Who Will Win the AI Race?

In his book, AI Superpowers, Kai-Fu Lee talks about how China and the US are battling for global supremacy using AI as the main weapon. Lee knows a bit about tech and AI, he’s a computer scientist and worked at Apple, Microsoft and was President of Google China and founded Sinovation Ventures, an AI-focused VC. 

According to Lee, there are three pillars on which AI is built: knowledge, computing power, and data.

Knowledge used to be the key component that gave the US advantage. Having the best engineers and data scientists from all over the world working for American companies gave it the advantage. But as knowledge grows, this tends to become a commodity.

Computing power is based on chips and servers, which have become accessible to any company or government in the world. The almost trillion-dollar valuation that Nvidia has reached validates this.

Finally, data. We’re entering an execution era, where any minimally qualified data scientist or engineer with open source software and access to AWS or any other cloud service with AI features can build incredible AI products, from chats, image editors, business plan builders, or even personal finance advisers.

This is where China is winning. While American tech companies like Google, Meta, and Netflix are focused on generating more data through their products to increase product usage and ultimately revenue (increase stock price), which is a short-term goal with extremely narrow thinking. At the same time, Chinese companies like Tencent, Alibaba, and Baidu are focused on increasing their digital and physical presence in China, creating products to become part of their users' daily lives, including super-apps that offer entertainment, transportation, loans, and payments, among many other things.

Imagine a person using their app to schedule a visit to the doctor, then call a taxi to get there, then pay for the doctor, find the nearest drugstore through their app, and pay for those drugs using the same app.

Clearly, that person is sick, and maybe offering them health insurance isn’t crazy. Now, imagine 1 billion people doing that every day.

This is where China is winning the AI war. By creating enormous amounts of high-quality data about all aspects of a person's daily activities (hello Big Brother!).

In my opinion, the closest to these Chinese companies in Latin America are banks. Almost everything we do is connected to payment and this creates an amazing footprint that very few companies in the region can match. Maybe this is why every company is becoming a fintech company.

Once a month you pay Netflix and Spotify, maybe you do your groceries once a week, and every day you buy lunch at work or go to the gym (which is usually paid once a month). As soon as you receive your paycheck, you pay the loan for the car you bought a year ago and you save a percentage of your salary for the house you want to buy next year. Every Friday night, you get a taxi to go out for dinner with friends and on Saturdays, you go to see a movie with your family.

You get the idea. Banks and fintech have their customers’ lifestyle footprint and could know exactly what type of person, lifestyle, or life stage you’re at. I say they could because very few of them are really using this data. The main problem isn’t quantity but quality.

Digital transactions are dirty and full, with codes and legal business names that don’t match the commercial name that customers recognize. It doesn’t matter if the payments processor is an old-school one with legacy technology or a new one with modern tech, dirty payment transactions in Latin America are among the biggest barriers for AI in banking. 

Working with dirty and unstructured data is the biggest barrier for AI in banking. I understand that there are other types of data that banks can use to train their AI models, but transactions are in my opinion the best way to understand who your customers are and to identify the next best action that can drive user engagement and long-term relationships.

That’s where transactional enrichment comes in. Enrichment is the process of adding value to something that’s raw; in this case, transactions, which are just a string of unrecognizable information that data analysts in banks take days or weeks just to make a data set useful for analysis. Where you transform CGF15884STARB1 4063556 into Starbucks with the merchant name, logo, category, and additional tags like URL, location, and payment method.

So the follow-up question to close this post is “How is your bank or fintech using your customer's transactions to know them better and offer them personalized experiences?”.

If you want to be the Wechat of banking in Latin America, how strong is your data game?