Multi-agent based business consulting report generation method and computer program product

Abstract

The application discloses a kind of commercial consulting report generation method and computer program product based on multi-agent.The method is: obtaining multiple industry research original data;For each original data, label source mark (including file name and this page / section code) for each page / each section thereof;From all original data, key business element text is extracted and the source mark thereof is attached to each section;According to the key business element text, in the order of report generation process node, corresponding link scheme is generated in turn, in this process, after generating corresponding scheme, if receiving user modification instruction, modify this scheme and take it as the standard, each text in each scheme is attached with its source mark;According to each link scheme, generate business consulting report, each text in the report is attached with its source mark.The application can inject fine-grained source mark in the process of generating business consulting report and accept user input adjustment instruction to ensure that the report meets the actual demands of enterprises.

Description

Technical Field

[0001] This invention relates to the field of AI large model technology, and in particular to a method and computer program product for generating business consulting reports based on multi-agent systems. Background Technology

[0002] Many companies typically conduct brand research before product development. Traditionally, this involves domain experts manually reviewing vast amounts of industry research data, performing detailed analysis, and then writing a brand research business consulting report. This traditional method is time-consuming and labor-intensive. With the development of Large Language Models (LLM), some companies have attempted to use AI models to automatically generate brand research business consulting reports. However, existing AI models often lack source citations for the conclusions and data generated, making it difficult for users to verify their authenticity and resulting in a lack of credibility in the business consulting reports. Furthermore, existing AI models usually generate brand research business consulting reports with a single click, leading to rigid, inflexible content that may deviate from the company's actual needs. Summary of the Invention

[0003] The purpose of this invention is to provide a method for generating business consulting reports based on multi-agent systems and a computer program product that can implement the method. This method can inject traceability markers during the generation of business consulting reports and accept user input adjustment instructions to ensure that the reports meet the actual needs of enterprises.

[0004] To achieve the above objectives, the present invention provides a method for generating business consulting reports based on multi-agent systems, characterized by comprising the following steps: S1. Obtain multiple sets of heterogeneous industry survey raw data from various sources; S2. For each piece of raw industry research data, mark each page / section with a source traceability tag. The source traceability tag includes the file name and the code for this page / section. S3. Extract key business element texts from all industry survey raw data. Each paragraph in the key business element text is annotated with the page / paragraph from which it was taken in the raw data. S4. Based on the key business element text, generate the corresponding step plan for each node in the order of the report generation process. During this process, if a user modification instruction is received after generating the corresponding plan, modify the plan and the modified plan shall prevail. Each paragraph in each plan shall be marked with the traceability mark of the page from which it was taken in the original data. S5. Based on the solutions generated at each node, generate a business consulting report. Each paragraph in the report should include a source tag indicating the page from which it was taken from the original data.

[0005] Furthermore: The report generation process includes the following nodes, which are connected sequentially: brand positioning node, positioning dimension reliability verification node, and brand expression form node. In step S4, "based on the key business element text, generate the corresponding process plan for each node in sequence according to the report generation process node order" specifically: generate a brand positioning plan based on the key business element text, generate a positioning dimension reliability demonstration plan based on the brand positioning plan and the key business element text, and generate a brand expression form plan based on the brand positioning plan and the positioning dimension reliability demonstration plan. S5 specifically generates a business consulting report based on the brand positioning plan, the reliability verification plan for the positioning dimensions, and the brand expression plan.

[0006] Furthermore, in steps S4 and S5, each solution and business consulting report is automatically generated using the corresponding intelligent agent.

[0007] Furthermore, in step S3, each piece of original data marked with traceability markers is divided into multiple logical blocks according to a preset character threshold. Then, all logical blocks are read, parsed, and extracted to obtain the key business element text.

[0008] Furthermore, in step S3, each text segment marked with a traceability tag in each piece of raw data is categorized, and the number of segments in each category is counted. If the number of segments in a category does not reach the threshold for that category, the execution of subsequent process steps is paused and the user is prompted to supplement the raw data of the industry survey. After receiving the new raw data, steps S1 to S3 are executed on the new raw data.

[0009] Furthermore, this includes S6, which is executed after step S5, to perform a logical check on the business consulting report.

