AI holds huge promise for healthcare, but not just on the medical side; many startups are convinced machine learning-based systems can do a lot of good on adjacent tasks such as appointment scheduling and confirmations.

Brazilian startup Carecode is among these AI believers. It’s coming out of stealth with an ambition to reduce healthcare costs and improve medical outcomes by developing AI agents that focus on tasks that happen before and after a medical appointment — and would typically be run by a call center.

“We tend to think that only the moment with the doctor is what matters, but after having spent 10 years in healthcare I realized that those moments [around the appointment] are as important as the medical encounter,” CEO Thomaz Srougi (pictured right in the above image) told TechCrunch.

Srougi comes from a family of doctors, but he’s not one himself; his firsthand knowledge of healthcare comes from founding Dr. Consulta, a private medical service provider scaleup that’s raised some $168 million in funding, and where he remains chairman.

Carecode is still early in its fundraising journey, but already has an impressive cap table: Its $4.3 million pre-seed round was mostly funded by a16z and QED, with participation from Endeavor Catalyst, K50 Ventures, and Latitud Ventures, as well as high-profile figures from Brazil’s tech scene, including Nubank founder David Vélez.

Venture capital in Latin America is still going through a “startup winter,” but — as this pre-seed raise underscores — there is still funding for entrepreneurs with track records like Srougi and his co-founder, Pedro Magalhães, a former CTO at several other startups, including BEES Bank Brasil and Zé Delivery.

“I think that counts a lot for major VCs, especially when the world is upside down,” Srougi suggested, also noting that a16z partner Gabriel Vasquez helped mature the idea for Carecode and move the startup from planning to the execution phase.

Srougi is also drawing on relationships he’s built during his career to find partners that can test Carecode. Early results with one partner it can’t name look promising, per the startup, which said the results suggest its AI agents can do most of the work of a typical healthcare call center for a fraction of the cost, and even go one step beyond by proactively filling canceled slots while leaving employees to take care of the most complex cases.

Importantly for Brazil, Carecode meets users where they are, which is usually on WhatsApp — where it supports both text and audio messages. “That’s really important, because older individuals and the majority of low-income individuals prefer to send WhatsApp audio instead of typing,” Srougi said, adding that voice calls are also on the roadmap.

These localization tweaks are one aspect that makes Carecode different from U.S. benchmarks such as Sierra, the AI startup co-founded by Bret Taylor.

Another difference is Carecode’s vertical focus. According to QED partner and head of Brazil, Camila Vieira Fernandes, this gives the startup an edge over horizontal approaches which “often necessitate multiple solutions to achieve subpar results, negatively impacting customer experience and leaving significant value untapped.”

Market size can be a limitation to a vertical model, but healthcare in Brazil isn’t exactly a small niche, and neither is the problem Carecode is going after. According to Srougi, healthcare companies in Brazil spend 50% of their revenue on contact centers and administrative payroll — some $100 billion annually.

Srougi and his team believe that going vertical in a market with specific requirements like healthcare will help Carecode build a moat compared to more generalist competitors, but the startup could also diversify later on. “We may be able to go in the future to insurance, for instance life insurance and other sectors related to healthcare. We want to tap into payments. We may tap into financing. So all of that derives from healthcare,” he added.

TechCrunch has an AI-focused newsletter! Sign up here to get it in your inbox every Wednesday.