Small Business, Big Tech

There’s a farmer in Honduras who can’t sleep.

He has newborn piglets and is trying to decide when to sell them. A week? A month? A year? The animals will grow in size and value over time, but they’ll also consume more feed, require more care and face greater risk of illness. A buyer may offer less than expected. Markets may shift. Every option carries uncertainty.

For microentrepreneurs like this farmer, nearly every day brings decisions like these. He may never use terms like cash flow, margins or risk tolerance, but he wrestles with those concepts all the same. He knows his animals, his customers and his costs. What he doesn’t have is something executives at larger companies often take for granted: someone to think through those decisions with him.

Marcela Aguilar (Ph.D. ’26) believes artificial intelligence may be able to fill that gap.

As part of her doctoral research at Rice Business, Aguilar developed Mia — short for mentora con inteligencia artificial — a generative AI mentor that operates through WhatsApp, the messaging platform already used by millions across Latin America. Rather than teaching accounting as a standalone subject, Mia helps entrepreneurs apply business principles while making real decisions about pricing, inventory, suppliers and growth.

“The goal isn’t to teach accounting,” Aguilar says. “It’s to help people make better business decisions.”

That distinction lies at the heart of Aguilar’s research, which focuses on one of the world’s largest — and least understood — business communities: microenterprises.

According to the International Finance Corporation, micro, small and medium enterprises account for roughly 90% of businesses worldwide, employ more than 70% of the global workforce and generate about half of global GDP. Yet millions of these businesses operate informally, without access to sophisticated financial tools, business advisers or professional training.

Many owners don’t think of themselves as practicing accounting. They’re simply trying to decide whether to extend credit to a customer, when to buy inventory or how much to charge for a product.

“They don’t want to learn accounting,” Aguilar says. ”They want marketing. They want sales.” Her research suggests those goals may not be separate after all.

Instead of presenting accounting as a series of isolated lessons, Aguilar views it as the language underlying every business decision. Pricing requires accounting. Purchasing inventory requires accounting. Deciding whether to keep piglets another month requires accounting — even if the entrepreneur has never seen a financial statement. The question became whether AI could help entrepreneurs recognize those connections while making decisions in real time.

To find out, Aguilar designed a field experiment in Tegucigalpa, Honduras, working with the International Labour Organization and the local chamber of commerce.

Participants — primarily women operating small businesses — were randomly assigned to one of three groups.

One group received traditional classroom instruction from a human trainer. A second interacted with Mia, which taught the same curriculum through WhatsApp. A third used a different version of Mia that integrated accounting concepts into every business conversation rather than teaching them as separate lessons.

Entrepreneurs could communicate with Mia naturally, sending text messages, voice notes and even photos. Mia responded the same way.

“A lot of people said it felt like a friend,” Aguilar says.

The format solved a problem traditional training often cannot. In a classroom, some participants hesitate to ask questions, especially when educational backgrounds vary widely. But chatting privately with Mia removes that social pressure. Questions can be asked anytime — whether at noon in a marketplace or midnight after a long day of work.

During the monthlong study, 173 entrepreneurs exchanged more than 30,000 text messages, 3,000 voice notes and 1,500 images with Mia, asking questions about pricing, inventory, customers and everyday business challenges.

The conversations created something researchers rarely have: a window into how entrepreneurs actually think through business decisions as they happen.

The first finding surprised even Aguilar.

Entrepreneurs trained by Mia performed about as well as those taught by human instructors — but at roughly 3% of the cost.

For governments and nonprofit organizations that collectively spend more than $1 billion each year on entrepreneurship training, the implications are substantial. AI cannot replace the relationship-building and shared experiences of classroom instruction, Aguilar says, but it may dramatically expand access to high-quality business guidance.

The second finding proved even more significant.

Entrepreneurs using the version of Mia that embedded accounting into everyday decision-making consistently outperformed those who learned accounting as a separate topic.

The result reinforces Aguilar’s central insight: entrepreneurs don’t become better business owners by memorizing accounting concepts. They become better decision-makers when those concepts are woven into the choices they’re already making.

“It’s like when we’re in school and someone teaches us math, and we don’t know how to apply it,” Aguilar says. “But if you teach mathematics for making decisions — in your specific context — you transfer that knowledge from the classroom to the real world.”

The approach appears to produce results that extend well beyond the study itself.

Since participating in Aguilar’s study, roughly 30% of the businesses have begun formally registering with the government — a first step toward accessing bank financing, legal protections and new opportunities for growth. After the research concluded, the chamber of commerce in Tegucigalpa has continued offering Mia as a service to local entrepreneurs.

The Future of Business Research

Aguilar sees Mia as more than an AI chatbot.

Every interaction creates data — not just about business outcomes, but about how entrepreneurs learn, question assumptions and make decisions. Thousands of conversations reveal patterns that traditional surveys could never capture.

The technology also raises important questions. Like any large language model, Mia can hallucinate or generate incorrect information. Aguilar emphasizes that AI should support human judgment, not replace it.

“At the end of the day,” she says, “it has to be the entrepreneur making the decisions.”

Her next study will expand Mia into Honduras and Costa Rica, testing a hybrid model that combines human instructors with AI mentors.

“In the ideal world, you have both working together,” Aguilar says. “You have training in a classroom but can ask questions at any moment to the AI.”

For Aguilar, the project ultimately returns to the question that first drew her to economics years ago: how to improve people’s lives through better decisions.

Millions of entrepreneurs around the world already possess the determination, expertise and resilience to build businesses under extraordinarily difficult circumstances. What many lack is someone to help them think through the next move.

With Mia, Aguilar is exploring whether that trusted adviser might one day fit inside a WhatsApp conversation.