. This capability may ensure that users have access to the most current information, facilitating informed decision-making based on up-to-date financial and operational data. In some embodiments, the systems and methodsherein may employ application programming interfaces (APIs) to facilitate communication between the Al and external platforms. The APIs may allow for the extraction and transfer of data between systems, beneficially providing the seamless flow of information. The API may allow for automated data updates, reducing the need for manual data entry and minimizing the risk of errors. In some embodiments, data migration solutions may be provided for any other ERP and accounting systems employing any other or future technologies. The systems and methods herein may comprise tools and processes to facilitate the migration of historical data from legacy systems to the new environment. This capability may ensure that users can retain valuable historical insights while transitioning to the Al-enabled platform.

[0034] In some embodiments, the configuration may support customizable mapping of data fields between the Al-platform herein and integrated ERP or accounting systems. This mapping may allow users to define how data elements from their existing systems correspond to those utilized by the Al-platform. By establishing clear relationships between data points, users may ensure that information is accurately represented and analyzed. In some embodiments, the integration may also encompass features for data validation and cleansing. Before data is migrated or synchronized, the system may conduct checks to identify inconsistencies or errors in the data. This capability may enhance the quality of the data being used, providing users with more reliable insights. In some embodiments, the systems and methods herein may feature user- friendly interfaces that guide users through the integration and migration processes. Step-by-step wizards may be provided to assist users in configuring settings, mapping data fields, and monitoring progress during data transfers. In some embodiments, the seamless integration capabilities may promote interoperability across various supply chain functions. Users may access insights derived from financial, operational, and customer data consolidated from multiple systems, facilitating a holistic view of supply chain performance.

[0035] In some embodiments, the predictive analytics functionality may provide users with foresight into potential future scenarios based on historical data trends. The Al and ML models may generate forecasts regarding demand fluctuations, pricing changes, supply chain disruptions, or any combination thereof. In some embodiments, the conversational format of the platform may enhance user engagement by allowing for interactive dialogue with the system. Users may pose questions, request clarifications, and explore data iteratively, leading to a more dynamic and responsive analytical experience. The configuration may support multi-turn conversations, enabling users to delve deeper into specific topics without losing context.

[0036] The systems or methods described herein may beneficially allow users to query any piece of information or access insights through a user interface by inputting a question in natural language. The system may generate answers to the user's question, along with one or more AI- suggested drill-down questions. For example, a user may input a question via the chat interface and may receive an answer with an option to download the response or insight, along with a list of suggested follow-up questions. This question-and-answer user interface may significantly save time on searching for answers and may assist in discovering insights from unsorted data, both structured and unstructured.

[0037] In some embodiments, an information query related to a supply chain may comprise a retrieval of relevant data and / or presenting it in a user-friendly format. In some embodiments, the method may comprise receiving, via a graphical user interface (GUI), a user prompt expressed in natural language. This prompt may inquire about specific information related to the supply chain, such as inventory levels, order statuses, or supplier performance metrics. The GUI may be designed to facilitate easy input, enabling users to communicate their queries without the need for technical jargon or complex query languages.

[0038] In some embodiments, the systems and methods herein may comprise a smart retriever component configured to generate a database query to retrieve relevant supply chain data based on a user inquiry. In some cases, the smart retriever may comprise a thought-based query generator. The generator may analyze the user prompt to identify key terms and concepts, interpret some of the terms to predefined custom terms based at least in part on an Enterprise Dictionary, details of which will be described later herein. The smart retriever component may employ a unique thought-based query generation algorithm that may allow the system to break down the user prompt into structured reasoning steps, where each thought serves as a mini-agent that contributes to formulating the final database query. The smart retriever component may leverage various sources, including classifiers and rules-based engines, to ensure that the resulting query is accurate and contextually relevant. Details about the smart retriever and / or the query generator are described later herein.

[0039] In some embodiments, the generated database query may be executed within a connected database to retrieve the requested data. This execution may involve querying relational databases, NoSQL databases, or any other data storage solutions pertinent to the supply chain. Upon successful execution of the query, the system may transmit the retrieved data to a large language model (LLM) for further processing. The LLM may be capable of understanding complex data structures and generating insights based on the retrieved information. In someembodiments, an output from the LLM may culminate in generating results. This output may include synthesized insights, summaries, or detailed answers to the user’s original query. The generated output may then be displayed in the GUI, allowing the user to view and interact with the information seamlessly. The interface may also provide options for further exploration, enabling users to refine their inquiries or request additional data as needed.

[0040] The database may comprise an enterprise resource planning (ERP) system, a third-party database, a public database, or a custom database. The user prompt may comprise an instruction to generate a visual illustration of the information. The output may comprise the visual illustration, and the user may be permitted to interact with the visual illustration via the graphical user interface (GUI). The contextual data may further comprise data retrieved from a dictionary comprising one or more entries. Each entry may consist of a name present in the user prompt and a standard name. The dictionary may be built by a user via the GUI, allowing for customization and adaptation to specific organizational needs.

[0041] FIG. 1 shows an example of a system 100 in accordance with some embodiments of the present disclosure. In some embodiments, a knowledge base (DB 110) may receive data input via various methods, including application programming interface (API) integration, direct connection to an on-premise database (DB), import of comma-separated values (CSV) files, JSON files, and files in other text formats. In some embodiments, DB 110 may receive data from various sources. For example, the various sources may comprise enterprise resource planning (ERP) data 112 and third-party data 114, which may encompass pricing data, weather data, shipment data, and other relevant information. Custom data 116 of any type may also be received as required by a particular instance of the system as dictated by a user. Additionally, any kind of public data 118 may be received by DB 110. In some embodiments, DB 110 may contain both structured and unstructured data. Knowledge database 110 may communicate with a smart retriever 104 to upload relevant data to smart retriever 104 and receive lookups from smart retriever 104. As detailed further below, the smart retriever 104 may employ Al technology and / or unique query generation algorithm to retrieve information based on limited user input (e.g., inquiry in natural language). In some embodiments, the system may be capable of extracting accurate insights and answers to a user query by providing a smart retriever component. In some embodiments, the smart retriever component may employ a tree of thoughts prompt technique that structures prompts hierarchically. Details about the smart retriever mechanism are described with reference to FIG. 2. The smart retriever 104 may receive usergenerated prompts and prior user conversations 102. After processing the user input 102 andcommunicating with DB 110, the smart retriever 104 may generate a series of lookup instructions that can be utilized by different downstream applications. In some embodiments, the system may then use these instructions to extract the relevant data and pass the data, as well as the original conversation, to large language model (LLM) generator 106. LLM generator 106 may output a response 108 to the query. The response may comprise downloadable file(s) such as Microsoft Excel.

[0042] It should be noted that though EPR database is shown in the example, the system and its various components can be integrated with or coupled to any other systems, databases or data sources that may or may not be related to supply chain. For example, the data sources may include, without limitation, systems, nodes, or devices in a computing network or other systems used by an enterprise, company, customer or client, or other entity, databases of customer or company information. The data sources may include data stored in an unstructured database or format, such as a Hadoop distributed file system (HDFS). The data sources may include data stored by a customer system, such as a customer information system (CIS), a customer relationship management (CRM) system, or a call center system. The data sources may include data stored or managed by an enterprise system, such as a billing system, financial system, supply chain management (SCM) system, asset management system, and / or workforce management system. The data sources may include data stored or managed by operational systems, such as a distributed resource management system (DRMS), document management system (DMS), content management system (CMS), energy management system (EMS), geographic information system (GIS), globalization management system (GMS), and / or supervisory control and data acquisition (SCADA) system.

[0043] Smart retriever and query generator

[0044] In some embodiments, the system may utilize prompt engineering techniques to enhance the effectiveness of user interactions with the system. The technical setup may involve a combination of natural language processing, machine learning algorithms, the Enterprise Dictionary, or a combination thereof, to ensure that user prompts are accurately interpreted and processed. As an example, when a user submits a prompt, the smart retriever may review the input and correlate it with defined custom terms within the dictionary, enabling a contextual understanding that aligns with the specific needs of the organization. This process may be closely connected to a thought-based algorithm as described elsewhere herein, where the smart retriever or query generator generates candidate thoughts that represent structured reasoning steps. Each thought may serve as a mini-agent or subquery that helps decompose the userprompt into actionable components. In some embodiments, the smart retriever may employ classifiers and rules-based engines to evaluate these thoughts concurrently, allowing for backtracking and exploration of alternative interpretations. By integrating prompt engineering with a thought-based approach, the system may generate precise database queries that facilitate efficient data retrieval and actionable insights, ultimately enhancing user experience and operational efficiency.

[0045] FIG. 2 shows a diagram of an embodiment of smart retriever 104. In some embodiments, the smart retriever component may employ a tree of thoughts prompt algorithm that structures prompts hierarchically. As shown in the example, the smart retriever 104 may receive a usergenerated prompt and historical inquiries (e.g., prior conversation) 102 and operate upon it iteratively using natural language processing and other artificial intelligence methodologies to eventually output a final thought and lookup query 208.

[0046] In some embodiments, the term “thought” as utilized herein may refer to a sub-prompt, an input into a model, an engine, or an algorithm as understood in artificial intelligence.

