The Business Intelligence and Development Lab (BUILD) at Ateneo de Manila University is exploring the application of artificial intelligence (AI) in small businesses, aiming to augment rather than replace human effort. As a vital pillar of the Philippine economy, small businesses widely rely on paper-and-pen logs in their daily operations. While these logs are low-cost, simple, and reliable, they are labor-intensive when it comes to data tabulation and interpretation.
BUILD researchers Zachary Matthew Alabastro, Joseph Benjamin Ilagan, Lois Abigail To, and Jose Ramon Ilagan proposed using AI technology to address this issue. They noted that although handwritten logs are indispensable in small businesses, they are inefficient for generating business insights. In contrast, AI shows great potential in business data analysis, easily identifying product performance, tracking sales trends, and providing recommendations on inventory, pricing, and restocking.
In response to the concerns of small business owners and employees regarding digital transformation, the researchers innovatively proposed a "co-pilot" mode, positioning AI as a complement and support to human efforts. At the 2025 International Conference on Human-Computer Interaction with Artificial Intelligence held in Sweden, they presented their research findings, demonstrating how optical character recognition (OCR) and large language model (LLM) technologies can convert handwritten sales logs into manageable digital data. The system was successfully tested at an actual food stall in the Ateneo University Student Enterprise Center. It was built using Python, utilized Amazon Web Services for OCR processing, and employed Anthropic's Claude 3 Haiku LLM to interpret the data.
This system allows personnel without digital training to easily grasp inventory trends. By scanning photos of the logs, the AI can identify products, match prices, and summarize sales data. This helps businesses quickly identify best-selling and slow-moving products, better meeting customer needs. The researchers noted that although the early prototype system has moderate accuracy, there is significant room for improvement. In the future, it can also handle handwritten data such as inventory sheets, delivery logs, and even payroll records. This AI tool is designed to be simple, affordable, and easy to upgrade, enabling local stalls to gain business insights previously available only to large enterprises.
