Gabriel Monroy, Colektia | Building Artificial Intelligence Out of LatAm

The audio and video are in Spanish. These notes are a translation.

These days, saying your startup uses artificial intelligence has become almost a meme, but a few years ago the reality was completely different.

Today I have as my guest Gabriel Monroy, cofounder and CEO of Colektia, a platform that develops artificial intelligence to optimize collections management for companies such as Nubank, Rappi, Claro, and Telefónica. Colektia has raised close to 80 million dollars in equity and debt from funds such as Angel Ventures, Nazca, and Devlabs.

Colektia was born 6 years ago and was one of the first companies in the world to develop its own artificial intelligence algorithms for collections.

Today Gabriel and I talk about:

How to solve the chicken-or-the-egg problem in artificial intelligence
The most common mistakes when building solutions with AI
How to sell artificial intelligence to companies
And what we can expect from the future of AI in Latin America

This story is proof that you can build proprietary technology from Latin America to solve Latin American problems. I hope you enjoy this conversation with Gabriel Monroy of Colektia.

Youtube Timestamps*

[02:15] - Insight and Collections

How did Gabriel arrive in the fascinating world of startups?

Colektia is born from a very counterintuitive insight, contrary to conventional wisdom: that people do not pay their debts because there is a problem that keeps them from doing so.

It is a simple but powerful idea because it runs against the industry and it is central to Colektia’s success.

Before validating the idea, where did this insight come from? How did Gabriel arrive at it?

The collections industry in Latin America is enormous.

Collecting manually through call centers, or even using threat tactics, seems so obsolete.

When he starts studying the collections industry, what struck him as the most broken or obsolete?

Colektia is a company building proprietary artificial intelligence from Latin America to solve a Latin American problem.

But similar companies are emerging in Africa, Asia, and even the US with similar ideas.

Why Latin America first? What does the collections industry in these countries have that created a fertile breeding ground for a technology company like Colektia to emerge?

[15:52] - How to Use AI

Gabriel has mentioned the chicken-or-the-egg dilemma in artificial intelligence. “You need data to build AI, but without AI I have no customers.”

What did they try, and how did they land that first customer?

Colektia’s insight is an idea that sounds almost too good to be true, to the point that potential customers such as banks or retailers would probably tell them: validate it with others and then come back to us.

How did they convince that first customer to run the pilot? Any lesson from this experience of landing the first pilot that is worth highlighting for other AI founders going through this dilemma?

Gabriel has described Chile as an ecosystem where companies are very open to trying new and innovative solutions. It sounds like starting in Chile was key to Colektia’s development.

How would he describe the level of “early adopter” mindset among large companies in the main countries of Latin America today, especially in the wake of the new AI wave?

It is fascinating when a corporate has a problem so large that it is willing to open its doors and collaborate with a startup to build a solution for its problem.

What did the first version of the algorithm look like? How much co-creation was there with this retail customer?

Gabriel has talked about the many challenges they faced implementing artificial intelligence solutions: from having to automate the collections process because it is very manual and, if you do not, you cannot capture data, to how hard it is to sell something that is not tangible, like an algorithm.

What were the biggest mistakes they made at Colektia building artificial intelligence solutions? The ones that hurt the most, but without which Colektia would not be where it is today.

There are call centers that replaced people with robots and then went back to hiring operators.

When implementing AI in processes and software, what has Gabriel learned about the optimal combination or interaction between artificial intelligence and people?

[39:44] - Business Model Innovation

They started as a software as a service. Then they became a collections agency leveraged on their technology and, recently, they started acquiring portfolios from banks and finance companies.

Why did they start as SaaS, who was the customer at each stage, and what were the reasons that led them to evolve their business model twice?
How do they decide to start refinancing people with a history of not paying their debts? It sounds like a more financial and risky business, versus data and software. Where does the idea come from, how did they test it, and what have they learned that others do not see?

Colektia’s solution runs on its own artificial intelligence algorithms. They are not using Microsoft or Google APIs.

It sounds like an incredible technical milestone, and it shows that this technology can be built with Latin American talent. But, from a business perspective:

Why was it necessary? What is the business logic, or the specific characteristics of the collections problem, that require building proprietary AI algorithms?

A year ago, ChatGPT came out and the wave of large language models (LLMs) exploded:

When does it make sense for a new company to build proprietary artificial intelligence algorithms? Put another way, if he were starting Colektia today from scratch with the new technologies that are available, what would he do differently?

[52:20] - The Future of AI in LatAm

At Colektia they have been working with AI for more than 6 years.

Is this technology more powerful for existing companies that have products with large databases and teams, compared with new startups that, for example, would want to build a Colektia competitor from scratch starting tomorrow and have this new technology at their disposal?

When asked why Colektia had been successful, Gabriel answered that they pulled off 2 things that seemed crazy: 1) developing their own algorithms with Latin American talent, not APIs; and 2) being the first in an industry from Latin America.

These are limiting beliefs Latin Americans hold, and if they do not let them go, they are going to miss the great wave of value creation with artificial intelligence.

How does he overcome these beliefs, and how does he pass that on to his team?

People often criticize Latin American investors for not investing in unique models and for looking for clones.

How did being a startup with few or no comparables globally affect Colektia’s ability to raise capital? How did that show up in his conversations with investors, and how did he handle them?

It is impressive that they had to create their own trainee program in data science and data engineering inside Colektia because they could not find trained talent.

At the business level, it has probably become an unfair advantage because of the access they have to talent, but it is also a symptom of a talent problem.

How would Gabriel describe the state of knowledge about AI and data in Latin America? What would his request be to universities and education companies?

[01:07:48] - Quick Questions

If Gabriel looks 10 years into the future and puts on his dreamer hat, what is the most extraordinary vision for Colektia?