AI tools turn handwritten sales logs into digital insights for small and medium-sized businesses

Ateneo de Manila University's Business Insights for Development (Build) is seeking ways to expand and enhance artificial intelligence (AI) to increase and enhance human labor in small and medium-sized businesses that make up the majority of the Philippine economy.

Build researchers Zachary Matthew Alabastro, Joseph Benjamin Ilagan, Lois Abigail to, and Jose Ramon Ilagan pay particular attention to the humble pen and paper logbook, a very ubiquitous tool in everyday business.

Low-cost, simple and reliable, this analog solution is essential for tracking a large number of small items that change hands throughout the day with small businesses, from shopping centre stalls to local “sarisari” or convenience stores.

Handwritten ledgers can be counted for record keeping, even in the bustling and often busy environments of kitchens and back rooms where electronic tablets can be awkward or fragile. But despite its reliable simplicity, handwritten logs can be laborious to create meaning by drawing them on tables, especially when they are trying to gather insight into how a business is run.

On the other hand, AI is completely useful for business data analysis. This is a short task of identifying which products are not working well. Track sales trends over time. Provides stock, pricing, and restocking recommendations.

But, of course, many small business owners and workers are reluctant to digitize them due to concerns about a sharp learning curve and job redundancy. However, Ateneo researchers have proposed a “co-pilot” model in which AI complements and supports human efforts.

Their research was presented at the Human Computer Interaction Conference 2025 in Sweden at Artificial Intelligence, which explores how Optical Character Recognition (OCR) and leading Language Model (LLM) technologies will transform handwritten sales logs into more manageable digital data.

The researcher system tested on the real food stalls at Ateneo's Student Enterprise Center is built in Python and uses Amazon Web Services to interpret handwritten logbook data with OCR and Anthropic's Claude 3 Haiku LLM.

This system makes it easy for people without digital training to grasp inventory trends. Scan your logbook photos and use AI to recognize products, match prices and provide a sales overview. This makes it easier for businesses to meet customer demand by quickly identifying bestsellers and slow-moving inventory.

The researcher's early prototypes show moderate accuracy and offer much hope for improvement. It can also be adapted to process other types of handwritten data, such as inventory sheets, distribution logs, and even pay ledgers.

Unlike the analog logbook itself, this AI tool is intended to be simple, affordable and easily upgradeable. When AI accuracy improves with training in more shorthand writing, local food stalls will ultimately gain reliable, low-cost access to business insights once reserved for large businesses.