Authenticated LLM Agent Tools for Financial Task Reliability

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing financial systems face challenges in securely and efficiently utilizing Large Language Model (LLM) agents due to concerns over factual accuracy and hallucinations, necessitating a more cautious approach by regulators.

Innovation Solution

A system is developed that includes an authentication engine, tools engine, and execution engine to manage LLM-based agents, providing secure access to financial tools and mitigating risks through an approval engine that requires user confirmation for critical actions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If LLM agents are deployed to perform financial tasks, then productivity and automation are improved, but reliability and factual accuracy deteriorate due to hallucinations

Engineering Contradiction:

Improveautomation capabilityVSAvoidfactual accuracy

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent introduces an intermediary system comprising a tools engine and execution engine that sits between the LLM agent and the actual task execution. The tools engine provides structured tool definitions and descriptions to the LLM agent, while the execution engine validates and executes only those tools that are properly authenticated and verified. This intermediary layer prevents the LLM agent from directly executing hallucinated or incorrect actions, thereby maintaining both automation capability and factual accuracy.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system performs preliminary actions by pre-defining and registering available tools with the tools engine before the LLM agent attempts to use them. The execution engine pre-authenticates user credentials and pre-validates tool parameters before actual execution. This preliminary preparation ensures that when the LLM agent generates tool calls, they are already constrained to valid, pre-verified options, reducing hallucinations while preserving automation.

Inventive Principle:

Principle #10Preliminary action

2Reliability

If security measures are implemented to verify user credentials, then reliability is improved, but device complexity and system overhead increase

Engineering Contradiction:

Improvesecurity assuranceVSAvoidsystem structure

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent merges the authentication functionality into the execution engine, which also handles tool execution. Rather than having separate authentication and execution systems, the execution engine performs both credential verification and tool execution in a unified process. This consolidation maintains strong security verification while reducing overall system complexity by eliminating redundant components and interfaces.

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS20250322374A1Tools for large language model agents

Publication Date: 2025.10.16 WELLS FARGO BANK NA

AI summary

An example computer system for providing one or more tools for large language model agents can include: one or more processors; and non-transitory computer-readable storage media encoding instructions which, when executed by the one or more processors, causes the computer system to: authenticate a user; provide a description of the one or more tools available for use by the large language model agents; execute the one or more tools upon receipt of a request from the large language model agents; and provide information in response to the request.