Get Your Weekly Digest

Receive the best podcast summaries and insights delivered to your inbox every week.

We respect your privacy. Unsubscribe anytime.

TED Talks Daily

Can AI uplift entrepreneurs that traditional banks reject? | Mercedes Bidart

with Mercedes Bidart

Published November 29, 2025
View Show Notes

About This Episode

Impact entrepreneur Mercedes Bidart explains how informal entrepreneurs across Latin America are highly trusted within their communities yet are excluded from formal banking because they lack conventional financial records. She describes an AI-driven approach that transforms alternative data from phones, telecom records, videos, and social media into financial identities and risk scores, enabling micro-business owners to access fair, tailored credit instead of relying on violent, predatory lenders. Over three years, these models have reached market-level accuracy and helped tens of thousands of entrepreneurs gain access to formal loans, illustrating how AI can make finance more inclusive when designed intentionally.

Topics Covered

Artificial intelligence Financial technology Small business financing Poverty and inequality Business growth and scaling Data-driven decision-making Informal economy Financial inclusion Microfinance Latin America

Disclaimer: We provide independent summaries of podcasts and are not affiliated with or endorsed in any way by any podcast or creator. All podcast names and content are the property of their respective owners. The views and opinions expressed within the podcasts belong solely to the original hosts and guests and do not reflect the views or positions of Summapod.

Quick Takeaways

  • In many Latin American neighborhoods, trust and reputation function as a powerful informal currency, yet this is invisible to traditional banks.
  • Half of Latin America's population is excluded from formal credit, forcing many micro-entrepreneurs into predatory lending with extreme interest rates.
  • Bidart's team built their own dataset on informal entrepreneurs because conventional financial data did not exist for this population.
  • They use AI models on telecom data, short videos of businesses, and social media activity to infer income patterns, business robustness, and willingness to pay.
  • Starting with $10 loans allowed them to both help entrepreneurs refill inventory and gradually train accurate risk models.
  • Their models now reach an accuracy above 0.83, have been trained on over 150,000 data samples, and have served more than 26,000 entrepreneurs.
  • AI, when designed with intention, can not only improve efficiency but also make financial services fairer and more personalized.
  • The approach turns millions of small behavioral signals into a live financial monitor so entrepreneurs no longer have to wait years to build a traditional credit history.

Podcast Notes

Podcast introduction and framing of the topic

Host introduces TED Talks Daily and herself

Elise Hu identifies the show as TED Talks Daily and notes its purpose[1:51]
• She says the show brings listeners new ideas to spark curiosity every day

Introduction of the problem faced by small business owners

Success is not one-size-fits-all, but traditional metrics still dominate access to finance[2:06]
• Elise notes that small business owners who find success outside traditional metrics struggle to secure bank loans

Introduction of Mercedes Bidart and the core idea

Host describes Bidart as an impact entrepreneur[2:13]
• She says Bidart shows how AI tools in Latin America are helping make the "currency of trust" visible and verifiable
• This is framed as opening doors to fairer finance for those excluded by traditional banking

Mercedes Bidart's personal background and early observations about trust-based economies

Family business upbringing in Argentina

Bidart grew up in a family of small business owners[2:29]
• Her parents ran a curtain and carpet shop in Argentina
She witnessed the challenges of growing a small business[2:43]
• She notes that trust and support from the community were key to keeping her parents' business alive

Decision to pursue political science and technology rather than the family business

Bidart chose not to continue with her family business[2:54]
• Instead, she studied political science, driven by a desire to change the world
Shift from policy to technology[5:37]
• She initially thought she would create change through policy but realized policy did not move at the speed people needed
• She turned to technology because it does not recognize geographic boundaries

MIT thesis and move into real-world pilots

MIT master's thesis on AI and economic development[3:12]
• In 2019, her master's thesis on AI and economic development at MIT was awarded funding to become a real-world pilot
Work in informal settlements in Colombia[3:19]
• The thesis funding led her to work in informal settlements in Colombia

Observation of trust-based local economies in Colombian neighborhoods

Informal credit at local shops[3:28]
• In the neighborhoods she studied, people did not need a credit card to buy lunch; it was enough for the shopkeeper to know who they were
• Shopkeepers would front rice, sugar cane, or bread based on knowledge of the customer and their family
Community-based criteria for informal credit[3:28]
• Factors like a mother's good record with loans, whether someone greeted others in the morning, or whether their shop was known by neighbors influenced trust
Trust as an "invisible currency"[3:52]
• She describes the economy as not running solely on cash but on trust, an invisible currency built over time

Recognition of similar trust dynamics between Argentina and Colombia

Parallel between her childhood in Argentina and Colombian businesses[4:03]
• She noticed that the same principles she observed growing up in Argentina were present in Colombian businesses
Trust as the strongest currency in many Latin American neighborhoods[4:06]
• She states that in many Latin American neighborhoods, trust-a good name-is the strongest currency