[0047] The smart retriever may comprise a thought generator that generates a plurality of thoughts based on the prompt and prior conversation. The plurality of thoughts may be converted to or translated into a plurality of candidate lookup queries that are executable on the data objects. In some embodiments, the plurality of candidate lookup queries may be evaluated to identify one or more top optimal lookup queries. The one or more top optimal lookup queries may be further refined by promoting correct partial solutions or queries that can be predicted within a few lookahead trials while eliminating impossible partial solutions based on commonsense reasoning, such as "too big" or "too small" evaluations. A thought from the plurality of thoughts may represent a step, equation, transformation, model prediction, rule application, function call, or sub-query generation. Each thought may serve as a distinct component of the reasoning process, contributing to the comprehensive understanding of the user prompt. For instance, a thought may involve applying a specific transformation to the data or generating a sub-query that focuses on a particular subset of information.

[0048] In some embodiments, a thought of the plurality of thoughts may generate a refined input or a candidate lookup query. In some embodiments, refined inputs or candidate queries may provide the foundation upon which the final database query is constructed. The thought-based query generator may be configured to evaluate a plurality of refined inputs or a plurality of candidate lookup queries generated by the plurality of thoughts. This evaluation process may comprise assessing the quality and relevance of each candidate query based on predefinedcriteria, such as precision, contextual alignment, and potential for effective data retrieval. Ultimately, the generator may identify a final input and the corresponding database query that accurately reflects the user’s original intent. In some embodiments, the evaluation process may be performed by an evaluation model that scores or ranks the candidate lookup queries. The evaluation model may be a neural network (e.g., an MLP or another LLM) to score thoughts trained using backpropagation to distinguish between good and bad paths of the tree.

[0049] In some embodiments, the final optimal thoughts or queries output by the smart retriever may be used to retrieve data and insights from the knowledge base and passed to the large language model (LLM) generator to generate the final response, which may include insights or answers / response to a user query. As illustrated in FIG. 2, a “thought” may be converted to a candidate lookup query. When input 102 is received, a classifier model (thought / node) may read the input and create outputs (lookup query / instruction) along with “edges” that connect to different thoughts (child nodes). Each unique classifier model may comprise a mix of rule-based engines and small models that are trained with various kinds of questions and instructions. For example, for a specific type of question, the model may follow a defined set of steps (creating a tree with multiple edges and child nodes); for another type of question, it may follow a different set of steps. The model may run thoughts concurrently and may also backtrack to explore alternative directions. In some embodiments, thoughts 202, 204, and 206 may represent a first iteration of the process (level 201). Thought 206 may be rejected as not being appropriate or relevant. Thought 204 may be considered appropriate and relevant and may be passed on to evolve further. Thought 202 may not be rejected or accepted but may instead be returned (203) for additional processing. At levels 205, the thoughts may evolve and generate refined inputs. Thoughts 306 may be discarded as unviable, while thought 304 may be passed on to the next level as thought 404. Thoughts 406 may then be discarded as unviable. There may be no limit to the number of levels or iterations that the system can process. At level 207, the “final” thought 208 may be determined by the smart retriever model to be sufficient to answer the original prompt, and the resulting lookup query 208 may be executed. This process may ensure that the query generated is optimized and relevant to the user’s initial input.

[0050] In some embodiments, the smart retriever or the thought-based query generator may comprise a classifier model in the tree of thoughts algorithm. The classifier may be trained to generate a plurality of thoughts by taking the user prompt and contextual data as input. In some cases, the classifier may be trained by updating the model's parameters via gradient descent based on a loss computed from the training dataset (e.g., prompt, contextual data and thoughts).As an example of the training process, a dataset (e.g., prompt, contextual data, thoughts domainspecific data, etc.) may be tokenized to match the model, then backpropagation is used to compute gradients of a loss function (e.g., CrossEntropyLoss) with respect to the model's parameters, and then update the parameters to reduce the loss. In some cases, algorithms such as gradient checkpointing and mixed precision training (e.g., fp 16 / bf 16) may be employed to manage memory usage. In some cases, the method may employe Low-Rank Adaptation (LoRA) and PEFT (Parameter Efficient Fine-Tuning) methods to backpropagate through fewer parameters. The classifier model may utilize machine learning techniques to analyze the natural language input provided by the user, enabling it to identify key components and underlying intents within the prompt. Upon receiving the user prompt, the classifier model may first tokenize the input, breaking it down into manageable elements, such as keywords and phrases. This tokenization process may allow the model to assess the semantic meaning of the user’s inquiry in relation to the supply chain context. The classifier may also leverage contextual data, which may include information from the Enterprise Dictionary and historical user interactions, to enhance its understanding of the prompt.

[0051] The classifier model may generate a plurality of thoughts, with each thought representing a potential direction or aspect of the user prompt. These thoughts may serve as structured reasoning steps, guidingthe system in formulating an appropriate database query. Each thought may capture different interpretations of the user’s request, considering various factors such as relevant entities, relationships, and required data attributes. In some embodiments, the evaluation of the plurality of candidate lookup queries may comprise assigning a score to each candidate lookup query. The score may be assigned based at least in part on the relevancy of the retrieved data with respect to the user prompt or a confidence metric. The thought-based query generator may be configured to refine the inputs until the score exceeds a predetermined threshold. The final input may be identified based at least in part on performing one or more lookahead trials for each refined input or candidate lookup query.

[0052] In some embodiments, the thought-based query generator or the algorithm may be configured to run the plurality of thoughts concurrently. This concurrent execution may allow the system to process multiple interpretations of the user prompt simultaneously, enhancing the efficiency of query generation. By evaluating various paths of reasoning at the same time, the system may identify the most relevant thoughts that contribute to formulating the final database query.

[0053] The thought-based query generator may be configured to backtrack to explore different thoughts upon executing the algorithm. This capability may enable the system to revisit previously considered thoughts if new insights or data emerge during processing. Backtracking may allow the generator to refine its approach dynamically, ensuring that no potentially valuable paths are overlooked in the quest for the most accurate and relevant query.

[0054] In some embodiments, the system may employ the thought-based algorithm / lookup to generate final lookup queries (e.g., SQL, NoSQL, or API calls). In some embodiments, the process of thoughts may comprise evaluating the candidate queries using the following exemplary criteria: schema consistency check (e.g., ensuring that the correct tables andfields are utilized in the proposed queries), semantic relevance to user intent (e.g., assessing how well each candidate aligns with the user's expressed needs), basic constraints validation (e.g., verifying that parameters, such as date formats, meet expected criteria).

[0055] In some cases, the smart retriever may assign scores to candidate queries based on various factors such as precision (e.g., evaluating how directly each candidate answers the user's prompt, coverage (e.g,. determining whether the query retrieves all relevant data), confidence metrics (e.g., utilizing model predictions or heuristic rules to gauge the reliability of the candidate queries) or other factors.

[0056] The smart retriever may perform refinement of the query through promotion and elimination. For instance, the algorithm may promote partial solutions that: are correct but incomplete, indicating they can lead to a final solution, can likely reach a complete solution within a few lookahead trials, and the like. The algorithm may eliminate based on: Commonsense filters: Discarding results that are deemed implausible (e.g., result sets that are "too big" or "too small"), Known impossibilities: Identifying and removing queries that contain logical errors (e.g., bad joins or illogical filters) or other factors.

[0057] In some cases, for the promoted partial solutions, the smart retriever may perform Lookahead Trials such as by predicting next refinement steps (lookahead trials) to explore potential continuations, or using predictive search methods, such as beam search or heuristic pruning, to identify likely pathways toward a full, correct query.

[0058] Next, from the refined candidates, the smart retriever may select the top optimal lookup query or set of queries based on the best-scoring queries after the refinement process and queries that successfully pass logical, semantic, and commonsense checks.

[0059] The system herein may then execute the final optimal query or queries against the database and / or knowledge base. The retrieved data may then be passed to the LLM generatorfor: insight synthesis such as by aggregating the data into meaningful insights, and natural language answer generation by producing a user-friendly response that addresses the original prompt.

[0060] FIGs. 3-10 show illustrations of how the system interacts with a user via an Al-powered assistant. FIG. 3 shows what a user would see when interacting with the user interface of the system. In some embodiments, the user may be presented with potential areas of inquiry, such as supply chain analysis and inventory analysis. Additionally, suggested questions may be presented to the user to provide insight into the capabilities of the system and potentially invite the user to train the system to meet their individual needs. The GUI comprises adding selected communications in favorites 310 and / or browsing the chat history 320. The GUI may show some relevant prompts such as business analysis questions 330 and / or inventory analysis questions 340, or ALsuggested questions 350. The user can insert custom questions about the method / or system 360.

[0061] FIG. 4 shows an answer to a query 410 such as “What were my top selling products in 2023?”. In some embodiments, the system may show a text response 420. In some embodiments, the user can download the data in various formats such as a spreadsheet 430. In this embodiment, the screen may also display suggested additional questions 440, known as drill-down questions, that the user may want to ask. The one or more drill-down questions may be generated by an Al model of the system. For example, the one or more drill-down questions may be derived from the user’s expressed interests and / orthe data accessible to the system. The user may be able to copy Al responses and / or generate email with Al response content 450. The user can insert additional custom questions 460. FIG. 5 shows another screen showing suggested questions that the user might want to ask. This feature may enhance user engagement by providing additional avenues for inquiry, allowing users to explore various aspects of their data. FIG. 6 shows the answer to a query such as “Who are my top customers?”. In this embodiment, the screen may present the answer along with further queries that allow the user to refine the original query or gain a better understanding of the data. This capability may educate the user about their industry, business, operations or supply chain and how to analyze it, while also assisting in training the system to better meet the user’s needs.

