LLM Prompt Structuring for Access-Controlled Record Analysis

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Large language models struggle to understand complex records and maintain data security, often providing incorrect responses due to lack of record relationship understanding and inadequate security measures, especially when users have limited access.

Innovation Solution

A prompt generation model that identifies user access levels, extracts accessible record portions, and generates prompts in natural language format, ensuring only relevant data is provided to the large language model, thus enhancing understanding and security.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If large language models are provided with extensive data to improve understanding of complex records, then the model's analytical capability is improved, but security risks increase and computational costs increase

Engineering Contradiction:

Improvemodel understanding accuracyVSAvoidsecurity risk

Core Design Contradiction:

ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The patent segments the data provided to the large language model by identifying and separating only the specific portions of records that are relevant to the user's question and accessible to the user. This prevents the model from receiving unnecessary or sensitive data while still providing sufficient context for accurate analysis, thereby improving security without sacrificing understanding capability.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system extracts only the necessary data elements from complex records that are required to answer the user's question. By taking out only the relevant portions and excluding extraneous or sensitive information, the system maintains model accuracy while reducing security risks and computational overhead.

Inventive Principle:

Principle #2Taking out (Extraction)

2Reliability

If large language models are provided with extensive data to improve understanding of complex records, then the model's analytical capability is improved, but computational costs increase

Engineering Contradiction:

Improvemodel understanding accuracyVSAvoidcomputational cost

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system applies partial action by providing the model with exactly the amount of data needed to answer the question - not too little to be inaccurate, not too much to waste computational resources. This optimized level of data provision reduces computational costs while maintaining understanding accuracy.

Inventive Principle:

Principle #16Partial or excessive action

3Ease of operation

If simple prompting is used for large language model interaction, then ease of operation is improved, but the model's ability to understand complex record relationships deteriorates

Engineering Contradiction:

Improveprompting simplicityVSAvoidrecord relationship understanding

Core Design Contradiction:

Ease of operationVSReliability

Solution Approach 1:

The patent introduces an intermediary system between the user's simple prompt and the large language model. This intermediary automatically identifies relevant record portions, extracts necessary data, and formulates optimized prompts that preserve the simplicity of user interaction while enhancing the model's understanding of complex record relationships through properly structured input.

Inventive Principle:

Principle #24Intermediary (Mediator)

4Object-affected harmful factors

If users are granted access to portions of records based on access levels, then data security is improved, but the model's ability to provide comprehensive analysis deteriorates

Engineering Contradiction:

Improvedata securityVSAvoidanalysis completeness

Core Design Contradiction:

Object-affected harmful factorsVSReliability

Solution Approach 1:

The patent segments record access by identifying specific portions of records that are both relevant to the user's question and accessible according to the user's access level. This segmentation allows the system to provide comprehensive analysis within security boundaries by including all necessary accessible information while excluding restricted data.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system applies local quality by tailoring the data provided to each user based on their specific access level and the particular question being asked. Different users receive different portions of records appropriate to their needs and permissions, optimizing both security compliance and analysis completeness for each individual case.

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS20250371263A1Large language model prompt generation and structuring

Publication Date: 2025.12.04 APPIAN CORP

AI summary

Methods, systems, apparatuses, and non-transitory computer-readable media are provided for generating a prompt for a large language model. Operations may include receiving an input from a user, identifying an access level of the user, identifying a portion of a record associated with the input based on the access level of the user, identifying metadata associated with the portion of the record, identifying data associated with the portion of the record, generating the prompt based on a combination of the input, the data associated with the portion of the record, and the metadata associated with the portion of the record in a natural language format, and providing the prompt to the large language model.