At the Ateneo de Manila University, the Business Insights Laboratory for Development (BUILD) is pioneering new approaches to integrate artificial intelligence (AI) into small-scale enterprises, emphasising support rather than substitution. In a country where micro and small businesses form the backbone of the economy, the initiative focuses on enhancing, not displacing, human labour through the thoughtful use of AI technologies.

Researchers Zachary Matthew Alabastro, Joseph Benjamin Ilagan, Lois Abigail To, and Jose Ramon Ilagan have centred their attention on one of the most familiar yet overlooked tools of everyday commerce: the handwritten logbook. Ubiquitous in neighbourhood sari-sari shops and mall-based food stalls alike, this analogue method of recording sales remains popular for good reason. It is inexpensive, dependable, and resilient—well-suited to the fast-paced and sometimes chaotic environments of small kitchens or cramped storerooms where electronic devices may be impractical.

Yet while paper logs excel in simplicity, they pose significant challenges when it comes to extracting meaningful data. Calculating totals, spotting patterns in customer demand, or analysing product performance from handwritten notes can be time-consuming and error-prone. This is precisely where AI shows its promise—not as a replacement, but as a digital partner. By automating the laborious task of data extraction and analysis, AI can empower small business owners with insights that would otherwise be out of reach.

Despite this potential, many entrepreneurs are understandably wary of digital solutions. Concerns over technological complexity, cost, and potential job displacement often hinder adoption. In response, the BUILD team has proposed a “copilot” model in which AI complements human decision-making, allowing workers to remain firmly in control while benefiting from the speed and analytical capabilities of machine learning.

This concept was recently showcased at the 2025 Artificial Intelligence in Human-Computer Interaction Conference held in Sweden. The Ateneo team demonstrated a prototype system designed to digitise and interpret handwritten sales logs using a combination of optical character recognition (OCR) and large language model (LLM) technologies. Built in Python, the system employs Amazon Web Services for text recognition and Anthropic’s Claude 3 Haiku LLM to make sense of the scanned data. Field tests were conducted at a food stall located in the university’s Student Enterprise Center.

The resulting tool offers a user-friendly interface that even individuals with little to no digital training can navigate. By simply uploading photographs of a handwritten ledger, users receive a readable digital breakdown of sales data. The system identifies product names, tallies quantities, matches prices, and compiles summaries of daily or weekly sales activity. Such insights can dramatically ease the task of inventory management, helping stall owners swiftly pinpoint fast-moving items, slow sellers, and opportunities for restocking or repricing.

Although still in its early stages, the prototype has demonstrated promising accuracy and is adaptable to a range of other paper-based records, including stock inventories, delivery logs, and payroll sheets. Its simplicity and scalability are key features: the tool is designed to be lightweight, low-cost, and continually upgradable as it learns from a broader range of handwriting styles and local terminology.

In many ways, the AI system mirrors the strengths of the traditional logbook it seeks to augment. It is robust, flexible, and shaped with the needs of the everyday user in mind. As the technology matures, tools like this could provide small businesses—long excluded from data-driven decision-making—with access to powerful insights previously reserved for larger corporations. In doing so, it supports a vision of inclusive technological progress, where innovation uplifts rather than displaces, and where human judgement remains central in an increasingly digital world.

More information: Zachary Matthew Alabastro et al, Applied Optical Character Recognition and Large Language Models in Augmenting Manual Business Processes for Data Analytics in Traditional Small Businesses with Minimal Digital Adoption, Artificial Intelligence in HCI. DOI: 10.1007/978-3-031-93429-2_18

Journal information: Artificial Intelligence in HCI Provided by Ateneo de Manila University