A method and system for a business to solicit a quote from a bank

Abstract

The application relates to a kind of enterprise to bank inquiring method and system, obtain the business type and inquiry application information of enterprise;According to the inquiry application information filled in by enterprise, the mail body containing core inquiry demand is dynamically generated, and the mail with fill report link is sent to bank;After bank side clicks fill report link, according to the business type of enterprise inquiry, corresponding fill report form is called, and bank side fills quote data and submits;Based on the quote data submitted by bank, according to preset rule and big model, comparison and selection analysis is carried out, and scheme recommendation result is generated;Wherein, big model is constructed with enterprise-level financing knowledge base, through multi-agent collaborative decision and scheme deduction, the quotation scheme is generated;Enterprise refers to recommended result, and carries out contract registration.Compared with prior art, the application realizes whole-process online management, reduces manual intervention, improves efficiency, standardization and accuracy through multi-agent collaborative decision, and assists enterprise scientific decision.

Description

Technical Field

[0001] This invention relates to the field of financial data processing technology, and in particular to a method and system for enterprises to solicit quotations from banks. Background Technology

[0002] To meet business needs, companies often need to inquire about pricing from multiple banks during the financing process. Currently, the mainstream operating model still relies on manual, offline communication via non-systematic channels such as telephone and WeChat, with personnel manually compiling and comparing quotes. This model has significant drawbacks: first, it is inefficient, incurring substantial manual communication costs; second, it is prone to data errors, with operational risks in manual recording and comparison; and third, the lack of unified process standards leads to inconsistent pricing criteria, potentially creating compliance risks and difficulties in auditing. This situation urgently requires the standardization, automation, and intelligent management of the financing pricing process through digital means.

[0003] Currently, the process of obtaining financing or deposit quotes for businesses largely relies on manual operations: businesses submit applications offline or through fragmented online tools, banks provide quotes via email or in person, and businesses need to manually compile and compare these quotes. This process is inefficient, error-prone, and lacks standardized procedures. Existing online systems only handle some steps, requiring businesses to manually send emails → banks to manually fill out quotes → businesses to manually compare and select quotes → and manual decision-making. This does not achieve full automation and lacks intelligent comparison capabilities. Summary of the Invention

[0004] The purpose of this invention is to solve the problems existing in the above-mentioned existing systems, and to provide a method and system for enterprises to solicit quotations from banks.

[0005] The objective of this invention can be achieved through the following technical solutions: As a first aspect of the present invention, a method for an enterprise to solicit quotations from a bank is provided, comprising the following steps: Obtain information about the company's business type and inquiry application; Based on the inquiry application information filled in by the enterprise, an email body containing the core inquiry requirements is dynamically generated and sent to the bank with an email containing a link to fill in the form; After the bank clicks the form filling link, the corresponding form will be retrieved based on the business type of the company's inquiry. The bank will then fill in the quotation data and submit it. Based on the quotation data submitted by banks, comparative analysis is performed according to preset rules and a large model to generate recommended solutions. The large model is equipped with an enterprise-level financing knowledge base and generates quotation solutions through multi-agent collaborative decision-making and solution deduction. Businesses should refer to the recommendations when registering contracts.

[0006] As a preferred technical solution, the emails are sent using a one-to-one sending mechanism, with time limits and data isolation: each email is sent to only a single bank; after the quotation deadline, the bank's quotation data is not allowed to be modified and is only visible to the applicant company.

[0007] As a preferred technical solution, the business types of the enterprise include: Financing types include loans, bill issuance, bill discounting, factoring, guarantees, letters of credit, bonds, consolidated trusts, market-oriented debt-to-equity swaps, supply chain electronic debt instruments, financial leasing, fund financing, and asset securitization. Among these, loans include working capital loans, fixed asset loans, entrusted loans, and project loans. Fund deposit types include fixed-term and current deposits, with fixed-term deposits including time deposits and large-denomination certificates of deposit.

[0008] As a preferred technical solution, the comparative analysis using the aforementioned large model is as follows: Demand Analysis and Task Planning: Receive financing requests from enterprises, extract structured elements including financing amount, term, purpose, and risk appetite based on a large language model, and generate an executable task plan; Parse and extract external knowledge from multi-source heterogeneous data documents, and convert unstructured external knowledge into vectors to build an enterprise-level financing knowledge base; Multi-agent collaborative decision-making enables multi-dimensional and parallel evaluation and optimization of financing schemes. The RAG enhancement engine retrieves external knowledge from the enterprise-level financing knowledge base as the basis for generating content and generating a ranking of recommended schemes. The candidate financing schemes are deduced and simulated, and the final recommended scheme is output.

