Real-World AI Doesn’t Look Like ChatGPT

How many meeting note-taking AI tools can you think of? We asked ChatGPT to list them, and it gave us forty right off the bat. Don’t get us wrong, we use and love them, but Silicon Valley often dazzles us with breakthroughs designed for speed and comfort — refinements to problems we sometimes didn’t know we had. 

Elsewhere, a different kind of AI revolution has been underway for years: one born not out of convenience, but out of necessity. In regions like Africa, Asia, and Latin America, too many hard problems go unsolved. They are so big and obvious that founders often either experience them firsthand or are close to people who do. 

Emerging markets have the unique opportunity to use that to their advantage and build AI rooted in real-world complexity, tailored to regional languages, customs, and needs. Meet three Endeavor Entrepreneurs whose solutions do just that.

Precision data for alternative credit access

Benjamin Njenga grew up on a one-acre farm in rural Kenya. It was through that piece of land that his mom fed and educated her kids. But like her, small-scale farmers, with no formal education and only inherited knowledge to guide them, often rely on intuition. Some harvests work out; some don’t. What changes that dynamic isn’t luck, it is data.

“I saw the problems my mother went through and I made a vow to find out how to help other smallholder farmers achieve better results on their plots,” says Benjamin. In 2016, he co-founded Apollo Agriculture, which has been an Endeavor Entrepreneur company since 2020. (Our rules-based VC fund, Endeavor Catalyst, backed it in 2022.) 

Leveraging machine learning models and alternative data including GPS, the company offers financial inclusion and support even to those in extremely remote areas. The customer applies for a loan and picks up high-quality seeds and fertilizer at their nearest agrodealer — all automated and cashless. Agronomic training arrives via voice recordings and is supplemented by a dedicated field officer. If pests come along or a drought hits, insurance cushions the blow.

“We have to do this in a super-challenging environment where farmers have no financial records like bank statements,” he explains. Apollo doesn’t require any proof or collateral to give them access to credit. By having precise data on the crops’ performance and focusing on helping farmers succeed, they guarantee their ability to repay loans.

With that, incomes rise and entire communities experience greater food security and stability.

Affordable customization in education

When Vu Van left Vietnam to pursue a dual MBA and Masters in Education at Stanford, she was confident in her English vocabulary. Her strong Vietnamese accent, however, often got in the way. “My opinions in class were often disregarded because professors couldn’t understand what I was saying,” she says. 

Vu soon realized she was not alone. More than 1.5 billion English speakers are non-native, struggling with pronunciation, rhythm, and fluency, leading to loss of work and overall career opportunities. 

Existing solutions didn’t provide personalized feedback, and while speech therapy and private tutoring do, they can also cost a few hundred dollars an hour. She set out to identify partners and talent in AI and voice recognition — specialties often dominated by tech giants — and joined forces with her professor, Xavier Aguerar. Together, they founded ELSA, an acronym for English Language Speech Assistant.

ELSA focuses on real-world scenarios, like small talk at work, job interviews, or professional presentations. The technology identifies pronunciation errors at the phoneme level (like the subtle sound differences between “pad” and “pat”) with 95% accuracy, providing precise and personalized feedback instantly, no matter the user’s accent — something no human tutor could efficiently scale.

ELSA focuses on real-world scenarios, like small talk at work, job interviews, or professional presentations. The technology identifies pronunciation errors at the phoneme level (like the subtle sound differences between “pad” and “pat”) with 95% accuracy, providing precise and personalized feedback instantly, no matter the user’s accent — something no human tutor could efficiently scale.

After launching in San Francisco in 2015, ELSA began experiencing significant momentum and by 2020, joined the Endeavor network, reaching almost 13 million users so far from more than 100 countries.

The company now has offices in Vietnam, India, Portugal, Japan, and Indonesia. In 2021, Endeavor Catalyst was among the investors who participated in their $15M Series B round.

Human English tutors and speech therapists could never personally coach tens of millions simultaneously efficiently or affordably — but ELSA’s AI can.

ELSA is helping millions of learners globally overcome language barriers, opening doors previously locked by resource constraints.

Knowledge building for the environment

Matías Muchnick has always had a passion for food and sustainability, yet he knew these two worlds conflicted when it came to animal products. But Matías wasn’t (and still isn’t) vegan, so he set out to create plant-based alternatives to animal products that didn’t compromise taste and texture compared to the real thing. 

That led to his first startup, the vegan mayo company Eggless, founded in 2012 and sold in 2015. With lessons from the consumer-goods industry, he then founded NotCo in Chile. 

“I hired a company to produce the formulation for me,” he remembers. “It was three guys in lab coats doing trial and error, reading research papers from the 1980s on how to replace animal-based food. It was so obsolete, and this was a company that made food formulations for really big companies around the world.”

He enlisted the help of a computer-science expert and a genetics expert to reimagine the process. Together, they developed a platform that uses AI and machine learning algorithms to reverse-engineer animal products and recreate them with plants. It’s what led them to discover that cabbage and pineapple together could taste like milk, for example, or that tomato plus strawberry equals chicken.

“Taking animals out of the equation was the top driver for me,” he says, “because it is absolutely inefficient.” 

But perhaps the biggest impact is that AI can not only break down the formula, but also share it with the world. Matías and team are causing a shift in mindset and ingredients used by food and beverage manufacturers, now able to leverage NotCo’s B2B AI platform to create their own plant-based products.

The company has operations in Brazil, Argentina, Chile, Colombia, the United States, and Canada, and intends to expand the brand to Europe and Asia. 

Scalable impact

The founders above aren’t building to make life that little bit easier. They are solving the kinds of issues that prevent entire communities from moving forward — problems that are systemic, not cosmetic.

For them, AI is an enabler for scaling impact. They now reach further and faster, tackling challenges that were simply too complex, too large, or too labor-intensive to solve before.

This is a pattern we see across our global community of Endeavor Entrepreneurs. In Mexico and across Asia, Trusting Social and Kueski use AI for mobile usage patterns and alternative data analysis that bring credit access to those with no financial footprint. From Saudi Arabia, Noon Academy helps millions of students learn from top teachers and each other, regardless of geography or income. When AI meets real-world urgency, it unlocks solutions that aren’t just innovative, but transformative.

Being in 45+ countries, we get to see the full spectrum of what AI is capable of, from the strong improvements to the truly life-changing solutions, and support the builders behind both types of value generation. It’s this global lens that shows us that the next frontier of AI is, first of all, human.