The contradiction: trusted locally but rejected by banks

How trusted community members are treated by formal banks

Bank rejection of informal entrepreneurs[4:17]
• When someone who is trusted locally goes to a bank asking for a loan to grow their business, they are rejected
• Banks tell them they lack a "collection" or financial history and claim there is no way to prove who they are
Scale of exclusion in Latin America[4:34]
• Bidart says that in Latin America, half of the population is excluded from formal credit

Central question driving Bidart's work

Connecting neighborhood creditworthiness to bank creditworthiness[4:57]
• She has dedicated her life to answering: What if what makes you creditworthy in your neighborhood could also make you creditworthy in the eyes of a bank?
Incorporating personal reputation into risk assessment[5:07]
• She asks what if your word could be part of the risk assessment for a loan
Scaling capital access by measuring potential[5:08]
• She wonders whether access to capital can be scaled by making a person's potential measurable
Core idea: measuring trust with AI[5:15]
• She explicitly poses the question: What if trust could be measured with AI?

From local marketplaces to recognizing financial exclusion

Early tech project at MIT: building local marketplaces

Goal of local marketplaces[5:58]
• At MIT, Bidart and classmates worked on platforms where community members could upload what they were selling and become visible locally
Field visits to help businesses[6:06]
• They visited businesses to help them upload more product pictures and become better known to increase sales

Discovery that visibility was not the main barrier

Entrepreneurs lacked money to buy supplies[6:26]
• Businesses reported they were not growing sales because they did not have enough money to buy more supplies
Lack of working capital prevented inventory expansion[6:36]
• Even after running businesses for years, entrepreneurs could not access working capital to increase inventory
Shift from visibility problem to financial exclusion problem[6:45]
• Bidart realized they were not facing a visibility problem but a financial exclusion problem

The high cost of poverty and predatory lending

Being poor is very expensive

Higher unit costs when buying in small quantities[7:01]
• She emphasizes that being poor is very expensive because products cost more when they are bought in small quantities
• If someone cannot afford a whole bottle of shampoo, they buy a sachet and pay more per unit
• If they cannot buy groceries for a week, they buy daily and again pay more overall

Limited credit options for those without financial history

Reliance on predatory lenders[7:28]
• Without a great history or bank account, people are limited to predatory lenders, called gota-gota or loan sharks
• These lenders do not require paperwork but charge brutal costs
Extreme interest rates and violence[7:38]
• She notes they can charge 20% interest per week, even per day, and they are violent and abusive

Case study: Maria and the reality of micro-businesses in Latin America

Introduction of Maria, a Venezuelan migrant entrepreneur

Maria's business making handcrafted bags[7:54]
• Maria is a Venezuelan migrant living in a low-income neighborhood in Colombia who makes beautiful handcrafted bags
Need for upfront capital to fulfill custom orders[7:58]
• She receives custom orders and must buy materials before she can produce and get paid

Barriers Maria faces in accessing finance

Lack of bank account and credit history[7:25]
• As a migrant, Maria does not have a bank account or credit history
Dependence on predatory lenders for materials[8:19]
• Her only option to buy materials is to borrow from dangerous predatory lenders

Scale of micro-businesses in Latin America

Maria as representative of millions[8:34]
• Bidart explains that Maria is not an exception but the rule, representing millions of micro-business owners
Prevalence and economic importance of micro-businesses[8:38]
• Micro-businesses include corner shops, restaurants, and beauty salons
• She states that almost every business in Latin America is a micro-business; 99% of businesses are micro
• These micro-businesses contribute one-third of the region's GDP
Ongoing exclusion from formal banking[9:55]
• Despite their contribution, micro-businesses cannot access even one dollar from a bank because they lack the paperwork the financial system requires

Recognizing the opportunity in mobile phone ownership

Distinguishing between formal records and digital traces[9:55]
• Bidart notes that Maria may not have a credit history or bank account, but she has a phone
Strategic shift: change how people are seen, not who they are[9:20]
• She says they saw an opportunity not to change who these entrepreneurs are, but to change how they are seen by financial systems

Building a dataset for the invisible informal economy

Challenge: lack of existing data on informal entrepreneurs

AI models require prior data[9:36]
• Bidart highlights that AI models can only predict what they have already seen
Informal entrepreneurs are invisible in existing datasets[9:55]
• She explains that informal entrepreneurs leave no formal records, making them invisible to existing systems
Need to build a dataset from scratch[9:49]
• To help this population, they realized they needed to create their own dataset of informal entrepreneurs

Traditional banking risk assessment vs. AI-enabled approach

Manual, subjective risk assessment by bank officers[10:06]
• In traditional banking, a risk officer visits the borrower's home, checks the business in person, and talks to neighbors
• Decisions are based on the officer's experience, which brings bias, subjectivity, and slowness

