Artificial Intelligence Agent Management System Based on a Three-Dimensional Strategy of Roles, Tasks, and Data

Abstract

An AI agent management system based on a three-dimensional strategy of roles, tasks, and data is disclosed. The AI ​​agent management system is connected to the user device. A role attribute module is connected to a transmission module, a task type judgment module is connected to a transmission module, a data range selection module is connected to both the transmission module and the task type judgment module, a strategy verification module is connected to the role attribute module, the task type judgment module, and the data range selection module, an AI agent control module is connected to the task type judgment module, the strategy verification module, and the AI ​​agent, an execution result recording module is connected to the AI ​​agent control module, and an audit tracking module is connected to all other modules except the AI ​​agent control module. This system obtains role attribute information, task type information, and data access range information corresponding to user identification information. Based on the AI ​​agent management strategy, a three-dimensional cross-verification of the combination of these three elements is performed to generate AI agent deployment and execution information. The system enables, restricts, downgrades, or disables agent execution, thereby achieving the technical effect of effectively controlling the deployment and execution of AI agents through a three-dimensional strategy verification mechanism integrating roles, tasks, and data.

Claims

1. An AI agent management system based on a three-dimensional strategy of roles, tasks, and data, providing management and control over the use of at least one AI agent, the system comprising: a transmission module, which receives a natural language message input to an AI agent from a user device, wherein the user device logs into the AI ​​agent management system using user identification information; a role attribute module, connected to the transmission module, which collects role attribute information corresponding to the user identification information from an identity and access management system (IAM); a task type determination module, connected to the transmission module, which parses the natural language message into a task and task type information using natural language processing (NLP) technology; and a data range selection module, connected to both the transmission module and the task type determination module, which uses NLP technology to filter data access range information corresponding to the task request message based on data source, data domain, data sensitivity, and / or data access permissions associated with the task information. A strategy verification module, connected to the role attribute module, the task type judgment module, and the data range selection module, verifies the deployment and execution of the combination of the role attribute information, the task type information, and the data access range information using a pre-established AI agent management strategy to generate AI agent deployment and execution information. An AI agent control module, connected to the task type judgment module and the strategy verification module, enables, restricts, downgrades, and disables the deployment and execution of the AI ​​agent performing the task based on the AI ​​agent deployment and execution information, thereby controlling the AI ​​agent. An execution result recording module, connected to the AI ​​agent management module, records the deployment and execution results of the AI ​​agent as an agent execution result; and an audit tracking module, connected to the transmission module, the role attribute module, the task type judgment module, the data range selection module, the policy verification module, and the execution result recording module, respectively, records the user identification information, the natural language message, the role attribute information, the task type information, the data access range information, the AI ​​agent deployment and execution information, and the agent execution result as audit information. The audit information provides a control mechanism for tracking, verifying, and legally reviewing the deployment and execution of the AI ​​agent.

2. The AI ​​agent management system based on a three-dimensional strategy of roles, tasks, and data as described in claim 1, wherein the role attribute module obtains the role attribute information through the identity management system, and the role attribute information includes at least one or a combination of the following: user unique identifier, department, title, permission level, data access restrictions, management rights, approval rights, or agency rights.

3. The AI ​​agent management system based on a three-dimensional strategy of roles, tasks, and data as described in claim 1, wherein the task type determination module performs text normalization, word segmentation, part-of-speech analysis, named entity recognition, and semantic intent recognition on the natural language message through the natural language processing technology to generate the task and the task type information, and the task type information includes one or a combination of query tasks, analysis tasks, generation tasks, control operation tasks, or prediction tasks.

4. The AI ​​agent management system based on a three-dimensional strategy of roles, tasks, and data as described in claim 1, wherein the data range selection module determines the data source and the data domain by mapping the task information to a data directory or data domain mapping table, and determines the data sensitivity and the data access permission based on the task information, wherein the data sensitivity includes at least low sensitivity, medium sensitivity, or high sensitivity.

5. The AI ​​agent management system based on a three-dimensional strategy of roles, tasks, and data as described in claim 1, wherein the strategy verification module establishes a three-dimensional correspondence between roles, tasks, and data based on the AI ​​agent management strategy through a rule engine, decision tree model, or strategy matrix to generate AI agent deployment execution information, and the AI ​​agent deployment execution information includes execution decision results of allow, conditional allow, downgrade execution, or block.

Images