AI in Finance and Energy: How Brazilian Banks, Insurers and Utilities Are Measuring Real Results

Introduction

Three sectors that control Brazil's money and energy reached 2026 with one thing in common: they abandoned the pilot stage and started measuring results. Brazilian banks, insurers, and utilities are no longer asking whether AI works — they are publishing how much it returned. And the numbers are specific enough to change boardroom conversations.

CONTEXT

The context making this relevant now is structural. In financial services, Brazil's Open Finance framework reached operational maturity in 2026, creating data infrastructure that makes AI-driven personalization viable at scale. In insurance, a CNseg survey conducted between October 2025 and January 2026 — covering 26 member insurers and 17 executives — found that 80% of companies are already actively using AI, and 68% expect to have fully automated processes within five years. In the energy sector, the gradual opening of Brazil's free energy market, set to complete by 2028, created urgent demand for data-driven energy efficiency rather than estimates. The practical outcome: three sectors with mature regulation, rich historical data, and real competitive pressure — conditions that make applied AI more effective than in less structured environments.

EXAMPLE 1 — Banco do Brasil: AI that negotiates debt better than humans

The most documented case of AI in Brazilian financial services in 2026 is not the most glamorous — it is debt negotiation via WhatsApp. BB Tecnologia e Serviços, in partnership with AWS and BRQ Digital Solutions, developed a conversational credit recovery model combining generative AI with business rules and direct integration into banking systems. The delinquent customer enters WhatsApp, describes their situation in natural language, and the system simulates personalized conditions, consults information, formalizes agreements, and issues payment slips — without leaving the channel. The bank's published figures: a 306% increase in debt negotiation conversions, 50% of interactions completed without any human intervention, and a reduction in average installment count from 33.17 to 14.22 — meaning shorter agreements and therefore lower risk of recurring default. In parallel, the conversational platform launched in July 2026 — integrated into the app and WhatsApp for individuals — recorded 69% more Pix conversions and a 21% drop in key entry error rates. Transacted financial volume grew 4.2% in the period. For SMEs, the bank has operated since 2024 the ARI, a GenAI recommendation agent offering guidance on cash flow, credit, marketing, and strategy in natural language.

EXAMPLE 2 — MAG Seguros and TYR Energia: different sectors, identical logic

MAG Seguros, a 191-year-old insurer, closed 2025 with R$ 3.5 billion in premiums and net income of R$ 423 million — up 36% year-over-year. Over the past five years, it grew an average of 18.7% per year, nearly double the Brazilian insurance market's average growth of 8–10% annually. The declared strategy: AI embedded across virtually the entire operation. In claims processing, the company processes and pays surgical indemnities on the same day the claim is received. In customer service, sentiment analysis algorithms route clients to the specialist best matched to the emotional context of each call. In distribution, tools convert conversations between brokers and clients into personalized financial protection studies. A collateral result: in 2025, the group's fintech, MAG Finanças, received authorization to operate as a direct Pix participant. In the energy sector, TYR Energia documented an equally precise case: working with Duloren, AI agents mapped operational losses at industrial units in Queimados and Vigário Geral (RJ), identifying charges for excess reactive energy and contracted demand incompatible with the actual consumption profile of the factories. Corrections at both units generated annual savings of R$ 326,000. Combined with migration to the free energy market, total savings exceeded R$ 1.4 million annually — a figure that had never appeared in any of the client company's management reports before.

EVIDENCE

Sector-wide data confirms these are not isolated exceptions. A Topaz/Celent survey of 1,023 financial institution leaders across 20 Latin American countries found that 53.9% of Brazilian institutions already use AI as the foundation for fraud detection — and 40.7% place fraud prevention as their top investment priority in 2026. In the insurance sector, the Evident AI Index 2026 tracked 30 major global insurers and identified a strategic shift: Manulife, Generali, and Intact Financial project generating more than US$ 1 billion in combined AI-driven value by the end of their respective reporting periods. Allianz currently runs 900 simultaneous AI use cases worldwide. In Brazil, the insurance sector projects R$ 2.6 billion in AI investment through 2026, according to CNseg — but 77% of Brazilian insurers still classify their gains as incremental rather than structural. The gap between what the sector invests and what it actually transforms remains real.

IMPLICATION

The pattern emerging across all three sectors is identical: AI delivers measurable results where there is structured data, high-volume repetitive processes, and regulation that requires traceability. Banks have transaction histories. Insurers have claims histories. Utilities have consumption series by meter. In all three cases, the AI model does not need to create intelligence from scratch — it needs to find patterns in data that already exists but has never been processed at the speed required for real-time decision-making. The implication for companies outside these sectors is direct: if historical data exists and the process is repetitive, the logic applies. If data is scarce or the process is case-by-case, ROI will take longer to materialize — and will likely be smaller than promised.

SYNTHESIS+

The next move the data points to is not adoption — it is selection. With AI already present in 80% of Brazilian insurers and conversational platforms in production at the country's largest banks, competition in 2026 and 2027 will center on who can measure results rigorously enough to justify scaling. MAG Seguros publishes growth 2x above market. Banco do Brasil publishes 306% conversion. TYR publishes R$ 1.4 million with the client company's address and tax ID. Companies that cannot publish a specific number are still in pilot stage — regardless of what their press releases say.