[0062] FIG. 7 shows that the user has entered an inquiry that was not suggested by the system, such as “What about just for December?”. This illustrates the system’s ability to engage in a back-and-forth dialogue with the user, maintaining context from historical communications, inquiries or prior conversations when responding to new inquiries. FIG. 8 is a screenshotillustrating the system’s ability to receive a variety of different data file types uploaded by the user. Referring back to FIG. 1, this data could come from any platform that the user wants to incorporate into DB 110, including but not limited to custom data 116 andERP data 112. FIG. 9 is a screenshot illustrating the system’s ability to instruct the user regarding data structure and allow the user to map fields for imported data. This feature may streamline the data integration process, ensuring that data is correctly aligned with the existing schema in the system. FIG. 10 is another screenshot illustrating the system’s ability to allow the user to preview data prior to import. This capability may enable users to verify the integrity and relevance of the data before finalizing the import process, reducing the likelihood of errors and enhancing data quality.

[0063] The systems and methods of the present disclosure can be integrated into any third-party platforms or systems i.e., platform-independent. The platform, system and methods herein may be capable of reading and interpreting data from any third-party supply chain software systems / platforms, such as enterprise resource planning (ERP), accounting, bookkeeping, or other relevant applications. In some embodiments, the system herein may provide a variety of integration and / or data import methods. In some embodiments, the integration and / or data import methods may be automatic, requiring little user intervention, or may necessitate manual mapping, as illustrated in FIGs. 8-10. For instance, the system herein may comprise an application programming interface (API) that allows for direct integration with a third-party system via API integration. In some embodiments, the API may be configured to automatically map data fields of the imported data to the data fields of the data model used by the system herein. This capability may streamline the data integration process, reducing the time and effort required to align disparate data sources. In some cases, the API may comprise artificial intelligence algorithms for data ingestion. For instance, the API may provide a file reader that is trained to automatically identify data fields in the connected third-party software and map the identified data fields to predetermined data fields of the system, such as SKU, Item Name, Type, other relevant identifiers, or any combination thereof.

[0064] As shown in FIG. 8, the user can import data based on various categories 810 such as addresses, choose corresponding templates 820 to the data categories, and import each category 830. As shown in FIG. 9, the user can map data by custom Al fields 910, import custom file column 920, identify / choose data type 930, set / view example values 940, or any combination thereof. As shown in FIG. 10, the user can review and validate the import items before starting the import process.

[0065] Data Visualization Tool

[0066] In some embodiments, the data visualization tool may support various visualization formats, such as charts, graphs, heat maps or a combination thereof. Each format may be selected based on the specific nature of the data being analyzed and the preferences of the user. These visualization options may help users interpret complex data sets easily. In some embodiments, the tool may incorporate natural language processing capabilities, allowing users to query data using conversational language. This feature may enhance user interaction and streamline the data exploration process.

[0067] In some embodiments, the platform herein may prioritize data security and privacy. The systems and methods herein may implement robust encryption protocols and / or access controls to protect sensitive supply chain information. In some embodiments, user feedback may be collected to continuously improve the Al algorithms and / or visualization techniques. In some cases, the system may provide a GUI configured to guide a user for providing feedback and the received feedback may be used to improve the Al algorithms. FIG. 24 shows an example of GUI for user feedback. As shown in the example, a user may upvote or downvote a response generated by the system herein via the GUI. In some cases, upon a user selecting downvote, the GUI may prompt a user to provide reason or additional information for the downvote. The information or reason may be utilized to improve the Al algorithms as feedback. The configuration may adapt over time, ensuring that the insights generated remain relevant and valuable. In some embodiments, the systems and methods herein may utilize a modular architecture that may facilitate easy upgrades and maintenance. Each module may be dedicated to a specific function, such as data input, processing, output or a combination thereof. In some embodiments, the modular configuration may allow for the addition of new functionalities without disrupting existing processes. In some embodiments, the processing unit may be equipped with multiple cores that may operate simultaneously to handle tasks in parallel. This configuration may lead to improved throughput and reduced latency in data processing.

[0068] In some embodiments, the memory module of the platform may employ a hierarchical structure that may optimize data access speeds. Different tiers of memory may be utilized, with faster access memory being used for frequently accessed data.

[0069] In some embodiments, the user interface may support various input methods, including touch, voice, traditional keyboard entry, or a combination thereof. This flexibility may ensure that users may interact with the systems and methods herein in a manner that suits their preferences. The systems and methods herein may incorporate encryption techniques to protect sensitive information during processing and storage. In some embodiments, the systems andmethods herein may feature real-time monitoring capabilities that may allow users to track data processing performance. Dashboards may be provided to display key metrics, enabling users to make informed decisions based on current system performance.

[0070] In some embodiments, the configuration may comprise feedback mechanisms that may allow users to report issues or suggest improvements. This feedback may be analyzed by the systems and methods herein to enhance future iterations and performance. In some embodiments, the systems and methods herein may integrate with external data sources, such as cloud storage solutions or enterprise resource planning systems. This integration may enable seamless data exchange and ensure that users have access to the latest information. In some embodiments, the configuration may prioritize user experience by providing customizable settings that may allow users to tailor the systems and methods herein to their specific needs. Preferences for data visualization and reporting may be adjusted according to user requirements. In some embodiments, the systems and methods herein may continuously improve the Al algorithms and / or visualization techniques based on user interactions and feedback. This capability may ensure that the configuration remains relevant and effective in meeting user needs. In some embodiments, the systems and methods herein may comprise a charts feature that may empower users to visualize key performance indicators (KPIs) graphically. This capability may facilitate the interpretation of large volumes of data by decision-makers.

[0071] In some embodiments, the charts feature may allow users to customize visualizations according to their specific needs. Users may modify parameters such as color schemes, data ranges, and axis labels to enhance clarity and relevance. In some embodiments, the charts feature may support real-time data updates, ensuring that visualizations reflect the most current information available. This capability may enable decision-makers to make timely and informed choices based on up-to-date insights. In some embodiments, the charts feature may be integrated with filtering options that may allow users to focus on specific data sets. Users may apply filters based on criteria such as date ranges, product categories, or geographic regions, thereby streamlining the analysis process. In some embodiments, the systems and methods herein may provide interactive charts that may allow users to engage with the data actively. Features such as tooltips, drill-down capabilities, and zoom functionalities may enhance user experience and facilitate deeper insights.

[0072] In some embodiments, the systems and methods herein may comprise functionality for tracking price trends over time. This capability may allow users to analyze fluctuations in pricing, which may inform strategic pricing decisions. In some embodiments, the systems andmethods herein may analyze sales performance of different product categories, sales representatives, or a combination thereof. This analysis may provide insights into which products are performing well and which sales representatives are achieving their targets, enabling users to optimize their sales strategies. In some embodiments, the systemsand methods herein may monitor the relative profitability of different vendors, customers, or a combination thereof. This monitoring may assist users in identifying which vendors provide the best margins and which customers generate the most profit, facilitating informed decisions regarding supplier relationships and customer engagement strategies. In some embodiments, the configuration may support customizable reporting options that may allow users to generate detailed reports based on the aforementioned analyses. Users may tailor reports to focus on specific timeframes, product categories, or sales representatives, enhancing the relevance of the insights provided.

[0073] In some embodiments, the systems and methods herein may visualize these analyses through dynamic charts and graphs, making complex data easier to interpret. This visualization may empower decision-makers to quickly grasp performance metrics and trends. In some embodiments, the systems and methods herein may allow users to generate intuitive visual reports tailored to their specific needs. This capability may beneficially allow users to create customized reports that reflect the unique aspects of their enterprise, organization or any other entity’s data (e.g., operations data, supply chain and inventory data, business process data, manufacturing, etc.). In some embodiments, the visual reports may be accessible via both desktop and mobile platforms. This accessibility may ensure that users may engage with their data on various devices, enhancing convenience and flexibility. In some embodiments, the reporting feature may be useful for managers, analysts, sales teams, and others who require an effective way to understand their enterprise, business and / or operational data at a glance. The configuration may empower users to quickly grasp insights and make informed decisions based on the visualizations presented. In some embodiments, the visual reports may comprise various visualization options, allowing users to select formats that best convey their data narratives. Options may include charts, graphs, and infographics, providing diverse ways to interpret and present information.

[0074] In some embodiments, the systems and methods herein may comprise a charts feature that improves over the traditional supply chain intelligence tools due to its flexibility and ease of use. Unlike conventional solutions that may require extensive setup only to deliver data in fixed formats, the charts feature may adapt to the user's unique requests in real-time. In some embodiments, users may generate visualizations without the need to pre-configure dashboards orperform time-consuming data setup. This capability may streamline the data analysis process, allowing users to focus on interpreting insights rather than managing complex configurations. In some embodiments, the charts feature may facilitate immediate data exploration, enabling users to manipulate parameters and visualize results on-the-fly. This adaptability may enhance user experience and empower decision-makers to respond quickly to changing supply chain needs. In some embodiments, the systems and methods herein may leverage artificial intelligence (Al) to generate the exact queries, datasets, and visual representations needed for each prompt. This capability may be based on the user's input and may occur within seconds, significantly enhancing the efficiency of data analysis. In some embodiments, the Al engine may analyze user prompts to understand context and intent such as using the smart retriever as described above, allowing it to produce tailored results that align with specific user requirements. This rapid response time may empower users to access insights quickly, facilitating informed decisionmaking.