[0010] Furthermore, in step S6, if a user modification instruction is received, the business consulting report is modified.

[0011] Furthermore, in step S6, if a user-specified jump instruction is received, the process jumps back to the user-specified scheme generation node and regenerates the corresponding process schemes for that node and its subsequent nodes in sequence.

[0012] Furthermore, in step S5, all traceability markers in each stage of the solution are obtained, all traceability markers are deduplicated and assigned numbers in sequence, the traceability markers attached to each paragraph of the business consulting report are replaced with their corresponding numbers, and then a list of cited sources is automatically generated at the end of the business consulting report based on the traceability markers and their corresponding numbers.

[0013] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the multi-agent-based business consulting report generation method described above.

[0014] The present invention provides a multi-agent-based business consulting report generation method. In step S2, a source marker is added to each page of the original data. In subsequent steps S3, S4, and S5, the text generated during the business consulting report generation process is always marked with the source marker of the page from which it was taken in the original data. This ensures that each conclusion in the business consulting report can be accurately traced back to the specific page / paragraph of the original data, with source tracing accurate to the page / paragraph level, achieving fine-grained source tracing. In addition, the method adopts a human-machine collaboration mechanism. Specifically, after generating the corresponding solution in step S4, the method accepts user input modification instructions. If a user input modification instruction is received, the solution is modified and the modified solution is taken as the standard. The business consulting report is generated collaboratively by AI agents and expert (user) opinions, ensuring that the final generated business consulting report meets the actual needs of the enterprise. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the multi-agent-based business consulting report generation method provided in this invention.

[0016] Figure 2 This is a block diagram illustrating the principle of the multi-agent-based business consulting report generation method for labeling and tracing the source of raw data, as presented in this invention.

[0017] Figure 3 This is the agent architecture diagram of the multi-agent business consulting platform provided by the present invention. Detailed Implementation

[0018] The present invention will be further described in detail below with reference to specific embodiments.

[0019] This embodiment was programmed by technicians using computer code, following the business consulting report generation process, and constructed using a Large Language Model (LLM) to create a multi-agent business consulting platform client (i.e., a computer program product) capable of automatically generating business consulting reports. The business consulting report generation process in this embodiment includes sequentially connected nodes for brand positioning, positioning dimension reliability verification, brand presentation, report generation, and quality auditing. Therefore, the multi-agent business consulting platform client, as shown... Figure 3As shown, the system includes, in sequence, a brand positioning agent (Agent 1), a positioning dimension reliability verification agent (Agent 2), a brand performance agent (Agent 3), a report generation agent (Agent 4), and a quality audit agent (Agent 5). Before generating the business consulting report according to the above process, data extraction is required from the acquired raw industry research data. Therefore, the multi-agent business consulting platform client also includes a data extraction agent (Agent 0) located before the brand positioning agent (Agent 1). The multi-agent business consulting platform client is installed on a computer, and the computer processor executes the multi-agent business consulting platform client to achieve the following... Figure 1 The method for generating business consulting reports based on multi-agent systems is shown below. The steps of this method are described in detail below.

[0020] S1. Obtain multiple sets of raw industry survey data from diverse sources.

[0021] To generate a business consulting report, a user runs the multi-agent business consulting platform client (hereinafter referred to as the multi-agent business consulting platform) on their computer, enters the business consulting report generation interface, and then uploads multiple sets of raw industry research data. The raw data document types include PDF documents, Word documents, PPT documents, Excel documents, etc. In this embodiment, the raw industry research data is uploaded by the user; in other embodiments, it can be automatically retrieved via a webpage.

[0022] S2. For each piece of raw data, mark each page / segment with a source tag. The source tag includes the file name and the code for this page / segment.

[0023] like Figure 2 As shown, after acquiring multiple heterogeneous industry survey raw data from various sources, the multi-agent business consulting platform performs page / paragraph parsing and text cleaning on each piece of raw industry survey data to obtain page-level / paragraph-level metadata. Each page / paragraph is then marked with a source tracing tag, which includes the filename and the current page / paragraph code. For raw data in document formats such as PDF, Word documents, PPT, and Excel spreadsheets, each page is marked with the source tracing tag "filename-current page code," where page numbers are used for PDF, Word, and PPT documents, and worksheet labels are used for Excel spreadsheets. For webpage-type raw data, each paragraph is marked with the source tracing tag "filename-current paragraph code." In other embodiments, this can be changed to marking each paragraph of raw data in all formats with the source tracing tag "filename-current paragraph code."

