HMN 2025: How AI device turns handwritten gross sales logs into digital insights for small companies

The Ateneo de Manila University’s Business Insights Laboratory for Development (BUILD) is methods for synthetic intelligence (AI) to reinforce and improve—slightly than change—human labor in small companies, which make up the majority of the Philippine financial system.

BUILD researchers Zachary Matthew Alabastro, Joseph Benjamin Ilagan, Lois Abigail To, and Jose Ramon Ilagan pay specific consideration to a really ubiquitous device of on a regular basis enterprise: the standard pen-and-paper logbook.

Low-cost, easy, and dependable, this analog answer is indispensable for preserving observe of the multitude of small objects that change palms all through the day in small companies, from procuring heart meals stalls to neighborhood “sari-sari” or comfort shops.

A handwritten ledger could be counted on for record-keeping even within the bustling and oftentimes hectic setting of a kitchen or backroom, where an digital pill is likely to be too cumbersome or fragile to make use of. But regardless of their reliable simplicity, handwritten logs could be painstaking to tabulate and make sense of—particularly when one is making an attempt to glean insights into find out how to better run a enterprise.

Meanwhile, AI lends itself completely to enterprise information evaluation: it makes quick work of figuring out which merchandise are performing properly or poorly; monitoring gross sales tendencies over time; and providing suggestions on stock, pricing, and restocking.

Understandably, nonetheless, many small enterprise homeowners and employees hesitate to digitize out of concern over steep studying curves and job redundancy. But the Ateneo researchers suggest a “copilot” model, during which AI enhances and helps human effort.

Their study, introduced on the Artificial Intelligence in Human-Computer Interaction Conference 2025 in Sweden, explores how optical character recognition (OCR) and enormous language model (LLM) expertise can flip handwritten gross sales logs into extra manageable digital information.

Tested in an precise meals stall on the Ateneo’s Student Enterprise Center, the researchers’ system is constructed with Python and makes use of Amazon Web Services for OCR and Anthropic’s Claude 3 Haiku LLM to interpret the handwritten logbook information.

The system permits even somebody with no digital coaching to know stock tendencies with ease. It scans logbook pictures and makes use of AI to acknowledge merchandise, match costs, and tabulate gross sales summaries. This helps companies shortly establish bestsellers or slow-moving stock, thereby making it simpler to maintain up with buyer demand.

The researchers’ early prototype exhibits reasonable accuracy and gives a lot hope for enchancment. It can be tailored, they are saying, to deal with other forms of handwritten information comparable to stock sheets, supply logs, and even payroll ledgers.

Not in contrast to analog logbooks themselves, this AI device is supposed to be easy, reasonably priced, and simply upgradable. As AI accuracy improves by coaching on extra shorthand writing, native stalls can ultimately acquire dependable, low-cost entry to enterprise insights as soon as reserved for bigger enterprises.