[0009] As a preferred technical solution, the enterprise-level financing knowledge base is constructed as follows: It accesses and parses internal enterprise data, external market data, unstructured bank quotation documents, and policy and regulatory texts. Using multimodal parsing technology and combined with a pre-trained financial entity identification model, it extracts key financing fields, including interest rates, fees, and guarantee terms, from the documents. Implement cross-document validation and information fusion to ensure data consistency; By using a vectorization model, unstructured key financing fields are converted into vectors and stored in a vector database, thus building an enterprise-level financing knowledge base.

[0010] As a preferred technical solution, the specific division of labor among the intelligent agents is as follows: Financing plan analyst AI agent: invokes cost assessment and risk quantification sub-models to conduct preliminary screening and scoring; Rule parsing agent: Dynamically loads and applies business rules defined by the user in natural language; Compliance assessment agent: Based on RAG enhanced retrieval, it matches and reviews whether the proposal complies with the latest regulatory policies from the enterprise-level financing knowledge base; Case expert intelligent agent: retrieves similar historical cases to provide reference for current decision-making; Risk warning intelligent agent: invokes risk prediction model to identify potential risks of the plan and generate alerts; Intelligent agent collaborative integration: The output results of each intelligent agent are merged, conflict resolved, and comprehensively optimized to generate a ranking of recommended financing schemes.

[0011] As a preferred technical solution, the candidate financing schemes are deduced and simulated, and Monte Carlo simulation and sensitivity analysis are used to conduct multi-scenario, forward-looking risk and cost assessments of the candidate pricing schemes, and the final scheme is output.

[0012] As a second aspect of the present invention, a system for enterprises to solicit quotations from banks is provided, the system executing the method for enterprises to solicit quotations from banks as described above, including: Business selection module: Used for businesses to select their business type; Application form module: Used for companies to fill in quotation request information; Approval module: performs verification and approval processes for quotation requests; Email sending module: After verification and approval, it automatically generates an inquiry email and sends it to the bank. The inquiry email contains core inquiry requirements and a form filling link. Bank Quotation Module: After the bank clicks the form filling link, a differentiated form will be displayed according to the business type. The bank fills in the quotation data and submits it. Comparison and Recommendation Module: Performs comparison and analysis according to preset rules and large model, and generates solution recommendation results; Contract registration module: Based on the comparison and recommendation results, the contract registration process is carried out.

[0013] As a preferred technical solution, the comparison and recommendation module specifically includes: Demand Analysis and Task Planning Engine: Receives users' financing requests, performs deep semantic understanding based on a large language model, extracts structured elements including financing amount, term, purpose, and risk preference, and generates an executable task plan for the system. Multi-source heterogeneous data parsing engine: Accesses and parses internal enterprise data, external market data, unstructured bank quotation documents, and policy and regulatory texts; adopts multimodal parsing technology, combined with a pre-trained financial entity identification model, to extract key fields from documents as external knowledge; and converts unstructured external knowledge into vectors and stores them in a vector database through a vectorization model to build an enterprise-level financing knowledge base; Multi-agent collaborative decision engine: It adopts a cognitive center based on a large language model for overall control, coordinates multiple agents with specific functions to work together, simulates the division of labor and cooperation of professional teams, and conducts multi-dimensional and parallel intelligent evaluation and optimization of financing plans. Dynamic rules and RAG enhancement engine: Acquire business rules defined or adjusted by business personnel through natural language and compile them into executable logic; during the decision generation process, retrieve relevant clauses, cases and policies from the vectorized knowledge base in real time as the basis for generating content; Solution simulation and modeling engine: Through Monte Carlo simulation and sensitivity analysis, the engine conducts multi-scenario, forward-looking risk and cost assessments of candidate financing solutions and outputs the final solution.