[0024] S3. Extract key business element texts from all industry survey raw data. Each paragraph in the key business element text is annotated with the page / paragraph from which it was taken in the raw data.

[0025] like Figure 2 As shown, the multi-agent business consulting platform divides each piece of raw industry research data with traceability tags into multiple logical blocks according to a set character threshold (Chunk Size). Then, it calls the data extraction agent (Agent 0) to read all logical blocks in parallel or sequentially, extracting key business element text (such as market pain points, competitor strengths and weaknesses, etc.) for subsequent business consulting report generation. The data extraction agent (Agent 0) adds a traceability tag to each paragraph in the key business element text, indicating the page / segment from which it was extracted from the raw data.

[0026] The multi-agent business consulting platform categorizes each text segment marked with a traceability tag in each piece of raw data, counts the number of segments in each category, and if the number of segments in any category does not reach the threshold for that category, the workflow is suspended, the execution of subsequent process steps is paused, and the user is prompted to supplement the raw data of the industry survey. After receiving the new raw data, steps S1 to S3 are executed on the new raw data until the number of segments in all categories reaches the corresponding threshold, and then the workflow is restarted to execute the subsequent process steps.

[0027] S4. Based on the key business element text, generate the corresponding step plan for each node in the order of the report generation process. During this process, if a user modification instruction is received after the corresponding plan is generated, the plan is modified and the modified plan shall prevail. Each text in each plan shall be annotated with the traceability mark of the page from which it was taken in the original data.

[0028] The business consulting report generation process in this embodiment includes a brand positioning node, a positioning dimension reliability verification node, a brand performance node, a report generation node, and a quality audit node, connected sequentially. The multi-agent business consulting platform, based on the key business element text, sequentially calls the corresponding agents to generate the solution for each node according to the order of the report generation process nodes. The specific process is as follows: The multi-agent business consulting platform invokes the brand positioning agent (Agent 1). Agent 1 receives the key business element text extracted by Agent 0 and analyzes it across four dimensions: consumer perception of the brand, perception of competitors, industry gaps, and the company's own strategy. It then identifies the intersection of these four dimensions and generates 10 candidate brand positioning schemes. Each segment of text in each scheme is annotated with a source tag indicating the page from which it was taken from the original data. The multi-agent business consulting platform, as shown... Figure 3As shown, a human-machine collaborative dynamic decision-making mechanism is introduced into the multi-agent architecture workflow by utilizing a state graph and checkpoint mechanism. After the multi-agent business consulting platform calls the brand positioning agent (Agent 1) to generate 10 candidate brand positioning schemes, the platform automatically suspends the workflow and triggers a checkpoint, pausing all platform processes, persisting the platform state, and outputting the 10 candidate brand positioning schemes to the user through the user interface. After viewing the various candidate brand positioning schemes in the user interface, the user selects the most suitable brand positioning scheme and confirms, thus the user has entered the modification command "Select number 3: the safest private car". After receiving the modification instruction, the multi-agent business consulting platform restarts the workflow and injects it into the brand positioning agent (Agent 1) through hard-coded "highest priority instruction" prefix. The brand positioning agent (Agent 1) converges the derivation path according to this instruction to the final output "Sequence No. 3: The safest private car" brand positioning scheme. Each text in the scheme is annotated with the traceability mark of the page from which it was taken in the original data. Then, the key business element text and the brand positioning scheme are output to the positioning dimension reliability demonstration agent (Agent 2).

