Automated Social Media Advertising Poster Design Using a Multi-Agent AI Framework

Putri, Christiana Naida (Department of Informatics, Telkom University, Bandung, 40257, Indonesia)
Wiharja, Kemas Rahmat Saleh (Department of Informatics, Telkom University, Bandung, 40257, Indonesia)

Zero : Jurnal Sains, Matematika, dan Terapan
Vol 9, No 3 (2025): Zero: Jurnal Sains Matematika dan Terapan

This study proposes an automated system for social media advertising poster generation using a role-based multi-agent architecture implemented with CrewAI. The system decomposes the design task into three sequential agents responsible for text generation, visual asset recommendation, and grid-based layout optimization. A formal 12×12 discrete layout model is employed to represent spatial constraints, enabling consistent and structured poster composition. System performance was evaluated through user testing involving five respondents using a five-point Likert scale. The results show mean scores of 3.2 for content completeness, 3.6 for layout consistency, 3.8 for text relevance, and an overall performance mean of 3.4 (SD = 0.15), indicating satisfactory usability. From an applied mathematics perspective, this work contributes a computational layout formulation using grid discretization and rule-based optimization, as well as a quantitative evaluation of multi-agent coordination efficiency. The proposed framework demonstrates that agentic AI can effectively support structured visual content generation while maintaining user-controlled refinement.