MAIA, MYPE AsesorIA, is an initiative by researchers focused on entrepreneurship, small business productivity, and the practical application of artificial intelligence, funded by international donors.
Two of Innovation for Poverty Action's 14 "Best Bets" to improve the lives of people living in poverty are training and consulting services for micro, small and medium sized enterprises (MSMEs) in developing countries. McKenzie & Woodruff's meta-analysis shows that training improves MSME performance, with the impact of consulting services particularly large. But human-led training and consulting programs are very expensive and difficult to scale. As the literature puts it: "There is a need for further experimentation with alternative delivery methods, particularly online training".
Generative AI offers the potential to deliver support that is customized, interactive, timely, and continuous, at a fraction of the cost. It could deliver personalized training and consulting, automate specific tasks, and provide practical tools, scaling valuable assistance to hundreds of millions of small business owners globally.
But it isn't as simple as just giving MSMEs ChatGPT. We need an accessible on-ramp for entrepreneurs who have limited comfort with technology. We need to take the world's most powerful LLMs and adapt them to this purpose, by:
There are many ways this can be done, and it isn't yet clear what works best. Should the focus be on hard skills or on soft skills like personal initiative? What is the right balance between expert coaching and enabling owners with AI tools they can use directly? Should support focus on immediate quick wins, or longer-term goal setting? What is the right dosage: daily chats, weekly check-ins? The best way to figure this out is through rigorous testing and direct user feedback.
That is how MAIA runs today: continuous pre-registered randomized experiments on coaching design, an evaluation pipeline informed by the staged approach of the Center for Global Development and The Agency Fund, longitudinal outcome tracking, and new behavioral measures of whether coaching actually gets executed. See the measurement pipeline.
An AI coach is only useful if its advice is sound. MAIA's answers are grounded in country-specific knowledge bases built from validated local sources: ministry guidance, partner training curricula, and field experience, by sector and crop. Real extension agents and entrepreneurs test and adjust the tool iteratively, and live web grounding backs answers about current rules and programs. MAIA's guidance is advisory: it does not replace professional advice or issue binding decisions, and we say so to users. We monitor quality continuously and publish what we learn, including the hard parts, like how hallucination risk has evolved.
MAIA is part of Hacer Perú's AI & MSME Productivity Initiative, implemented with the ILO's FORLAC programme.
In 2026, MAIA was selected from more than 1,500 applications for the bootcamp of the Moonshots for Development (M4D) Open Innovation Challenge.
Contact: Bailey Klinger, project lead (maia@hacerperu.pe)
We dedicate our intellectual property, including all code, documentation, and related material, to the public domain. We make this dedication for the benefit of the public at large and to the detriment of our heirs and successors. We intend this dedication to be an overt act of relinquishment in perpetuity of all present and future rights under copyright law, to the extent allowed by law.
You can copy, modify, distribute, and perform the work, even for commercial purposes, without asking for permission. The work is offered as-is, without warranties of any kind.