[0029] After receiving the key business element text and brand positioning plan from Agent 1, the positioning dimension reliability demonstration agent (Agent 2) proves why the brand positioning plan can be supported according to preset supporting dimensions, such as product function, emotional resonance, and channel service. It generates a positioning dimension reliability demonstration plan, in which each text segment is annotated with a traceability tag indicating the page from which it was taken in the original data. After generating the positioning dimension reliability demonstration plan, the multi-agent business consulting platform automatically suspends the workflow and triggers a checkpoint, pauses all platform processes, persists the platform state, and outputs the positioning dimension reliability demonstration plan to the user through the user interface. After reviewing the positioning dimension reliability verification plan, if the user deems the plan unreliable and requires modification, they input the corresponding modification comments as a modification command into the user interface. The multi-agent business consulting platform receives the user's modification command and injects it into Agent 2 via a hard-coded "highest priority command" prefix. Agent 2 then modifies the positioning dimension reliability verification plan according to this command, generates a new positioning dimension reliability verification plan, and outputs it to the user through the user interface for review. This process continues until the user deems the plan reliable and clicks "YES" to confirm. Upon receiving the user's "YES" command, the multi-agent business consulting platform restarts the workflow, instructing Agent 2 to output the key business element text, brand positioning plan, and positioning dimension reliability verification plan to the brand representation agent (Agent 3).

[0030] After receiving key business element text, brand positioning plan, and positioning dimension reliability verification plan, the brand presentation agent (Agent 3) generates a brand presentation plan based on these. Each text segment in the plan includes a source tag indicating the page from which it was taken from the original data. The brand presentation plan may include, for example, suggested brand visual colors, slogan wording style, and marketing campaign formats. Once Agent 3 generates the brand presentation plan, the multi-agent business consulting platform automatically suspends the workflow and triggers a checkpoint, pausing all platform processes, persisting the platform state, and displaying the brand presentation plan to the user through the user interface. After reviewing the brand presentation plan, if a user deems it unreliable and requires modification, they input the corresponding modification comments as a modification command into the user interface. The multi-agent business consulting platform receives the user's modification command and injects it into Agent 3 via a hard-coded "highest priority command" prefix. Agent 3 then modifies the brand presentation plan according to this command, generates a new brand presentation plan, and outputs it to the user for review through the user interface until the user deems the plan reliable and clicks "YES". Upon receiving the user's "YES" command, the multi-agent business consulting platform restarts the workflow, instructing the brand presentation agent (Agent 3) to output the key business element text, brand positioning plan, positioning dimension reliability verification plan, and brand presentation plan to the report generation agent (Agent 4).

[0031] S5. Based on the solutions generated at each node, generate a business consulting report. Each paragraph in the report should include a source tag indicating the page from which it was taken from the original data.

[0032] The report generation agent (Agent 4) typeset and polished the brand positioning scheme, positioning dimension reliability demonstration scheme, and brand expression form scheme generated by Agent 1, Agent 2, and Agent 3 respectively, forming a business consulting report. It also obtains all traceability tags in each of the above schemes, removes duplicate traceability tags, assigns numbers to them in sequence, replaces the traceability tags attached to each paragraph of the report with their corresponding numbers, and then automatically generates a list of cited sources at the end of the business consulting report based on the traceability tags and their corresponding numbers.

[0033] S6. Perform a logical check on the business consulting report.

[0034] The multi-agent business consulting platform invokes the quality audit agent (Agent 5) to perform logical checks on the generated business consulting report. Once the report is confirmed to be free of logical errors, it is displayed to the user through the user interface. The user reviews the report and, if satisfied, clicks "Confirm END." Upon receiving the END command from the user, the multi-agent business consulting platform completes the business consulting report process. If the user finds minor flaws in the report, they input modification suggestions as a modification command. The platform then rewrites the identified flawed sections. If the user considers the report to have major flaws, they input a specified jumpback command on the user interface. This jumps back to the user-specified solution generation node and regenerates the corresponding steps for that node and its subsequent nodes. For example, if the user's specified jumpback command is to jump back to Agent 1, the platform will jump back to Agent 1 to regenerate the brand positioning solution. The platform will then automatically and sequentially call Agent 2 -> Agent 3 -> Agent 4 -> Agent 5 to execute the subsequent process steps.

[0035] The present invention provides a multi-agent-based business consulting report generation method. In step S2, a source marker is added to each page of the original data. In subsequent steps S3, S4, and S5, the text generated during the business consulting report generation process is always marked with the source marker of the page from which it was taken in the original data. This ensures that each conclusion in the business consulting report can be accurately traced back to the specific page / paragraph of the original data, with source tracing accurate to the page / paragraph level, achieving fine-grained source tracing. In addition, the method adopts a human-machine collaboration mechanism. Specifically, after generating the corresponding solution in step S4, the method accepts user input modification instructions. If a user input modification instruction is received, the solution is modified and the modified solution is taken as the standard. The business consulting report is generated collaboratively by AI agents and expert (user) opinions, ensuring that the final generated business consulting report meets the actual needs of the enterprise.