[0075] In some embodiments, the configuration may support a variety of data sources, enabling the Al to pull relevant information seamlessly. This integration may enhance the comprehensiveness of the visual representations generated for the user. In some embodiments, the systems and methods herein may convert vast amounts of data into actionable insights in real-time. The Al-powered charts feature may make it simple for users to generate the exact visuals they need, with no prior setup or technical knowledge required. In some embodiments, the charts feature may be positioned as the most accessible, flexible, and powerful data visualization tool in the food industry. This capability may enable even non-technical users to access complex insights with ease. In some embodiments, the configuration may focus on empowering customers with the right tools, at the right time, to make smarter, faster decisions. By simplifying the data visualization process, the systems and methods herein may enhance overall user engagement and decision-making effectiveness. In some embodiments, the systems and methods herein may comprise Al-driven chart capabilities that utilize generative artificial intelligence (Al), natural language processing (NLP), modern data visualization technologies, or any combination thereof. These capabilities may work together to enhance the user experience and improve data accessibility.

[0076] In some embodiments, generative Al may analyze user inputs to create tailored queries that align with the specific needs of the user. This process may involve understanding the context of the request and generating the appropriate datasets required for visualization. By leveraging large datasets, the generative Al may ensure that the visuals produced are relevantand insightful. In some embodiments, natural language processing may facilitate seamless interaction between users and the systems and methods herein. Users may input queries in natural language, which the NLP component may interpret to identify key elements such as data types, metrics, and desired visualization formats. This capability may eliminate the need for users to have technical knowledge about data querying or coding, making the system highly accessible. In some embodiments, modern data visualization technologies may enable the creation of dynamic and interactive charts that may respond to user inputs in real-time. Users may manipulate data points, apply filters, and explore various visualization options without requiring prior setup. This flexibility may empower users to gain insights efficiently and intuitively. In some embodiments, the integration of these technologies may allow for the generation of complex visualizations, such as multi-dimensional graphs and heat maps, which may convey intricate data relationships. The systems and methods herein may support collaborative features, enabling users to share visualizations and insights with team members, fostering a data-driven culture within organizations.

[0077] In some embodiments, the Al-driven chart capabilities may continuously learn from user interactions, enhancing their performance over time. Feedback loops may be established to refine the accuracy of generated queries and visual outputs, ensuring that the systems and methods herein remain responsive to evolving user needs. In some embodiments, the systems and methods herein may comprise Al-powered visual reports that may transform complex datasets into clear, actionable charts. This capability may provide users with real-time insights into various critical business, operational or supply chain metrics, such as inventory levels, pricing trends, customer demand, or any combination thereof. In some embodiments, the AL powered visual reports may utilize advanced data processing algorithms to analyze large volumes of data efficiently. By applying machine learning techniques, the systems and methods herein may identify patterns and correlations within the data, allowing for the generation of meaningful insights that may drive informed decision-making.

[0078] In some embodiments, the visual reports may focus on inventory management by providing users with real-time visibility into stock levels, turnover rates, reorder points, or any combination thereof. Users may view trends in inventory depletion, enabling them to optimize stock levels and reduce carrying costs. The configuration may also highlight items that are overstocked or underperforming, allowing users to adjust their purchasing strategies accordingly. In some embodiments, the Al-powered visual reports may analyze pricing trends by aggregating historical pricing data, market fluctuations, or a combination thereof. Users mayvisualize pricing patterns over time, helping them to identify optimal pricing strategies and respond proactively to competitive changes. The systems and methods herein may also generate predictive analytics that may forecast future pricing trends based on historical data, empowering users to make strategic pricing decisions. In some embodiments, the visual reports may provide insights into customer demand by analyzing sales data, customer feedback, market conditions or a combination thereof. The systems and methods herein may utilize segmentation techniques to categorize customer preferences and behaviors, allowing users to tailor their offerings effectively. Users may access visualizations that illustrate changes in market supply or demand across different regions, product categories, or customer demographics, facilitating targeted marketing efforts.

[0079] In some embodiments, the visual reports may be designed with user-friendliness in mind, offering customizable dashboardsthat may allow users to select the specific metrics they wish to track. Users may easily manipulate visual elements to display data in formats that best suit their analytical needs, whether through bar charts, line graphs, heat maps, or any combination thereof. In some embodiments, the Al-powered visual reports may support collaboration by enabling users to share insights and visualizations with team members or stakeholders. This capability may foster a data-driven culture within organizations, encouraging cross-functional teams to engage with the data actively and collaboratively. In some embodiments, the configuration may prioritize real-time data integration, ensuring that visual reports reflect the most current information available. The systems and methods herein may connect with various data sources, such as point-of-sale systems, inventory management software, customer relationship management platforms, or any combination thereof, to deliver comprehensive insights that are timely and relevant. In some embodiments, the systems and methods herein may comprise a user-friendly interface that allows users to create charts effortlessly using natural language. This capability may enhance user engagement and accessibility, enabling individuals without technical expertise to generate visualizations effectively.

[0080] In some embodiments, users may input requests in everyday language, which the systems and methods herein may interpret to create the desired charts. For example, a user may simply type a request such as "Show me monthly sales data," and the systems and methods herein may automatically generate a corresponding chart based on that input. In some embodiments, users may optionally specify desired attributes to tailor the chart display. For instance, users may select the chart type, including options such as bar chart, line chart, pie chart, or other visualization formats. This flexibility may allow users to choose the mostappropriate representation for their data, facilitating clearer communication of insights. In some embodiments, users may also specify the time interval for the data displayed in the chart. Options may include daily, weekly, monthly, custom time frames, or any combination thereof. This capability may enable users to focus on specific periods of interest, enhancing their ability to analyze trends and patterns over time. In some embodiments, users may have the option to customize number formatting within the charts. This feature may allow users to present data in formats that align with their preferences or industry standards, such as currency symbols, decimal places, or percentage representations. This level of customization may improve the clarity and relevance of the visualizations.

[0081] In some embodiments, the user-friendly interface may be designed to provide real-time feedback as users create their charts. As users modify their requests or specify attributes, the systems and methods herein may dynamically update the chart display, allowing users to visualize changes immediately. This interactive experience may enhance user satisfaction and facilitate a deeper understanding of the data. In some embodiments, the configuration may support tooltips and guidance prompts within the interface. These features may assist users in navigating the chart creation process, offering suggestions for attribute selection or clarifying available options. By providing contextual help, the systems and methods herein may further enhance usability. In some embodiments, the systems and methods herein may comprise a data visualization storage feature that may allow users to save charts for easy access in the future. This capability may ensure that important visualizations are always at the user's fingertips, facilitating quick retrieval, ongoing analysis, or any combination thereof. In some embodiments, users may organize saved charts into customizable folders or categories, enabling efficient management of visualizations according to their specific projects or areas of interest. This organization may enhance user experience by reducing the time spent searching for previously created charts.

[0082] In some embodiments, the systems and methods herein may support version control for saved charts. Users may have the option to maintain multiple versions of a chart, allowing them to track changes over time and compare different data representations. This feature may be valuable for analyzing trends and making informed decisions based on historical data. In some embodiments, the systems and methods herein may facilitate real-time data updates for saved charts. With a simple prompt, users may refresh saved charts with the latest data, ensuring that their reports remain current and relevant. This capability may eliminate the need for manual updates, thereby enhancing efficiency. In some embodiments, the refresh functionality mayutilize data integration techniques to pull the most recent information from connected data sources. This integration may ensure that the updated charts reflect the latest inventory levels, sales figures, other critical metrics, or any combination thereof. In some embodiments, users may receive notifications or alerts when saved charts are updated with new data. This feature may keep users informed and engaged, prompting them to review the latest insights and make timely decisions.

[0083] In some embodiments, the data visualization storage and real-time data update capabilities may work in conjunction to create a seamless user experience. Users may easily switch between saved visualizations and refreshed data, allowing for dynamic analysis and exploration of insights. In some embodiments, the configuration may prioritize data security within the storage feature, ensuring that saved charts and associated data are protected from unauthorized access. The systems and methods herein may implement encryption and access control measures to safeguard sensitive information. In some embodiments, the systems and methods herein may comprise features for downloadable visuals and / or seamless sharing. Users may easily download charts and graphs as images or any other format such as PDF, facilitating their use in presentations or for sharing across various platforms. This capability may enhance the versatility of visual insights, allowing users to incorporate them into reports, emails, and collaborative documents efficiently.

