What AI can give South American companies, and their people

The race for AI

Building the most capable AI has become a race, and governments have stepped into it. Between August and September alone, five labs in the United States and China shipped new models. On September 1, Anthropic released Fable 5.1. Two days later, OpenAI launched GPT-6 “Astra” and said it may represent artificial general intelligence.

On June 12, shortly after Anthropic released Fable 5, the US government issued an export-control directive suspending access to it for any foreign national. To comply, Anthropic had to switch the model off for every customer, everywhere. On June 26, at the government’s request, OpenAI limited its new GPT-5.6 to a small group of trusted partners. Fable 5 came back on July 1, and OpenAI opened GPT-5.6 to the public on July 9. In both cases, who could use the model, and when, came to depend on a government.

The competition has turned into a moral argument, too. On February 25, Anthropic refused to sign a new Pentagon agreement without written bans on fully autonomous weapons and mass domestic surveillance. On the 27th, the President ordered federal agencies to stop using Anthropic’s technology, and on the 28th OpenAI announced its own Pentagon deal. Sam Altman said his company shares those limits. OpenAI, as MIT Technology Review reported, relies on existing law rather than on prohibitions written into the contract. Months earlier, in July 2025, OpenAI, Anthropic, Google and xAI had all signed prototype contracts with the Pentagon. One disclosure: Ekua’s agents run on Anthropic’s models today.

Many people look at AI with suspicion. In North America, an Ipsos survey taken in March and April finds more people nervous than excited. Latin America is more excited than nervous, but almost half of the region is nervous too, and with reason.

Some of those reasons are concrete. The models the region’s companies rely on are decided far away, and June showed that one can disappear overnight on another government’s order. The region uses a lot of AI, but the investment goes elsewhere. CEPAL, the UN’s economic commission for Latin America and the Caribbean, finds that the region accounts for 14% of global visits to AI tools with 11% of the world’s internet users, yet receives 1.12% of global AI investment while producing 6.6% of global GDP. And the models fail more often than their launches suggest. In a study of ten models, a preprint published in March, the share of tasks done right on the first try falls from 76% to about 50% as tasks get longer.

Then there is the question that weighs most at any kitchen table, which is work. Over the next few years, the people who run the region’s companies will decide which technology to adopt, what for, and who to build it with. In doing so, they will also decide what place their own people have in what comes next.

Work and its dignity

In this region, the Church also has something to say about work, and what it says is part of the public conversation. Believer or not, its questions deserve an answer from those of us who build this technology.

Pope Leo XIV chose his name partly after Leo XIII, whose encyclical Rerum Novarum answered the first great industrial revolution. His first encyclical, Magnifica Humanitas, signed on May 15, 2026, is about safeguarding the human person in the age of artificial intelligence. The line that has stayed with us most is in paragraph 149: “work is not simply an instrument; it expresses and enhances the dignity of our lives.”

If work expresses the dignity of a life, an AI agent should take away the part that wears a person down and leave the part that makes them grow, which is judgment and dealing with other people.

The Pope also writes: “I wish to address a special appeal to those who develop artificial intelligence.” That appeal includes us. He asks them to build “with transparency, responsibility toward affected communities and careful attention to ensuring that what is being cultivated is a genuine good” (n. 111). On employment, he asks for more.

Its first sentence describes what we want our agents to do. The rest sets a limit on employment that applies to an AI company too, and we accept being judged by it. None of this is an endorsement of Ekua or of any company. We quote it because it demands an answer, and the rest of this piece is ours.

On September 12, in an address, Leo XIV said that “the decisive question is not simply whether we accept or reject certain tools, but what kind of humanity we wish to build and what place each person will have within it.” It is the same question now in front of the people who run the region’s companies.

Where we are going

In RSM’s survey of 461 Latin American business leaders, 53% expect AI to improve their people’s skills, against 15% who expect job cuts. We want to prove that 53% right. Put in one sentence, that is our mission.

Every company has its own list of work nobody wants to do, and we want our next agents to come from those lists. That is why this piece ends with a question for you.

What we do

We build AI agents for mid-size and large companies in South America. Each one takes on a repetitive task, the kind that wears a company’s people down and keeps its customers waiting: the fortieth request of the day, identical to the thirty-nine before it, or the figure someone copies by hand from one system into another. That gives the team time for work that needs judgment, and customers stop waiting on things that should be simple.

Every agent we build follows the same rules. They are how we think about AI, and we put them in writing so anyone can hold us to them.

1. An agent says what it is. From the first sentence, the person talking to it knows it is a virtual assistant.
2. An agent only states what it can check. The figures and facts it mentions come from the company’s own systems, and if something isn’t there, it doesn’t make it up.
3. An agent does one task. That is the practical lesson of the study above, since the longer the task, the more often models get it wrong.
4. People make the decisions that affect people, because the team knows the customers. The agent can remind, ask and take notes, but the hard conversation and the final decision stay with the team.
5. An agent has to understand the place where it works, and the region is already building its own tools. In February, Chile’s National Center for Artificial Intelligence (CENIA) presented Latam-GPT, a model it describes as open and trained on Latin American data. The Latam-GPT team has also published Trueque, an open benchmark of 500 questions on 20 countries in the region, Ecuador among them, which its authors present as a beta with limited coverage. We will run Trueque against the model our agents use, to learn where its knowledge of the region falls short.

For now we work with a few companies at a time, and we, the founders, work with each one ourselves.

Our first agent

I’m Ecuadorian, and I know how much this country has to offer. Living in the United States, close to the technology being built there, I caught a glimpse of what it could do for South America. Coming back to Ecuador to sign our first customer, an Ecuadorian company, and build our first agent for them has been the most rewarding part of all this.

They asked for a Spanish voice agent to follow up on pending payments. It is built to call customers with a payment coming due or overdue, remind them of the amount and the date, and work toward a commitment to pay. The idea is for the collections team to spend fewer hours on repeat calls and more on the cases that need a person.

It follows the rules above. In its case, it also cannot change anything in the company’s system or move money. When a conversation gets difficult or someone doesn’t keep a promise, it hands the case to a person on the collections team, who decides what to do. We designed it that way because we believe in it, and because Ecuador’s data protection law (Article 20) gives people the right not to be subject to decisions based on automated assessments that have legal effects on them or affect their fundamental rights.

It is the first of the agents we plan to build.

What we believe

Building AI is a moral decision as much as a technical one. Some of our rules will cost us conversations or customers, and we will keep them anyway, because none of this works without the trust of the people an agent talks to.

We believe AI in South America can give workers more creative freedom and take the tedium out of their day to day. As people, we understand how complex the world is, and we know there are decisions AI cannot make on its own. The change we want to see is something we will build together, and that is what Ekua is about: technology for the greater good, starting with the people whose work it will change first.

We think of the employee who answers the same question all day long, and of the customer who picks up an agent’s call.

We still have to prove it, one agent at a time, and on one condition: that whoever is on the other end always knows they are talking to a machine.