[0014] As a preferred technical solution, the specific division of labor among the agents in the multi-agent collaborative decision-making engine is as follows: Financing plan analyst AI agent: invokes cost assessment and risk quantification sub-models to conduct preliminary screening and scoring; Rule parsing agent: Dynamically loads and applies business rules defined by the user in natural language; Compliance assessment agent: Based on RAG enhanced retrieval, it matches and reviews whether the proposal complies with the latest regulatory policies from the enterprise-level financing knowledge base; Case expert intelligent agent: retrieves similar historical cases to provide reference for current decision-making; Risk warning intelligent agent: invokes risk prediction model to identify potential risks of the plan and generate alerts; Intelligent agent collaborative integration: The output results of each intelligent agent are merged, conflict resolved, and comprehensively optimized to generate a ranking of recommended financing schemes.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1) This invention automates the price inquiry process between enterprises and banks by constructing an automated system, covering steps such as application submission, email notification, bank quotation, system comparison, and manual adjustment. The system intelligently compares bank quotations using preset rules and large-scale model analysis technology, combined with a manual intervention mechanism, and finally outputs recommended results, completing closed-loop management of the entire process.

[0016] 2) This invention uses a multi-agent simulation of an expert team to comprehensively evaluate from multiple dimensions (cost, risk, compliance, case studies), overcoming the limitations of human decision-making and the selection of a single rule. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a method for enterprises to solicit quotations from banks according to the present invention. Detailed Implementation

[0018] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0019] Example 1 By introducing an advanced architecture of "large model cognitive center + multi-agent collaboration + knowledge base driven," not only was the price inquiry process automated, but also the financing decision-making process became more intelligent, in-depth, and scientific. For example... Figure 1 As shown, the specific steps of the method of the present invention include: Step 1: The company selects the business type, such as financing type or fund deposit type, and fills in the application information (such as amount, term, purpose, etc.), and stores the above information in the database.

[0020] Furthermore, the business types supported by this invention include: financing types, including loans (working capital loans, fixed asset loans, entrusted loans, project loans), bill issuance, bill discounting, factoring, guarantees, letters of credit, bonds, consolidated trusts, market-oriented debt-to-equity swaps, supply chain electronic debt instruments, financial leasing, fund financing, and asset securitization; and fund deposit types, including time deposits (fixed deposits, large-denomination certificates of deposit) and demand deposits.

[0021] Step 2: Verify and approve the application information.

[0022] Step 3: After verification and approval, an email containing key application information and a link to fill in the form will be automatically generated and sent to the bank. The dynamic email content is generated as follows: Based on the application information filled in by the company (such as financing type, amount, term, etc.), the email body containing the core requirements will be dynamically generated to ensure accurate information delivery.

[0023] Email system integration: Connect to the email servers of enterprises or banks (such as SMTP protocol, enterprise email system) to realize automatic email sending and status tracking (such as sending success / failure).

[0024] Emails are sent using a one-to-one sending mechanism: multiple banks can be selected as recipients in batches, but the "one-to-one" sending logic must be implemented (each email is sent to only a single bank) to avoid information leakage.

[0025] Step 4: After the bank clicks the link, it will retrieve the corresponding form based on the business information, fill in the quotation, and submit it.

[0026] Specifically, different forms are displayed based on the type of business. For example, information to be filled in for working capital loans includes interest rate, loan amount, withdrawal method, and required documents; information to be filled in for project loans includes initial loan interest rate, project loan interest rate, approval time, grace period, whether to participate in a group, and repayment method; information to be filled in for bills includes processing time, margin ratio, margin amount, and handling fee rate; and information to be filled in for time deposits includes yield, purchase period, and overall return.

[0027] Step 5: Accept and record the bank's quotation data. Once the deadline is reached, trigger the selection logic.

[0028] Time limits and data isolation are implemented: once the quotation deadline has passed, modifications are prohibited to ensure fairness; bank quotation data is only visible to the applicant companies.

[0029] Step 6: Perform comparative analysis according to preset rules and the large model to generate recommendation results; the large model analysis process includes requirement input and parsing, data preparation and knowledge retrieval, multi-agent parallel evaluation, result integration and solution deduction.

[0030] 6.1) Quotation data aggregation and parsing: Extract key parameters such as interest rates, fees, and service terms from the quotations submitted by banks, and perform standardized storage and multi-dimensional correlation.

[0031] 6.2) Demand Analysis and Task Planning: Receive financing requests from enterprises, perform deep semantic understanding based on a large language model, extract structured elements including financing amount, term, purpose, and risk preference, and automatically generate an executable task plan for the system.

[0032] 6.3) Intelligent Analysis and Governance of Multi-Source Heterogeneous Data: Unified access and analysis of internal enterprise data, external market data, unstructured bank quotation documents, policy and regulatory texts, etc. Employing multimodal analysis technology combined with a pre-trained financial entity identification model, key fields such as interest rates, fees, and guarantee terms are automatically extracted from documents. Cross-document verification and information fusion are implemented to ensure data consistency. Unstructured knowledge is converted into vectors through a vectorization model and stored in a vector database, constructing an enterprise-level financing knowledge base.