[0036] The above description is merely an embodiment of the present invention and does not limit the scope of patent protection. Any non-substantial changes or substitutions made by those skilled in the art based on the present invention will still fall within the scope of patent protection.

Claims

1. A method for generating business consulting reports based on multi-agent systems, characterized in that, Includes the following steps: S1. Obtain multiple sets of heterogeneous industry survey raw data from various sources; S2. For each piece of raw industry research data, mark each page / section with a source traceability tag. The source traceability tag includes the file name and the code for this page / section. S3. Extract key business element texts from all industry survey raw data. Each paragraph in the key business element text is annotated with the page / paragraph from which it was taken in the raw data. S4. Based on the key business element text, generate the corresponding step plan for each node in the order of the report generation process. During this process, if a user modification instruction is received after generating the corresponding plan, modify the plan and the modified plan shall prevail. Each paragraph in each plan shall be marked with the traceability mark of the page from which it was taken in the original data. S5. Based on the solutions generated at each node, generate a business consulting report. Each paragraph in the report should include a source tag indicating the page from which it was taken from the original data.

2. The method for generating business consulting reports based on multi-agent systems as described in claim 1, characterized in that: The report generation process includes the following nodes, which are connected sequentially: brand positioning node, positioning dimension reliability verification node, and brand expression form node. In step S4, "based on the key business element text, generate the corresponding process plan for each node in sequence according to the order of the report generation process nodes" specifically: generate a brand positioning plan based on the key business element text, generate a positioning dimension reliability demonstration plan based on the brand positioning plan and the key business element text, and generate a brand expression plan based on the brand positioning plan and the positioning dimension reliability demonstration plan. S5 specifically generates a business consulting report based on the brand positioning plan, the reliability verification plan for the positioning dimensions, and the brand expression plan.

3. The method for generating business consulting reports based on multi-agent systems as described in claim 1, characterized in that, In steps S4 and S5, each solution and business consulting report is automatically generated using the corresponding intelligent agent.

4. The method for generating business consulting reports based on multi-agent systems as described in claim 1, characterized in that, In step S3, each piece of original data marked with traceability markers is divided into multiple logical blocks according to a preset character threshold. Then, all logical blocks are read, parsed, and the key business element text is extracted.

5. The method for generating business consulting reports based on multi-agent systems as described in claim 4, characterized in that, In step S3, each text segment marked with a traceability tag in each piece of raw data is categorized, and the number of segments in each category is counted. If the number of segments in a category does not reach the threshold for that category, the subsequent process steps are paused and the user is prompted to supplement the raw data of the industry survey. After receiving the new raw data, steps S1 to S3 are executed on the new raw data.

6. The method for generating business consulting reports based on multi-agent systems as described in claim 1, characterized in that, This includes S6, which is executed after step S5, to perform a logical check on the business consulting report.

7. The method for generating business consulting reports based on multi-agent systems as described in claim 6, characterized in that, In step S6, if a user modification instruction is received, the business consulting report is modified.

8. The method for generating business consulting reports based on multi-agent systems as described in claim 7, characterized in that, In step S6, if a user-specified jump instruction is received, the process jumps back to the user-specified scheme generation node and regenerates the corresponding process schemes for that node and its subsequent nodes in sequence.

9. The method for generating business consulting reports based on multi-agent systems as described in claim 1, characterized in that, In step S5, all traceability markers in each stage of the solution are obtained, all traceability markers are deduplicated and assigned numbers in sequence, the traceability markers attached to each paragraph of the business consulting report are replaced with their corresponding numbers, and then a list of cited sources is automatically generated at the end of the business consulting report based on the traceability markers and their corresponding numbers.

10. A computer program product comprising a computer program, characterized in that, When executed by a processor, the computer program implements the multi-agent-based business consulting report generation method as described in any one of claims 1 to 8.

Images