[0084] In some embodiments, the charts feature may be directly integrated with sharing capabilities, enabling users to quickly send visual insights to supervisors or team members. This integration may streamline communication and ensure that relevant data is readily available to stakeholders, fostering a collaborative environment focused on data-driven decision-making. In some embodiments, the configuration may support multi-platform access, allowing users to access the charting features seamlessly acrossboth desktop and mobile devices. This flexibility may ensure that users can create, view, and share visual reports regardless of their location or device preference, promoting continuous engagement with data insights. In some embodiments, the systems and methods herein may prioritize security and reliability, ensuring that user data is protected with top-tier security protocols. The configuration may incorporate encryption techniques for data in transit and at rest, safeguarding sensitive information from unauthorized access. Additionally, access controls maybe implemented to restrict data visibility based on user roles, further enhancing data security.

[0085] In some embodiments, the charts feature may undergo regular security audits and updates to maintain compliance with industry standards and best practices. This commitment tosecurity may instill confidence in users, assuring them that their data is managed safely and reliably. In some embodiments, the systems and methods herein may facilitate the visualization of critical metrics such as pricing fluctuations, customer demand, inventory levels, or any combination thereof. By presenting these metrics in a clear and accessible manner, users may identify areas for growth, optimize operations, reduce inefficiencies, or any combination thereof. In some embodiments, visualizing pricing fluctuations may enable users to track changes in product prices over time. By analyzing historical pricing data alongside current market trends, users may uncover opportunities to adjust pricing strategies dynamically. This capability may help users remain competitive and maximize profit margins by responding swiftly to market changes. In some embodiments, the visualization of customer demand may provide insights into purchasing behaviors and preferences. Users may analyze trends in sales data, identifying which products are in high demand and which may require additional marketing efforts. This understanding may empower users to tailor their offerings and promotional strategies to align with customer needs, ultimately driving sales growth.

[0086] In some embodiments, monitoring inventory levels through visualizations may help users maintain optimal stock levels. By visualizing key metrics related to inventory turnover and stockouts, users may make informed decisions regarding restocking and inventory management. This capability may reduce carrying costs and minimize the risk of overstocking or stockouts, enhancing overall operational efficiency. In some embodiments, the systems and methods herein may support the identification of inefficiencies within supply chain operations by visualizing key performance indicators (KPIs). Users may analyze operational metrics, such as order fulfillment times and supply chain performance, to pinpoint bottlenecks and areas for improvement. By addressing these inefficiencies, users may enhance productivity, streamline workflows, or reduce waste of resources. In some embodiments, the configuration may allow users to customize visualizations based on their specific needs and objectives. Users may filter data by time periods, product categories, warehouse or geographic regions, enabling them to focus on the most relevant insights for their decision-making processes.

[0087] In some embodiments, the systems and methods herein may enable users to generate actionable reports that may summarize findings from visualized data. These reports may serve as valuable tools for strategic planning and performance evaluation, ensuring that users can adapt and thrive in a dynamic market environment. In some embodiments, the systems and methods herein may allow food supply chain companies to directly train artificial intelligence (Al) to rapidly learn and adapt to both industry -lev el context and supply chain-specific concepts andterminology. This capability may ensure that the responses generated by the Al are highly relevant and unparalleled in accuracy across diverse enterprises. In some embodiments, the training process may comprise feeding the Al system with datasets that reflect specific operational practices, terminologies, challenges faced by food supply chain companies, or any combination thereof. By utilizing historical data, current market trends, internal supply chain knowledge, or any combination thereof, the Al may develop a comprehensive understanding of the unique aspects of each enterprise. In some embodiments, the configuration may support continuous learning, enabling the Al to update its knowledge base in real-time as new data becomes available. This adaptability may allow the systems and methods herein to respond effectively to changing conditions in the food supply chain, such as fluctuations in demand, supply disruptions, and regulatory changes.

[0088] FIG. 11 shows an example response to a query such as “Show me year’s top 10 selling products in a chart.”. The system can visualize response data in various shapes and format such as a bar graph 1110, and / or text format 1120. The user can insert custom questions regarding the response data or further customize the response data. As an example, upon receiving a user request, the smart retriever may generate a database query as described above based on the user request, and / or context data (e.g., historical conversion), the relevant data maybe retrieved upon executing the query and the LLM may receive the relevant data and make a suitable tool API call to generate the chart.

[0089] FIG. 12 shows another example response to a query such as “Show me last year’s sales by salesperson in a pie chart.”. The system can visualize response data in various shapes and formats such as a pie chart 1210, and / or text format 1220. The user can insert custom questions regarding the response data or further customize the response data. The user can ask the Al to display the results in a chart. The user can specify a desired kind of chart. The user can specify parameters for time intervals / frequency. The user can request formatting changes, e.g. x-axis labels. In some embodiments, at least a portion of data visualization can be done through conversation with the Al. In some embodiments, a chart may be interactive. For instance, values may be displayed in hover states as shown in FIG. 13. In some embodiments, charts can be generated via tool / function calls and API. FIGs. 14-19 show additional response data visualization generated by the Al. These examples show various types of graphs such as a bar graph, a pie chart, or a line graph.

[0090] In some embodiments, the systems and methods herein may comprise an Al model configuration that utilizes a combination of Al reasoning, Al training, indexing, regularexpressions (regex), and other search technologies. This configuration may enhance the system's ability to interpret user queries accurately and retrieve relevant data efficiently. In some embodiments, Al reasoning may enable the model to understand context and intent behind user inputs. By applying logical frameworks and inferential techniques, the Al may derive conclusions that are pertinent to the user’s inquiries. In some embodiments, Al training may involve utilizing labeled datasets to improve the model's performance over time. The training process may include supervised learning techniques, where the model learns from examples, as well as unsupervised learning, where it identifies patterns within unlabelled data. This training may enhance the Al's ability to recognize and process diverse queries effectively.

[0091] In some embodiments, indexing techniques may be employed to organize and optimize the retrieval of information from large datasets. The system may create an index that maps keywords and phrases to specific data points within the knowledge base, enabling rapid access to relevant information. In some embodiments, regular expressions (regex) may be utilized to parse and manipulate text data efficiently. The use of regex may allow the Al to identify specific patterns, validate inputs, and extract essential elements from user queries or datasets. In some embodiments, the Enterprise Dictionary may serve as a comprehensive reference that maps common language terms to technical data representations within the system. This dictionary may facilitate the Al's understanding of industry-specific terminology and user-defined concepts, ensuring that queries are interpreted accurately. In some embodiments, the configuration may allow for continuous updates to the Enterprise Dictionary, enabling users to add or modify entries as needed. This adaptability may ensure that the Al remains aligned with evolving business language and operational requirements.

[0092] In some embodiments, the system may support various data formats for integration, including but not limited to CSV, JSON, XML, proprietary formats, or any combination thereof, used by third-party applications. This flexibility may enhance the system's adaptability to different supply chain environments and data ecosystems. In some embodiments, the system may also facilitate real-time data synchronization with third-party platforms, ensuring that updates in external systems are reflected promptly in the knowledge base. This capability may enable users to access the most current data, enhancing decision-making and operational efficiency. In some cases, the data integration methods may comprise an installed program that directly connects to an on-premise database or data cloud. The program may extract, transform, and send the data from the on-premise database to the database via API connection. In addition to the conventional ETL (extract, transform, load) or ELT (load and transform in datawarehouse) processes, the integration program of the system herein may automatically map data fields of the data in the on-premise database to the data fields of the system as described above.

[0093] In some embodiments, the system may execute workloads, such as queries, within the database or data cloud provider (e.g., AWS S3, Snowflake, Databricks, external data provider, etc.) coupled to the LLM and / or smart retriever and then stream the data and / or insights to the user via the user interface (UI) provided by the system. This approach may enhance efficiency by minimizing data movement and leveraging existing infrastructure. In some cases, the platform may employ buffering techniques, such as caching intermediary results generated during a session of insight or information retrieval. This capability may improve performance and user experience by reducing latency in accessing frequently requested data. In some cases, the data integration methods may comprise manual import of various file types, such as CSV, while allowing the user to control the mapping of fields from the imported file to the database. For instance, a graphical user interface (GUI) may be provided, allowing users to select data fields in the import file and map those data fields to the predetermined data fields (e.g., SKU, Item Name, Type, etc.) of the system. This manual mapping capability may empower users to ensure data accuracy and relevance based on their specific requirements.

[0094] In some embodiments, the systems and methods herein may comprise a Custom Al Model Instruction feature that utilizes system -gen erated prompts with a preset of questions to guide user interactions. This capability may enhance the user experience by providing structured guidance while allowing for flexibility in responses. In some embodiments, the systemgenerated prompt may initiate a conversation with the user by asking specific questions designed to clarify user intent. For example, the systems and methods herein may prompt the user to provide a date definition, such as “last week.” This approach may ensure that the Al has the necessary context to generate accurate instructions based on user input. In some embodiments, after the user responds to the prompt, the Al may process the input and formulate a systemgenerated instruction. For instance, if the user specifies “last week,” the Al may interpret this as a request for a date range that spans 7 days starting from a defined reference day (e.g., the current date). The Al may then generate the corresponding date range and present it to the user in a clear format.