[0033] 6.4) Multi-Agent Collaborative Decision Engine: Simulates the division of labor and collaboration within a professional team, performing multi-dimensional and parallel intelligent evaluation and optimization of financing plans. It employs a cognitive central control system based on a large language model to coordinate the collaborative work of multiple agents with specific functions. The division of labor among the agents is as follows: Financing plan analyst agent: invokes sub-models such as cost assessment and risk quantification to conduct preliminary screening and scoring.

[0034] Rule parsing agent: Dynamically loads and applies business rules defined by the user in natural language.

[0035] Compliance assessment agent: Based on RAG enhanced retrieval, it matches and reviews the proposals from the knowledge base to determine whether they comply with the latest regulatory policies.

[0036] Case Expert Agent: Retrieves similar historical cases to provide reference for current decision-making.

[0037] Risk warning agent: invokes the risk prediction model to identify potential risks of the plan and generate alerts.

[0038] Agent Collaborative Integration: The Agent integrates the outputs of each agent, resolves conflicts, and comprehensively optimizes them to generate a ranking of recommended solutions.

[0039] 6.5) Dynamic Rules and RAG Enhancement Engine: Provides flexible and evolvable rule management, and leverages retrieval-enhanced generation technology to improve decision-making accuracy. Business users can define or adjust business rules using natural language, and the system automatically compiles them into executable logic. Integrating the RAG engine, relevant clauses, cases, and policies are retrieved in real-time from the vectorized knowledge base during the decision generation process as the basis for generated content, reducing the "illusion" of large models. Supports reinforcement learning optimization of rule weights, automatically adjusting them based on historical decision performance.

[0040] Data quality, knowledge engineering, and RAG effectiveness: The accuracy of multi-source data needs to be guaranteed, and the timeliness of structured knowledge extracted from massive documents must be maintained. A multi-dimensional ranking approach, considering semantic similarity, keyword weight, and timeliness, is necessary to avoid decision-making quality being negatively impacted by simple vector similarity retrieval.

[0041] 6.6) Scheme derivation and simulation: Through Monte Carlo simulation and sensitivity analysis, conduct multi-scenario, forward-looking risk and cost assessments of candidate financing schemes, and output the final scheme.

[0042] Step 7: Enterprises refer to the recommended results and can make manual adjustments. After completion, they register the contract.

[0043] Manual intervention mechanism: Provides a visual interface that allows users to manually adjust the system's recommended results, and records the reasons for the adjustments and the approval process to ensure compliance.

[0044] Example 2 As another embodiment of the present invention, this embodiment also constructs an automated system for performing the method as described in Embodiment 1, specifically including: Business selection module: Enterprises select business types, such as financing types (loans / bills / guarantees, etc.) or fund deposit types (fixed-term / current accounts). Application form module: Companies fill in application information (such as amount, time limit, purpose, etc.); Approval module: The system verifies and approves applications. Email sending module: After approval, the system automatically generates an email containing key application information and a link to fill in the form; Bank Quotation Module: After clicking the link, the system displays differentiated forms based on the business type. For example, information to be filled in for working capital loans includes interest rate, loan amount, withdrawal method, and required documents; information to be filled in for project loans includes initial loan interest rate, project loan interest rate, approval time, grace period, whether to participate in a group purchase, and repayment method; information to be filled in for bills includes processing time, margin ratio, margin amount, and handling fee rate; and information to be filled in for time deposits includes yield, purchase period, and overall return.

[0045] Comparison and Recommendation Module: 1) Demand Analysis and Task Planning Engine: Receives users' financing requests and performs deep semantic understanding based on a large language model to accurately extract structured elements such as financing amount, term, purpose, and risk preference, and automatically generates executable task plans for the system.

[0046] 2) Multi-source heterogeneous data parsing engine: Unifies access and parses internal enterprise data, external market data, unstructured bank quotation documents, policy and regulatory texts, etc. Employing multimodal parsing technology combined with a pre-trained financial entity recognition model, it automatically extracts key fields such as interest rates, fees, and guarantee terms from documents. Cross-document verification and information fusion are implemented to ensure data consistency. Unstructured knowledge is converted into vectors through a vectorization model and stored in a vector database, building an enterprise-level financing knowledge base.

[0047] 3) Multi-agent collaborative decision-making engine: Simulates the division of labor and collaboration among professional teams, conducting multi-dimensional and parallel intelligent evaluation and optimization of financing plans. It employs a "cognitive hub" based on a large language model for overall control, coordinating the collaborative work of multiple agents with specific functions.