Seeing hidden economic signals in images from local marketplaces

Discovery of rich information embedded in product images[10:40]
• While building local marketplaces, they saw that uploaded product images contained many economic signals
• From images they could detect if there were customers in the background, whether products were handmade, and the potential for the product or service in that neighborhood
Mismatch between available data and what banks are trained to read[11:00]
• She explains that the data existed but not in the format banks had been trained to read

Starting small: micro-loans and intentional data collection

Pilot loans to build the dataset

Issuing very small loans[11:08]
• They started by giving out $10 loans, enough for entrepreneurs to refill inventory
• These small loans were also sufficient for them to begin building and expanding their dataset

Ensuring inclusivity in model training

Gender balance in loan recipients[11:20]
• They were very intentional about who received loans; half of the people they served were women
Fairness requirement for AI models[12:00]
• Bidart argues that if AI is to be fair, it must learn from everyone, including women like Maria

Transforming phone and digital activity into a financial identity

Types of digital traces informal entrepreneurs generate

Maria's digital footprint[11:32]
• Maria has a Facebook page where she uploads products she sells
• She receives text orders and has had her phone for years, storing videos of her products

Creating AI-powered scores from "invisible" data

Suite of AI scores to build a financial identity[11:58]
• Bidart's team built a suite of AI-powered scores that transform invisible data into a financial identity
• She notes they process multiple types of data but focuses on three proprietary scores they developed

Three proprietary AI scores: telecom, video, and social media

Telecom-based score using text messages

Using shortcode SMS as a proxy for financial activity[12:23]
• One main score analyzes shortcode text messages for bill payments, payment confirmations, mobile recharges, and transactions in digital wallets or bank accounts
Detecting income and spending patterns via ML and LLMs[12:40]
• Using large language models and machine learning, they detect patterns of income, spending, and disposable available balance per month
Analogy to open banking using telecom data[12:48]
• She describes it as a kind of open banking, but instead of bank account data they use telecom data

Video-based score replacing physical site visits

Replacing in-person visits with one-minute videos[13:09]
• They replace expensive and time-consuming risk officer visits with one-minute videos that users send showing their businesses
• In the videos, entrepreneurs explain what they are doing in their business
Information extracted via computer vision and audio analysis[13:13]
• Using computer vision, they infer stock levels, inventory, tone of voice, business description, location, type of business, and its potential
Estimating willingness to pay from video data[13:26]
• She states that they are detecting borrowers' willingness to pay from these videos

Social media-based score assessing online presence

Leveraging informal businesses' online presence[14:06]
• Bidart notes that most businesses, even informal ones, are present online through Facebook or Instagram
Connecting to borrowers' social media at application time[13:46]
• When applicants apply for a loan, they sign in with their social media, allowing access to their videos and pictures
Using engagement metrics and profile data to infer reliability[13:58]
• They use computer vision on the media and also collect likes, comments, engagement, and profile bios
• They have detected that a business with strong social and online presence has higher probability of paying back

From decision to personalization: how the models are used

Combining signals to assess trustworthiness of new borrowers

Aggregating multi-source data into risk decisions[14:12]
• All data flows into their models, which detect patterns and signals to determine whether a person can be trusted with a loan, even if they never had one

Beyond yes/no: tailoring loan terms with AI

Speed and depth of automated decisions[14:30]
• After three years, they can not only answer yes or no in seconds but also estimate how much a borrower can repay, when, and under what conditions
Simulating loan conditions and seasonal impact[14:47]
• Their models simulate interest rates, number of installments, and detect seasonal impacts
Hyper-personalized, needs-based credit products[14:54]
• This enables them to offer credit products that support people's everyday needs and are tailor-made, rather than a single product for everyone

Validation, scale, and systemic impact

Model performance and training data scale

Achieved accuracy and market standards[15:25]
• Bidart says their business and models have reached an accuracy level above 0.83, which is at market standards
Number of entrepreneurs served and data samples used[15:39]
• They have served more than 26,000 entrepreneurs
• Their models have been trained with more than 150,000 data samples of informal entrepreneurs and millions of data points

Changing the financial system's treatment of the informal sector

Impact beyond individual entrepreneurs[15:39]
• She stresses that the impact goes beyond supporting individual entrepreneurs and their families; it is changing the financial system
From multi-year credit histories to live financial monitors[16:00]
• What used to take years to build, such as a credit history, can now be created in months
• They are building a live financial monitor of financial well-being that can be updated daily
• This live monitoring removes the need to wait years to be eligible for a loan
Opening formal banking to informal entrepreneurs[16:15]
• This approach is allowing the informal sector to access loans from the formal banking system for the first time