[0095] In some embodiments, the configuration may allow for the customization of preset questions based on user preferences or specific operational needs. Users may define the types of prompts that the system generates, ensuring that the interactions are relevant and tailored to their business requirements. In some embodiments, the Custom Al Model Instruction feature mayalso incorporate a feedback mechanism. Users may provide input on the clarity and usefulness of the prompts, allowing the Al to learn and adapt its questioning strategies over time. This continuous improvement may enhance the effectiveness of user interactions and the overall functionality of the system. In some embodiments, the system may support multi-turn conversations, enabling users to refine their definitions further or provide additional context as needed. This capability may lead to more accurate outputs and a more engaging user experience.

[0096] In some embodiments, the process may commence when a user submits a prompt. The Al may then review the names and terms present in the user prompt alongside the custom defined terms in the Enterprise Dictionary. Following this review, the Al may generate a database query that incorporates the custom defined terms and the original user prompt. This query generation process may ensure that the resulting database query is aligned with the specific terminology and context relevant to the user’s request. This structured flow may facilitate accurate data retrieval and enhance the overall effectiveness of the Al in responding to user inquiries, thereby improving user experience and operational efficiency within the system.

[0097] Enterprise Dictionary

[0098] In some embodiments, the Al may be trained to recognize and utilize industry-specific terminology, developing an industry -specific dictionary, enhancing communication between users and the systems and methods herein. This capability may reduce misunderstandings and improve the relevance of the insights provided, as users may interact with the Al using familiar language and concepts. In some embodiments, the training of the Al may involve collaboration with subject matter experts within the food supply chain companies. These experts may provide insights and / or feedback that may refine the Al's understanding of specific processes, best practices, or strategic objectives. This collaborative approach may enhance the accuracy and applicability of the Al-generated responses. In some embodiments, the systems and methods herein may allow users to customize the training parameters based on their specific needs and / or goals. Users may prioritize certain areas of focus, such as inventory management, pricing strategies, customer engagement, or any combination thereof, ensuring that the Al aligns closely with their strategic objectives. In some embodiments, the configuration may leverage advanced analytics to evaluate the performance of the Al in generating insights and / or recommendations. Users may assess the effectiveness of the Al's responses and make adjustments to the training process as needed, fostering a cycle of continuous improvement.

[0099] In some embodiments, the ability to train the Al specifically for the food supply chain context may result in enhanced decision-making capabilities. Users may receive tailored insightsthat address their unique operational challenges, ultimately leading to improved efficiency, reduced costs, and better alignment with market demands. In some embodiments, the systems and methods herein may comprise Al model configuration capabilities that allow users to easily map common, everyday language used by team members to their supply chain's internal Enterprise Resource Planning (ERP) data within a proprietary Enterprise Dictionary. This feature may enhance the interaction between users and the Al, ensuring that responses are relevant and contextually appropriate. In some embodiments, the proprietary Enterprise Dictionary may serve as a bridge between user input and the underlying data structures within the ERP system. By cataloging terminology commonly used within the organization, the configuration may facilitate a better understanding of user queries and enhance the Al's ability to interpret language accurately. In some embodiments, the Al may adapt how it utilizes generative Al models based on the mappings established in the Enterprise Dictionary. This adaptability may enable the systems and methods herein to respond to prompts in a manner that aligns with the specific language and terminology of the enterprise. As a result, users may receive intelligent insights that are fine-tuned to meet the unique needs of their organization.

[0100] In some embodiments, the configuration may support continuous updates to the Enterprise Dictionary. As supply chain terminology evolves or new terms are introduced, users may have the ability to add or modify entries within the dictionary. This flexibility may ensure that the Al remains aligned with the current language and operational context of the enterprise. In some embodiments, the Al model configuration capabilities may enhance user engagement by providing a more intuitive interface for data interaction. Users may input queries in their natural language, and the systems and methods herein may translate these queries into precise data requests that the ERP system can process effectively. In some embodiments, the intelligent insights generated by the Al may encompass a wide range of supply chain metrics and performance indicators. Users may receive tailored reports that reflect the specific context of their queries, facilitating informed decision-making and strategic planning.

[0101] In some embodiments, the configuration may also support collaborative features, allowing team members to contribute to the development of the Enterprise Dictionary. This collaborative approach may foster a shared understanding of the language used within the organization, enhancing the overall effectiveness of the Al-driven insights. In some embodiments, the systems and methods herein may utilize various technical algorithms to create, train, and utilize a proprietary dictionary that enhances the interaction between users andthe Al. This dictionary may serve as a critical component in facilitating natural language processing and understanding within the system.

[0102] In some embodiments, the dictionary may be created via a graphical user interface (GUI). For example, a user may interact with the GUI to define various mappings and terms in the dictionary. FIGs. 20-22 show examples of Al model configuration and building an Enterprise Dictionary. As shown in FIG. 20, the system may comprise selecting and / or viewing field group 2010, field name 2020, standard name 2023, common name 2040, or any combination thereof. FIG. 21 shows examples of field group 2110 such as items, customers, vendors, sales reps, or location. FIG. 22 shows examples of field name 2210 such as item name, commodity, item category, or item subcategory.

[0103] In optional embodiments, the dictionary may be created using techniques such as data mining, deep learning, machine learning, natural language processing (NLP) or any combination thereof. Data mining algorithms may analyze large datasets sourced from user interactions, industry-specific terminology, and existing documentation. In some cases, the dictionary may be created utilizing text extraction algorithms to gather relevant textual data from various sources, including internal documents, customer communications, and market research reports, utilizing tokenization algorithms to break down the extracted text into individual terms or phrases, enabling the identification of commonly used language within the target domain, performing frequency analysis algorithms to determine the occurrence rates of identified terms, allowing the system to prioritize the most relevant and frequently used words or phrases in the dictionary, or employing contextual mapping algorithms to associate terms with specific meanings based on their usage in various contexts thereby enhancing the dictionary's ability to understand nuanced language.

[0104] In some embodiments, the systems and methods herein may allow organizations to add custom context to the artificial intelligence (Al) model. This capability may ensure that company-specific names, codes, customer identifiers, operational terms, or any combination thereof, are understood and processed correctly by the Al, enhancing its relevance and accuracy in generating insights. In some embodiments, the configuration may comprise a proprietary Enterprise Dictionary that may serve as a reference for mapping custom terms to their corresponding meanings and uses within the organization. The dictionary may comprise entries for specific names, codes, operational jargon, or any combination thereof, that may be unique to the organization’s industry and supply chain processes. In some embodiments, the Al model may utilize natural language processing (NLP) techniques to interpret user inputs that comprisethese custom terms. By applying tokenization, lemmatization, semantic analysis, or any combination thereof, the Al may recognize and contextualize the organization-specific language within user queries. This understanding may enable the Al to generate precise responses and recommendations that align with the organization's operational context.

[0105] In some embodiments, organizations may have the ability to update and maintain the Enterprise Dictionary dynamically. The configuration may support user-friendly interfaces for adding new entries, modifying existing terms, or removing outdated terminology. This flexibility may ensure that the Al model remains current and reflective of any changes in the organization’s operational language. In some embodiments, the systems and methods herein may employ machine learning algorithms to continuously refine the Al model's understanding of custom terms based on user interactions. By analyzing feedback and correction inputs from users, the Al may learn to associate specific terms with appropriate actions or data interpretations. This iterative learning process may enhance the accuracy of the Al's responses over time. In some embodiments, the configuration may support multiple layers of context customization. Organizations may categorize terms based on departments, functions, or specific use cases. For example, terms related to sales, marketing, and logistics may be defined separately, allowing the Al to apply the appropriate context based on the user's query or data input.

[0106] In some embodiments, the custom Al model instruction feature may integrate with existing data management systems to ensure that the Al has access to relevant datasets that correspond with the custom terms. This integration may facilitate real-time data retrieval and processing, allowing the Al to deliver insights that are pertinent to the organization’s specific operational scenarios. In some embodiments, organizations may have the capability to test and validate the customizations made to the Al model. The configuration may provide diagnostic tools that may assess the model's performance in understanding and processing custom terms, ensuring that the Al operates effectively in the organization’s context. In some embodiments, the systems and methodsherein may support communication in natural language, allowing users to interact with the Al by speaking prompts using common names and expressions for products, commodities, customers, vendors, warehouse locations, or other relevant entities. This capability may be accessible through both text-based chat interfaces, hands-free voice mode, or a combination thereof. In some embodiments, the natural language processing (NLP) component may analyze user inputs to identify key terms and phrases. The NLP algorithms may employ techniques such as tokenization, part-of-speech tagging, and named entity recognition to parsethe input, extracting the relevant entities and intents. This analysis may enable the system to accurately interpret user requests in the context of the underlying data.

[0107] In some embodiments, the systems and methods herein may utilize a proprietary Enterprise Dictionary that maps common language terms to technical database queries. The dictionary may comprise custom terms and codes that are specific to the organization’s operational context. When a user inputs a prompt, the system may cross-reference the recognized terms with entries in the Enterprise Dictionary to determine the corresponding technical representations. In some embodiments, the Al may generate database queries based on the interpreted natural language input. This process may involve the construction of structured query language (SQL) statements or other database query formats that align with the organization’s data architecture. The Al may account for various data relationships, filtering criteria, aggregation methods, or any combination thereof, to ensure that the queries return the desired results. In some embodiments, the systems and methods herein may support contextual understanding, allowing the Alto maintain the context of the conversation. This capability may enable users to issue follow-up queries without needing to restate previous information. The Al may leverage context retention algorithms to keep track of user interactions, thus enhancing the fluidity and coherence of the dialogue.

