# Can AI Uplift Entrepreneurs That Traditional Banks Reject? | Mercedes Bidart | TED Source: https://www.youtube.com/watch?v=rGOTVtXkaRs Recap page: https://rapidrecap.app/video/rGOTVtXkaRs Generated: 2025-12-30T16:37:46.097+00:00 --- ## Quick Overview Mercedes Bidart successfully developed AI models that leverage non-traditional data sources like text messages and social media activity to assess the creditworthiness of informal entrepreneurs in Latin America, enabling financial inclusion where traditional banking systems fail them due to lack of formal records. **Key Points:** - Bidart grew up in a family of small business owners in Argentina and chose to study political science instead of continuing the family business, eventually pursuing AI research at MIT focused on economic development. - Traditional financial systems reject half of the Latin American population who are informal entrepreneurs because they lack formal credit history or bank accounts, making them invisible to conventional risk assessment models. - Bidart and her team developed a system that uses non-traditional data—specifically text messages (Text Score) and social media presence (Social Score)—to create a financial identity for these entrepreneurs. - The Text Score analyzes factors like bill payments, order confirmations, and mobile charges, while the Social Score evaluates online presence and engagement, allowing them to assess repayment probability. - The resulting AI models can predict loan repayment metrics like amount, timing, and conditions, achieving accuracy levels above market standards (e.g., 0.83 vs. market standard) by analyzing this alternative data. - This AI-driven approach allows for the offering of tailored financial services, such as loans, to those previously excluded by formal banking systems, proving that credit history can improve in months, not years. - The core philosophy is that for AI to be fair, it must learn from everyone, not just those represented in existing, often biased, formal data sets. ![Screenshot at 04:44: The speaker illustrates the problem of financial exclusion by showing a slide stating 'Being poor is expensive,' highlighting the systemic barrier faced by entrepreneurs who cannot access credit without formal documentation.](https://ss.rapidrecap.app/screens/rGOTVtXkaRs/00-04-44.jpg) **Context:** Mercedes Bidart, inspired by her family's small business background in Argentina and later her studies at MIT, addresses the critical problem of financial exclusion faced by informal entrepreneurs in Latin America. These small business owners, who collectively account for 99% of companies and one-third of the GDP in the region, are typically denied loans by traditional banks because they operate outside formal financial structures, lacking credit history or bank statements, rendering them 'invisible' to conventional risk assessment. ## Detailed Analysis Mercedes Bidart details her work creating AI models to provide financial services to informal entrepreneurs in Latin America who are excluded by traditional banks. She recounts growing up around small businesses and her decision to pursue AI at MIT to solve this problem. The core issue is that half the population in Latin America, composed of micro-entrepreneurs, lacks the formal data (bank statements, credit history) required by traditional loan officers, leading to predatory lending practices like 'pay-by-the-day' loans with extremely high interest rates. To combat this, her team built a system using alternative data to establish a financial identity. They developed three scores: the Text Score, which analyzes text messages for indicators like bill payments and order confirmations; the Visual Score, which uses computer vision on uploaded videos and pictures to assess inventory, business activity, and social presence; and the Social Score, which analyzes social media engagement. By training these models on millions of data points from informal entrepreneurs, they can predict loan repayment capacity and terms with high accuracy (e.g., 0.83 score), allowing for the offering of fairer, tailored financial services that traditional systems cannot provide. This approach effectively grants financial visibility to previously invisible entrepreneurs, enabling credit history improvement in months rather than years. ### Personal Context and Problem Setup - Grew up around small businesses in Argentina - Chose political science over family business - Pursued master's thesis at MIT on AI and economic development to solve financial exclusion for informal entrepreneurs - Informal sector represents 99% of companies and 1/3 of GDP in Latin America. ### The Data Problem - Traditional banks reject informal entrepreneurs lacking credit history or bank accounts - Predatory lenders exploit this by charging high interest (e.g., 20% per week) for small, daily needs. ### Solution - Three AI-driven scores were developed: Text Score (analyzing text messages for payments/orders) - Visual Score (analyzing uploaded photos/videos of inventory/activity) - Social Score (analyzing social media presence). ### Results and Impact - Models achieve high accuracy (0.83 score) by analyzing non-traditional data - This allows for predicting loan repayment metrics (amount, terms, conditions) - Enables granting loans to those without formal history, fostering financial inclusion and honoring local knowledge/context. ![Screenshot at 00:11: The storefront of 'ALFOMBRAS ARGENTINAS' illustrating the type of small business the speaker grew up around.](https://ss.rapidrecap.app/screens/rGOTVtXkaRs/00-00-11.jpg) ![Screenshot at 01:22: A slide showing the key concept: 'TRUST: Invisible currency built over time,' emphasizing trust as the foundation for lending where formal data is absent.](https://ss.rapidrecap.app/screens/rGOTVtXkaRs/00-01-22.jpg) ![Screenshot at 02:13: A graphic showing a large portion of the population \(yellow figures\) being excluded from formal credit access \(darker figures\).](https://ss.rapidrecap.app/screens/rGOTVtXkaRs/00-02-13.jpg) ![Screenshot at 04:44: The slide 'Being poor is expensive,' summarizing the economic consequence of financial exclusion for micro-entrepreneurs.](https://ss.rapidrecap.app/screens/rGOTVtXkaRs/00-04-44.jpg) ![Screenshot at 09:56: A slide detailing the '1: Text Score' metric, showing examples of text messages analyzed by the AI model.](https://ss.rapidrecap.app/screens/rGOTVtXkaRs/00-09-56.jpg)