Customer Analysis and Text Generation for Small Retail Stores Using LLM-Generated Marketing Presence

A team of researchers from Japan has developed a novel prototype system designed to help small retail stores create more effective marketing materials. The system, detailed in a paper for the 2025 International Conference on Smart Computing and Artificial Intelligence, tackles a common problem: large language models (LLMs) like GPT-4 often produce generic, uncreative text, while small business owners typically lack professional marketing expertise. The proposed solution is a structured, collaborative workflow that combines AI's data-processing power with human insight and creativity.

The system guides users through four key stages: analyzing target customers, generating initial POP text drafts, refining the language and expressions, and finally evaluating candidate texts through simulated customer personas. This process ensures the final output is both data-informed and creatively tailored. The researchers' experimental validation is compelling—using their system resulted in a significant quality improvement, with an average score increase of 2.37 points on a detailed -3 to +3 evaluation scale compared to materials created without the system's structured support.

This research highlights a crucial shift in AI application design, moving beyond fully automated text generation toward frameworks that facilitate effective human-AI partnership. By providing structure and tools for customer analysis and persona-based evaluation, the system empowers non-experts to leverage AI as a creative co-pilot rather than a replacement. The work demonstrates that the greatest gains in practical AI tools may come from thoughtfully designed interfaces and processes that augment human judgment, especially in domains like marketing that require nuanced persuasion and audience understanding.