[0108] In some embodiments, the configuration may also incorporate feedback mechanisms that allow the Al to learn from user interactions. By analyzing user corrections or confirmations, the system may refine its understanding of common language terms and improve the accuracy of future translations from natural language to technical queries. In some embodiments, the systems and methods herein may feature a robust error-handling mechanism that may provide users with clarifications or suggestions if the Al encounters ambiguity in the input. For example, if multiple interpretations are possible, the Al may present options for the user to choose from, ensuring that the intended query is accurately executed. In some embodiments, the communication capabilities may be enhanced by utilizing advanced speech recognition technologies. The voice mode may incorporate natural language understanding (NLU) to accurately capture spoken language nuances, enabling users to interact with the system in a conversational manner. In some embodiments, the systems and methods herein may deliver accurate Al responses that are dynamically suited to the specific context of the enterprise. The Al may interpret any inbound prompt by analyzing it through the lens of the company's unique operational environment, culture, or terminology. This contextual understanding may enable theAI to achieve a deeper comprehension of the user's intent, leading to greater accuracy in its responses.

[0109] In some embodiments, the Al may employ a context-aware processing framework that leverages the proprietary Enterprise Dictionary. This framework may allow the Al to recognize and adapt to company-specific terms, jargon, and operational procedures, ensuring that the responses generated are relevant and aligned with the user's expectations. In some embodiments, the Al may utilize advanced natural language processing (NLP) techniques, including semantic analysis and intent recognition, to discern the nuances of user prompts. By identifying the underlying intent behind user queries, the Al may formulate responses that accurately address the specific needs of the user, rather than providing generic or irrelevant information. In some embodiments, the Al may also incorporate machine learning algorithms that may continuously refine its understanding of the enterprise context based on user interactions. Through feedback loops, the Al may learn from past queries and responses, progressively improving its accuracy and relevancy over time. This capability may ensure that the Al adapts to changes in the company, organization, business or operational process, supply chain processes and user preferences.

[0110] In some embodiments, the Al is capable of formulating responses using user-friendly language that is customized for each specific company. This customization may involve generating responses that not only align with the company's terminology but also reflect its communication style and tone. By tailoring language to fit the corporate culture, the Al may enhance user engagement and comprehension. In some embodiments, the systems and methods herein may support multi-turn conversations, allowing the Al to maintain context across multiple interactions. This capability may enable users to engage in more complex discussions without needing to repeat information, as the Al retains the relevant context from previous exchanges. In some embodiments, the configuration may allow for the integration of user- defined parameters that may further enhance response accuracy. For example, users may specify preferences for response formats, levels of detail, or specific areas of focus, enabling the Al to align its outputs with user expectations more closely. In some embodiments, the accuracy and relevance of Al responses may be further validated through user feedback mechanisms. Users may provide ratings or comments on the usefulness of responses, allowing the Al to learn from this feedback and adjust its response strategies accordingly.

[0111] In some embodiments, the Al interprets any inbound prompt based on the company's specific context. The context may be supplied through several mechanisms that enhance the Al'sunderstanding and responsiveness. The Enterprise Dictionary, as described herein, may serve as a comprehensive reference that contains custom defined terms relevant to the organization. This dictionary may enable the Al to interpret industry-specific terminology and internal jargon accurately, ensuring that user prompts are understood within the context of the company's operations. The Al model configuration may comprise preset question and answer pairs received from the customer. These pairs may represent common inquiries, and their corresponding responses based on the company's specific needs. By training the Al with this contextual information, the system may improve its ability to generate relevant answers and insights based on user prompts, aligning responses with the organization's objectives and data. In some embodiments, direct training may be conducted with the customer to tailor the Al model to the specific context of the organization. This training process may involve engaging with stakeholders to gather insights on key business processes, terminology, and use cases. The Al may be fine-tuned based on this training, allowing it to effectively understand and respond to user inquiries that reflect the company's operational context. In some cases, the training process may comprise updating the model's parameters via gradient descent based on a loss computed from a fine-tuning dataset. As an example of the fine-tuning process, a fine-tuning dataset (e.g., contextual data, enterprise-specific data, domain-specific text data, etc.) may be tokenized to match the Al model (e.g., LLM), then backpropagation is used to compute gradients of a loss function (e.g., CrossEntropyLoss) with respect to the model's parameters, and then update the parameters to reduce the loss. In some cases, algorithms such as gradient checkpointing and mixed precision training (e.g., fp 16 / bfl6) may be employed to manage memory usage. In some cases, the method may employe Low-Rank Adaptation (LoRA) and PEFT (Parameter Efficient Fine-Tuning) methods to backpropagate through fewer parameters.

[0112] User Friendly System

[0113] The systems and methods herein may comprise a conversational chat interface designed for retrieving unrestricted and relevant supply chain information. The system may summarize and deliver the final answer directly to the user, in contrast to traditional products that may provide only restricted, predefined sets of data for the user to parse through. This capability may enhance user experience by providing direct access to relevant insights without the need for extensive data exploration.

[0114] The systems and methods herein may allow the user to easily explore the data further and refine and customize the output through the conversational interface. This interactivity mayempower users to engage with the data dynamically, enabling them to ask follow-up questions and adjust their inquiries based on the insights received.

[0115] The systems and methods herein may facilitate the generation of responses and the quick downloading of reports with simple prompts. This streamlined process may differ from traditional tools that require a multitude of steps, selections, and filters. By minimizing the complexity of report generation, the system may enhance operational efficiency and user satisfaction.

[0116] The systems and methods herein may provide a glossary of core industry and supply chain concepts to be utilized for training the Al. These concepts may be customizable for each user, allowing the system to adapt to individual preferences and specific industry terminologies. This feature may improve the relevance and accuracy of Al interactions based on the user's context.

[0117] The systems and methods herein may be available for iOS and Android platforms, incorporating integration with Siri and Google Assistant for voice-activated prompts. This functionality may enable users to interact with the platform hands-free, enhancing accessibility and convenience for users on the go.

[0118] Computer Systems

[0119] The present disclosure provides computer systems that are programmed to implement methods of the disclosure. FIG. 23 shows a computer system 2301 that is programmed or otherwise configured to process, analyze, represent, or any combination thereof, in accordance with any of the systems or methods provided herein. The computer system 2301 can regulate various aspects of the systems or methods as provided herein. The computer system 2301 canbe an electronic device of a user or a computer system that is remotely located with respect to the electronic device. The electronic device can be a mobile electronic device.

[0120] The computer system 2301 includes a central processing unit (CPU, also “processor” and “computer processor” herein) 2305, which canbe a single core or multi core processor, or a plurality of processors for parallel processing. The computer system 2301 also includes memory or memory location 2310 (e.g., random-access memory, read-only memory, flash memory), electronic storage unit 2315 (e.g., hard disk), communication interface 2320 (e.g., network adapter) for communicating with one ormore other systems, andperipheral devices 2325, such as cache, other memory, data storage and / or electronic display adapters. The memory 2310, storage unit 2315, interface 2320 and peripheral devices 2325 are in communication with the CPU 2305 through a communication bus (solid lines), such as a motherboard. The storage unit2315 can be a data storage unit (or data repository) for storing data. The computer system 2301 can be operatively coupled to a computer network (“network”) 2330 with the aid of the communication interface 2320. The network 2330 can be the Internet, an internet and / or extranet, or an intranet and / or extranet that is in communication with the Internet. The network 2330 in some cases is a telecommunication and / or data network. The network 2330 can include one or more computer servers, which can enable distributed computing, such as cloud computing. The network 2330, in some cases with the aid of the computer system 2301, can implement a peer-to-peer network, which may enable devices coupled to the computer system 2301 to behave as a client or a server.

[0121] The CPU 2305 can execute a sequence of machine-readable instructions, which can be embodied in a program or software. The instructions may be storedin a memory location, such as the memory 2310. The instructionscan be directed to the CPU 2305, which can subsequently program or otherwise configure the CPU 2305 to implement methods of the present disclosure. Examples of operations performed by the CPU 2305 can include fetch, decode, execute, and writeback. The CPU 2305 can be part of a circuit, such as an integrated circuit. One or more other components of the system 2301 can be included in the circuit. In some cases, the circuit is an application specific integrated circuit (ASIC).

[0122] The storage unit 2315 can store files, such as drivers, libraries and saved programs. The storage unit 2315 can store user data, e.g., user preferences and user programs. The computer system 2301 in some cases can include one or more additional data storage units that are external to the computer system 2301, such as located on a remote server that is in communication with the computer system 2301 through an intranet or the Internet. The computer system 2301 can communicate with one or more remote computer systems through the network 2330. For instance, the computer system 2301 can communicate with a remote computer system of a user (e.g., a smart phone). Examples of remote computer systems include personal computers (e.g., portable PC), slate or tablet PC’s (e.g., Apple® iPad, Samsung® Galaxy Tab), telephones, Smart phones (e.g., Apple® iPhone, Android-enabled device, Blackberry®), or personal digital assistants. The user can access the computer system 2301 via the network 2330.