[0048] Division of labor among intelligent agents: Financing Solution Analyst (Agent): Utilizes sub-models such as cost assessment and risk quantification to conduct initial screening and scoring.

[0049] Rule parsing agent: Dynamically loads and applies business rules defined by the user in natural language.

[0050] Compliance Assessment Agent: Based on RAG enhanced search, it matches and reviews the scheme from the knowledge base to see if it complies with the latest regulatory policies.

[0051] Case Expert Agent: Retrieves similar historical cases to provide reference for current decision-making.

[0052] Risk Warning Agent: Invokes the risk prediction model to identify potential risks of the solution and generate alerts.

[0053] Agent Collaborative Integration: The agent integrates, resolves conflicts, and comprehensively optimizes the outputs of each agent to generate a ranking of recommended solutions.

[0054] (4) Dynamic Rules and RAG Enhancement Engine: Provides flexible and evolvable rule management, and utilizes retrieval-enhanced generation technology to improve decision-making accuracy. It allows business personnel to define or adjust business rules using natural language, and the system automatically compiles them into executable logic. It integrates a RAG engine to retrieve relevant clauses, cases, and policies from a vectorized knowledge base in real time during the decision generation process, serving as the basis for generated content and reducing the "illusion" of large models. It supports reinforcement learning optimization of rule weights, automatically adjusting them based on historical decision performance.

[0055] (5) Scheme deduction and simulation engine: Through Monte Carlo simulation and sensitivity analysis, the candidate financing schemes are evaluated for risks and costs in multiple scenarios and forward-looking perspectives, and the final scheme is output.

[0056] Contract registration module: After the selection process is completed, the contract registration process begins.

[0057] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0058] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A method for enterprises to solicit quotations from banks, characterized by the following steps: include: Obtain information about the company's business type and inquiry application; Based on the inquiry application information filled in by the enterprise, an email body containing the core inquiry requirements is dynamically generated and sent to the bank with an email containing a link to fill in the form; After the bank clicks the form filling link, the corresponding form will be retrieved based on the business type of the company's inquiry. The bank will then fill in the quotation data and submit it. Based on the quotation data submitted by banks, comparative analysis is conducted according to preset rules and a large model to generate recommended solutions. The large model is equipped with an enterprise-level financing knowledge base and generates pricing schemes through multi-agent collaborative decision-making and scheme deduction. Based on the recommendations, proceed with contract registration.

2. The method for enterprises to solicit quotations from banks according to claim 1, characterized in that, The emails are sent using a one-to-one sending mechanism, with time limits and data isolation: each email is sent to only a single bank; after the quotation deadline, bank quotation data cannot be modified, and the quotation data is only visible to the applicant company.

3. The method for enterprises to solicit quotations from banks according to claim 1, characterized in that, The business types of the aforementioned enterprises include: Financing types include loans, bill issuance, bill discounting, factoring, guarantees, letters of credit, bonds, consolidated trusts, market-oriented debt-to-equity swaps, supply chain electronic debt instruments, financial leasing, fund financing, and asset securitization. Among these, loans include working capital loans, fixed asset loans, entrusted loans, and project loans. Fund deposit types include fixed-term and current deposits, with fixed-term deposits including time deposits and large-denomination certificates of deposit.

4. The method for enterprises to solicit quotations from banks according to claim 1, characterized in that, The comparative analysis using the aforementioned large model is as follows: Demand Analysis and Task Planning: Receive financing requests from enterprises, extract structured elements including financing amount, term, purpose, and risk appetite based on a large language model, and generate an executable task plan; Parse and extract external knowledge from multi-source heterogeneous data documents, and convert unstructured external knowledge into vectors to build an enterprise-level financing knowledge base; Multi-agent collaborative decision-making enables multi-dimensional and parallel evaluation of financing schemes. The RAG enhancement engine retrieves external knowledge from the enterprise-level financing knowledge base as the basis for generating content and generating a ranking of recommended schemes. The candidate financing schemes are deduced and simulated, and the final recommended scheme is output.

5. The method for enterprises to solicit quotations from banks according to claim 4, characterized in that, The enterprise-level financing knowledge base is constructed as follows: It accesses and parses internal enterprise data, external market data, unstructured bank quotation documents, and policy and regulatory texts. Using multimodal parsing technology and combined with a pre-trained financial entity identification model, it extracts key financing fields, including interest rates, fees, and guarantee terms, from the documents. Implement cross-document validation and information fusion to ensure data consistency; By using a vectorization model, unstructured key financing fields are converted into vectors and stored in a vector database, thus building an enterprise-level financing knowledge base.

