Work

Agents at work: utilizing design patterns for AI - agent workflow automation

Year: 2025

Type: dissertation

Abstract: This thesis evaluates the automation potential of knowledge work tasks using multi-agent AI frameworks. A structured scorecard assesses task suitability across project roles. Selected tasks are implemented with distinct agent design patterns in CrewAI, using GPT-4o. Performance is evaluated via workflow-level metrics, including Task Success Rate and Effective Task Success Rate. Findings reveal that pattern-specific configurations significantly influence agent success, robustness, and resource efficiency. The study provides empirically grounded guidance onorchestrating design patterns for agents with varying task complexity in business analytics workflows.

Source: Portuguese National Funding Agency for Science, Research and Technology (RCAAP Project by FCT)

Author: Veltrup, Leon

Language: English

Topic: Artificial Intelligence Applications

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

SDG: Decent work and economic growth

Open Access status: green

Funder: Fundação para a Ciência e a Tecnologia

Awards: G3434864815, G388523895, Nova School of Business and Economics - Unidade de Investigação

OpenAlex ID: W7197587953

Cites: 82

Cited by:

Related to: 10