[0123] Methods as described herein can be implemented by way of machine (e.g., computer processor) executable code stored on an electronic storage location of the computer system 2301, such as, for example, on the memory 2310 or electronic storage unit 2315. The machineexecutable or machine-readable code can be provided in the form of software. During use, thecode can be executed by the processor 2305. In some cases, the code can be retrieved from the storage unit 2315 and stored on the memory 2310 for ready access by the processor 2305. In some situations, the electronic storage unit 2315 can be precluded, and machine-executable instructions are stored on memory 2310. The code can be pre-compiled and configured for use with a machine having a processer adapted to execute the code or can be compiled during runtime. The code can be supplied in a programming language that can be selected to enable the code to execute in a pre-compiled or as-compiled fashion.

[0124] Aspects of the systems and methods provided herein, such as the computer system 2301, can be embodied in programming. Various aspects of the technology may be thought of as “products” or “articles of manufacture” typically in the form of machine (or processor) executable code and / or associated data that is carried on or embodied in a type of machine- readable medium. Machine-executable code can be stored on an electronic storage unit, such as memory (e.g., read-only memory, random-access memory, flash memory) or a hard disk.“Storage” type media can include any or all of the tangible memory of the computers, processors or the like, or associated modules thereof, such as various semiconductor memories, tape drives, disk drives and the like, which may provide non-transitory storage at any time for the software programming. All or portions of the software may at times be communicated through the Internet or various other telecommunication networks. Such communications, for example, may enable loading of the software from one computer or processor into another, for example, from a management server or host computer into the computer platform of an application server. Thus, another type of media that may bear the software elements includes optical, electrical and electromagnetic waves, such as used across physical interfaces between local devices, through wired and optical landline networks and over various air-links. The physical elements that carry such waves, such as wired or wireless links, optical links or the like, also may be considered as media bearing the software. As used herein, unless restricted to non-transitory, tangible “storage” media, terms such as computer or machine “readable medium” refer to any medium that participates in providing instructions to a processor for execution.

[0125] Hence, a machine-readable medium, such as computer-executable code, may take many forms, including but not limited to, a tangible storage medium, a carrier wave medium or physical transmission medium. Non-volatile storage media include, for example, optical or magnetic disks, such as any of the storage devices in any computer(s) or the like, such as may be used to implement the databases, etc. shown in the drawings. Volatile storage media include dynamic memory, such as main memory of such a computer platform. Tangible transmissionmedia include coaxial cables; copper wire and fiber optics, including the wires that comprise a bus within a computer system. Carrier-wave transmission media may take the form of electric or electromagnetic signals, or acoustic or light waves such as those generated during radio frequency (RF) and infrared (IR) data communications. Common forms of computer-readable media therefore include for example: a floppy disk, a flexible disk, hard disk, magnetic tape, any other magnetic medium, a CD-ROM, DVD or DVD-ROM, any other optical medium, punch cards paper tape, any other physical storage medium with patterns of holes, a RAM, a ROM, a PROM and EPROM, a FLASH-EPROM, any other memory chip or cartridge, a carrier wave transporting data or instructions, cables or links transporting such a carrier wave, or any other medium from which a computer may read programming code and / or data. Many of these forms of computer-readable media may be involved in carrying one or more sequences of one or more instructions to a processor for execution.

[0126] The computer system 2301 can include or be in communication with an electronic display 2335 that can comprise a user interface (UI) 2340 for providing, for example, user input, raw data, processed data, Al analyzed data, Al response, or any combination thereof. Examples of UI’s include, without limitation, a graphical user interface (GUI) and web-based user interface. Methodsand systems of the present disclosure can be implemented by way of one or more algorithms. An algorithm can be implemented by way of software upon execution by the central processing unit 2305. The algorithm can, for example, process user input, analyze inventory, generate Al response, or any combination thereof.

[0127] While preferred embodiments of the present invention have been shown and described herein, it will be obvious to those skilled in the art that such embodiments are provided by way of example only. It is not intended that the invention be limited by the specific examples provided within the specification. While the invention has been described with reference to the aforementioned specification, the descriptions and illustrations of the embodiments herein are notmeantto be construed in a limiting sense. Numerous variations, changes, and substitutions will now occur to those skilled in the art without departing from the invention. Furthermore, it shall be understood that all aspects of the invention are not limited to the specific depictions, configurations or relative proportions set forth herein which depend upon a variety of conditions and variables. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in practicing the invention. It is therefore contemplated that the invention shall also cover any such alternatives, modifications, variations or equivalents. It is intended that the following claims define the scope of the invention and thatmethods and structures within the scope of these claims and their equivalents be covered thereby.

Claims

CLAIMSWHAT IS CLAIMED IS:1 . A method for querying information related to a supply chain, the method comprising:(a) receiving, via a graphical user interface (GUI), a user prompt in natural language inquiring information;(b) generating, using a thought-based query generator, a database query based at least in part on the user prompt;(c) executing the database query in a database to retrieve data and transmitting the data to a large language model (LLM); and(d) generating an output by the LLM and displaying the output in the GUI. The method of claim 1, wherein the thought-based query generator comprises a classifier model trained to generate a plurality of thoughts by processing the user prompt and contextual data.3 The method of claim 2, wherein the information is related to a supply chain company and wherein the contextual data comprises a previous conversion inquiring information related to the supply chain company. The method of claim 2, wherein the thought-based query generator is configured to run the plurality of thoughts concurrently.5 The method of claim 4, wherein the thought-based query generator is configured to backtrack to explore different thoughts.6 The method of claim 2, wherein a thought of the plurality of thoughts is a step, equation, a transformation, a model prediction, a rule application, a function call, or a sub-query generation.7 The method of claim 2, wherein a thought of the plurality of thoughts generates a refined input or a candidate lookup query.8 The method of claim 7, wherein the thought-based query generator is configured to evaluate a plurality of refined inputs or a plurality of candidate lookup queries generated by the plurality of thoughts to identify a final input and the database query.9 The method of claim 8, wherein evaluating the plurality of candidate lookup queries comprises assigning a score to each candidate lookup query, wherein the score is assigned based at least in part on a relevancy of the retrieved data with respect to the user prompt, or a confidence metric, and wherein the thought-based query generator is configured to refine theinputs until the score is above a predetermined threshold.

10. The method of claim 8, wherein the final input is identified based at least in part on performing one or more lookahead trials for each refined input or candidate lookup query.

11. The method of claim 1, wherein the database comprises an enterprise resource planning (ERP), a third-party database, a public database, or a custom database.

12. The method of claim 1, wherein the user prompt comprises an instruction to generate a visual illustration of the information.

13. The method of claim 12, wherein the output comprises the visual illustration and wherein a user is permitted to interact with the visual illustration via the GUI.

14. The method of claim 13, wherein the context data further comprises data retrieved from a dictionary comprising one or more entries.

15. The method of claim 14, wherein each entry comprises a name in the user prompt and a standard name.

16. The method of claim 14, wherein the dictionary is built by a user via the GUI.

17. A system for querying information related to a supply chain, the system comprising at least one processor and instructions executable to cause the at least one processor to perform operations comprising:(a) receiving, via a graphical user interface (GUI), a user prompt in natural language inquiring information;(b) generating, using a thought-based query generator, a database query based at least in part on the user prompt;(c) executing the database query in a database to retrieve data and transmitting the data to a large language model (LLM); and(d) generating an output by the LLM and displaying the output in the GUI.

18. The system of claim 17, wherein the thought-based query generator comprises a classifier model trained to generate a plurality of thoughts by processing the user prompt and contextual data.

19. The system of claim 18, wherein the information is related to a supply chain company and wherein the contextual data comprises historical inquiries about information related to the supply chain company.

20. The system of claim 18, wherein the thought-based query generator is configured to run the plurality of thoughts concurrently.

21. The system of claim 20, wherein the thought-based query generator is configured to backtrack to explore different thoughts.

22. The system of claim 18, wherein a thought of the plurality of thoughts is a step, equation, a transformation, a model prediction, a rule application, a function call, or a sub-query generation.

23. The system of claim 18, wherein a thought of the plurality of thoughts generates a refined input or a candidate lookup query.

24. The system of claim 23, wherein the thought-based query generator is configured to evaluate a plurality of refined inputs or a plurality of candidate lookup queries generated by the plurality of thoughts to identify a final input and the database query.

25. The system of claim 24, wherein evaluating the plurality of candidate lookup queries comprises assigning a score to each candidate lookup query, wherein the score is assigned based at least in part on a relevancy of the retrieved data with respect to the user prompt, or a confidence metric, and wherein the thought-based query generator is configured to refine the inputs until the score is above a predetermined threshold.

26. The system of claim 24, wherein the final input is identified based at least in part on performing one or more lookahead trials for each refined input or candidate lookup query.

27. The system of claim 17, wherein the database comprises an enterprise resource planning (ERP), a third-party database, a public database, or a custom database.

28. The system of claim 17, wherein the user prompt comprises an instruction to generate a visual illustration of the information.

29. The system of claim 28, wherein the output comprises the visual illustration and wherein a user is permitted to interact with the visual illustration via the GUI.

30. The system of claim 29, wherein the context data further comprises data retrieved from a dictionary comprising one or more entries.31 . The system of claim 30, wherein each entry comprises a name in the user prompt and a standard name.

32. The system of claim 30, wherein the dictionary is built by a user via the GUI.

Images