6. The method for enterprises to solicit quotations from banks according to claim 4, characterized in that, The specific division of labor among the intelligent agents is as follows: Financing plan analyst AI agent: invokes cost assessment and risk quantification sub-models to conduct preliminary screening and scoring; Rule parsing agent: Dynamically loads and applies business rules defined by the user in natural language; Compliance assessment agent: Based on RAG enhanced retrieval, it matches and reviews whether the proposal complies with the latest regulatory policies from the enterprise-level financing knowledge base; Case expert intelligent agent: retrieves similar historical cases to provide reference for current decision-making; Risk warning intelligent agent: invokes risk prediction model to identify potential risks of the plan and generate alerts; Intelligent agent collaborative integration: The output results of each intelligent agent are merged, conflict resolved, and comprehensively optimized to generate a ranking of recommended financing schemes.

7. The method for enterprises to solicit quotations from banks according to claim 4, characterized in that, The process involves extrapolating and simulating candidate financing schemes, employing Monte Carlo simulation and sensitivity analysis to conduct multi-scenario, forward-looking risk and cost assessments of candidate pricing schemes, and ultimately outputting the final scheme.

8. A system for enterprises to solicit quotations from banks, characterized in that, The system executes the method for enterprises to solicit quotations from banks as described in any one of claims 1-7, including: Business selection module: Used for businesses to select their business type; Application form module: Used for companies to fill in quotation request information; Approval module: performs verification and approval processes for quotation requests; Email sending module: After verification and approval, it automatically generates an inquiry email and sends it to the bank. The inquiry email contains core inquiry requirements and a form filling link. Bank Quotation Module: After the bank clicks the form filling link, a differentiated form will be displayed according to the business type. The bank fills in the quotation data and submits it. Comparison and Recommendation Module: Performs comparison and analysis according to preset rules and large model, and generates solution recommendation results; Contract registration module: Based on the comparison and recommendation results, the contract registration process is carried out.

9. A system for enterprises to solicit quotations from banks according to claim 8, characterized in that, The comparison and recommendation module specifically includes: Demand Analysis and Task Planning Engine: Receives users' financing requests, performs deep semantic understanding based on a large language model, extracts structured elements including financing amount, term, purpose, and risk preference, and generates an executable task plan for the system. Multi-source heterogeneous data parsing engine: Accesses and parses internal enterprise data, external market data, unstructured bank quotation documents, and policy and regulatory texts; adopts multimodal parsing technology, combined with a pre-trained financial entity identification model, to extract key fields from documents as external knowledge; and converts unstructured external knowledge into vectors and stores them in a vector database through a vectorization model to build an enterprise-level financing knowledge base; Multi-agent collaborative decision engine: It adopts a cognitive center based on a large language model for overall control, coordinates multiple agents with specific functions to work together, simulates the division of labor and cooperation of professional teams, and conducts multi-dimensional and parallel intelligent evaluation and optimization of financing plans. Dynamic rules and RAG enhancement engine: Acquire business rules defined or adjusted by business personnel through natural language and compile them into executable logic; during the decision generation process, retrieve relevant clauses, cases and policies from the vectorized knowledge base in real time as the basis for generating content; Solution simulation and modeling engine: Through Monte Carlo simulation and sensitivity analysis, the engine conducts multi-scenario, forward-looking risk and cost assessments of candidate financing solutions and outputs the final solution.

10. A system for enterprises to solicit quotations from banks according to claim 9, characterized in that, In the multi-agent collaborative decision-making engine, the specific roles of each agent are as follows: Financing plan analyst AI agent: invokes cost assessment and risk quantification sub-models to conduct preliminary screening and scoring; Rule parsing agent: Dynamically loads and applies business rules defined by the user in natural language; Compliance assessment agent: Based on RAG enhanced retrieval, it matches and reviews whether the proposal complies with the latest regulatory policies from the enterprise-level financing knowledge base; Case expert intelligent agent: retrieves similar historical cases to provide reference for current decision-making; Risk warning intelligent agent: invokes risk prediction model to identify potential risks of the plan and generate alerts; Intelligent agent collaborative integration: The output results of each intelligent agent are merged, conflict resolved, and comprehensively optimized to generate a ranking of recommended financing schemes.

Images