‘All LatAm banks knocked on our door for generative AI', says Microsoft

All of the major banks in Latin America have spoken to Microsoft to understand how to apply generative AI in their operating processes and these talks have already led to pilots in most cases.

That is according to Eduardo Joia, CTO and managing director for the financial services industry at Microsoft Latin America.

“All the banks, 100% of the financial institutions, especially those that are Microsoft customers in some way, all of them knocked on our door, and all of them are asking us to run pilots,” Joia told BNamericas on the sidelines of the Febraban Tech event in São Paulo.

Demand has been so intense that Microsoft now has a supply issue with graphics processing units (GPUs) for its datacenters and trouble finding staff to handle the projects.

“Today, if you take OpenAI's neural network, the only private cloud company that can offer it in a corporate domain is Microsoft," said Joia.

The executive refers to the private (corporate domain) rather than the public open version of ChatGPT developed by OpenAI, a company in which Microsoft is an investor and partner.

The corporate version offers additional layers of security and confines data to the company's own network, avoiding security risks and increasing information protection, according to the executive.

In this version, the company controls the documents being generated and trained in a Virtual Private Cloud (VPC), more specifically, Microsoft's Azure cloud.

CAUTION

More powerful and natural chatbots, coding and programming, and internal product training are the main applications that Latin American banks are testing with generative AI, said the executive.

However, they are being cautious.

Most institutions applying ChatGPT are using filters and internal tests before putting the platform into contact with customers, even for chatbots, said the executive.

“We were seized by the euphoria and we went out applying generative AI until our group of internal researchers flagged us about biases, data protection and other issues,” Marcelo Leite Ferreira, Banco do Brasil’s general IT manager, said during a ChatGPT panel at the Febraban financial technology event.

“We won't make the content available directly to the customer before internal validation, before curation and internal validation to avoid ethical issues, hallucinations and biases,” said the executive.

To help it understand how to make the most of AI – generative or not – since 2021 Banco do Brasil has had a consulting partnership with Quebec-based Mila, which is reportedly the world’s largest academic research center for deep learning. 

The bank also follows an internal AI ethics playbook, according to Ferreira.

On the side of open generative AI, which is being tested and used by the public, the three main debates in the market currently revolve around data lineage (basically, where and why the data originates), data privacy and copyright issues, Joia said.

Telma Luchetta, lead partner of data and analytics financial services at EY in South America, urged companies to test and train platforms internally prior to using ChatGPT publicly.

She added that enterprises should carry out a critical assessment of the need to adopt generative AI now.

Accenture and Microsoft recently announced a new partnership focused on generative AI for corporates.

Together with their Avanade joint venture, the companies are developing new solutions to help customers leverage generative AI.