Composite Agent Generation for Single-Instruction Task Automation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional robotic process automation systems require multiple user interactions and extensive command recording to perform sequential tasks, which is tedious and resource-intensive.

Innovation Solution

An agent cooperation graph (ACG) is generated based on machine-human interactions to identify cooperating digital agents, forming a composite sub-graph (CSG) that creates a composite agent capable of performing integrated tasks in response to a single user instruction, using a composite evaluator and actuator to automate multiple processes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If conventional robotic process automation systems perform sequential tasks through multiple user interactions and command recording, then task automation is achieved, but the process becomes tedious and resource-intensive

Engineering Contradiction:

Improvetask automationVSAvoiduser interaction time

Core Design Contradiction:

Extent of automationVSLoss of time

Solution Approach 1:

The patent merges multiple sequential digital agents into a single composite agent that can handle multiple tasks simultaneously. The composite agent integrates the functionalities of individual agents (e.g., flight booking, hotel booking, car rental) into one unified entity, eliminating the need for users to interact with each agent separately and reducing the overall interaction time.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The composite agent is designed with multi-functionality, capable of performing diverse tasks that were previously handled by separate specialized agents. This universal agent can process flight bookings, hotel reservations, car rentals, and other travel-related tasks within a single interaction session, improving automation efficiency without increasing user time investment.

Inventive Principle:

Principle #6Universality (Multi-functionality)

2Extent of automation

If conventional systems use multiple separate digital agents for different tasks, then task-specific automation is achieved, but processing resource demand increases

Engineering Contradiction:

Improvetask-specific automationVSAvoidprocessing resource demand

Core Design Contradiction:

Extent of automationVSUse of energy by moving object

Solution Approach 1:

The patent combines multiple separate digital agents into one composite agent, consolidating their processing resources. Instead of running multiple independent agent instances that each consume computational resources, the system executes a single integrated composite agent that shares underlying resources, thereby reducing overall processing demand while maintaining task-specific automation capabilities.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The composite agent provides universal functionality across multiple task domains, replacing the need for multiple specialized agents. This multi-functional approach allows a single agent instance to handle various tasks (flight booking, hotel reservation, car rental) that previously required separate agent instances,从而 reducing the total processing resource demand.

Inventive Principle:

Principle #6Universality (Multi-functionality)

3Adaptability or versatility

If multiple digital agents are used for related automated tasks, then comprehensive task coverage is achieved, but the complexity of system management increases

Engineering Contradiction:

Improvetask coverageVSAvoidsystem management complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent merges multiple digital agents into a single composite agent, simplifying system management. Instead of managing multiple independent agents with their own configurations, deployments, and coordination mechanisms, the system manages one unified composite agent. This reduction in the number of components directly lowers system management complexity while preserving comprehensive task coverage.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The composite agent provides universal coverage for multiple task types, replacing the need for multiple specialized agents. This multi-functional design maintains comprehensive task coverage (flight booking, hotel reservation, car rental, etc.) while simplifying the system architecture to a single agent entity, thereby reducing management complexity.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS12518760B2Automated composite agent generation by implicit derivation

Publication Date: 2026.01.06 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

Automated composite agent generation includes generating an agent cooperation graph (ACG) based on machine-human interactions between a user and digital agents pretrained to perform automated tasks. The ACG nodes represent digital agents connected by edges weighted according to cooperation densities. A composite sub-graph (CSG) whose nodes represent cooperating digital agents that cooperatively perform related tasks in response to multiple user requests is generated based on the cooperation densities. A composite agent configured to perform a composite process in response to a composite instruction is generated and includes a composite evaluator and composite actuator. The composite evaluator is generated based on the multiple requests using a language model. The composite actuator is generated using an automated processes compiler that compiles the related automated processes. The composite actuator of the composite agent actuates the composite process, which performs the related automated tasks as a single, integrated process.