Reflections on AI as a tool for fair and contextual finance

Clarifying what AI is and can do

AI as a non-magical tool[16:25]
• Bidart emphasizes that artificial intelligence is not magic but a tool
Processing scale beyond human capacity[16:37]
• AI helps process millions of data points that no human risk officer could read, watch, or analyze at scale

From efficiency to fairness through intentional design

Efficiency is only the baseline benefit[16:38]
• She notes that AI is, of course, improving efficiency
Intentional design makes AI fair and inclusive[16:45]
• If designed with intention, AI becomes more than efficient; it becomes fair
• Fair AI allows systems to see value where others previously saw risk
Shifting perception: seeing gold where others saw stones[16:54]
• She uses the metaphor that AI allows us to see gold where others saw stones
Balancing scale with respect for local context[17:04]
• AI is allowing services to be offered at scale while honoring local knowledge, culture, and context
• It enables hyper-personalization of financial services

Reframing who is considered creditworthy

Saying yes to overlooked entrepreneurs[17:22]
• She says this enables lenders to say yes to someone like Maria and to someone like her mother when she started a business
• It also allows them to say yes to millions of women entrepreneurs who are driving the economy forward
Basing trust on quiet behavioral signals rather than bank statements[17:31]
• The yes is given not because of a bank statement but because of millions of quiet signals indicating that an entrepreneur shows up, delivers, and can be trusted

Outro and TED production credits

Identification of the talk and event

Talk recorded at TED AI in Vienna[17:50]
• The host notes that the talk was given by Mercedes Bidart at TED AI in Vienna, Austria in 2025

Explanation of TED curation

Reference to TED curation guidelines[17:58]
• Listeners are invited to learn more about TED's curation at TED.com slash curation guidelines

Credits for TED Talks Daily production team

Fact-checking and production roles[18:03]
• The talk was fact-checked by the TED Research Team and produced and edited by a named team including Martha Estefanos, Oliver Friedman, Brian Green, Lucy Little, and Tansika Sangmarnivong
Audio mixing and additional support credits[18:17]
• The episode was mixed by Christopher Fasey-Bogan, with additional support from Emma Taubner and Daniela Balarezo

Host sign-off

Promise of future episodes[17:58]
• Elise Hu says she will be back tomorrow with a fresh idea for the audience's feed and thanks listeners for listening

Lessons Learned

Actionable insights and wisdom you can apply to your business, career, and personal life.

1

Critical economic value often resides in "invisible" forms of trust, reputation, and behavior that traditional systems ignore, but which can be systematically captured and honored with better tools and data.

Reflection Questions:

  • • Where in your work or community do people quietly create value that is not formally recognized or measured?
  • • How could you start documenting or surfacing the informal trust and reliability you and your collaborators already demonstrate?
  • • What is one system you rely on today that could be improved by incorporating more nuanced, behavior-based signals of trust?
2

When conventional infrastructures exclude people, a powerful strategy is to change how they are seen rather than trying to change who they are, by translating their existing behaviors into formats institutions can understand.

Reflection Questions:

  • • Who in your ecosystem is doing good work but is being overlooked because they don't fit standard metrics or formats?
  • • How might you reframe or repackage your own track record so that gatekeepers can better recognize your actual capabilities?
  • • What is one concrete step you could take this month to make an underappreciated group or person more legible to decision-makers?
3

Starting with very small, low-risk experiments can both deliver immediate value to users and generate the data needed to build more sophisticated, scalable solutions.

Reflection Questions:

  • • What is a small-scale version of your current idea or project that could still meaningfully help a few people?
  • • How could you design your next experiment so that it not only tests your concept but also produces reusable insights or data?
  • • What is one tiny, reversible bet you could place this week to move your initiative from theory into practice?
4

AI amplifies the intentions and data it is built on, so designing it with explicit goals of fairness and inclusion is essential if you want it to expand opportunity rather than entrench bias.

Reflection Questions:

  • • In the tools and processes you currently use, whose experiences or data are missing from the picture?
  • • How would your decisions change if you deliberately optimized not just for efficiency and profit but also for fairness and access?
  • • What is one practice you could adopt to regularly audit for bias or blind spots in your own decision-making or systems?
5

Hyper-personalization-adapting services to individuals' real patterns, constraints, and seasons-can dramatically increase the relevance and effectiveness of support compared to one-size-fits-all offerings.

Reflection Questions:

  • • Where are you currently offering or receiving generic, one-size-fits-all solutions that don't fully match real needs?
  • • How could you use the information you already have about your clients, team, or audience to customize support just a bit more?
  • • What is one process, product, or relationship in your life that you could redesign this week to better fit the actual context and rhythms involved?

Episode Summary - Notes by Alex

Can AI uplift entrepreneurs that traditional banks reject? | Mercedes Bidart
0:00 0:00