
Sergio Barbosa, CEO of FutureBank.
Sergio Barbosa, CEO of fintech firm FutureBank, says traditional, rule-based and rigid automated technology is giving way to agentic AI to enable intelligent, adaptive and customer-centric banking.
Agentic AI represents the future of financial services, adds Barbosa. “Traditional automation relies on fixed rules, which limits adaptability and intelligence. Agentic AI, by contrast, can dynamically determine the best next action based on context, available tools and goals, making workflows smarter and more autonomous.”
Barbosa says the shift from static automation to agentic intelligence leads to operational efficiency. “By automating complex, multi-step processes such as case management, reconciliation processes, deal-making and liquidity management, agentic AI reduces the need for human intervention, speeds up workflows and minimises errors.”
“This will allow banks to run using smaller, leaner teams, giving large banks the same kind of agility currently enjoyed by their smaller, more nimble counterparts. Barbosa says these efficiencies will translate directly into improved cost-to-income ratios, allowing banks to do more with less and remain competitive in an increasingly contested environment.”
AI agent customer representative
Barbosa sees a future where customers are empowered by their own AI agents, or digital representatives, that aggregate financial data, optimise spending and even negotiate better deals on their behalf.
“This is going to flip things on their head. Customers can have access to their personal AI that can get an aggregate view across their accounts to allow personal financial management, within the safety of the bank’s system,” he explains.
FutureBank’s view is that this shift from institution-centric to customer-centric AI marks a profound change in the relationship between banks and their clients, enabling hyper-personalised services and proactive, predictive financial management.
The company points to agentic AI’s ability to leverage both private and public data, combined with advanced techniques like using vector databases, retrieval/cache augmented generation and model context protocol servers, allowing for context-aware, forward-looking insights.
Customers can move from reactive problem-solving (responding to issues as they arise), to proactive, strategic decision-making.
“Agentic AI continuously monitors customer accounts, transactions and market conditions. It can proactively alert customers to potential risks and opportunities before issues arise. This not only enhances the customer experience but also helps banks build deeper, more trusted relationships with their clients,” Barbosa says.
Barbosa sees a significant opportunity for banks at the intersection of agentic AI and crypto.
“Currency programmability opens up new possibilities for automating interbank relationships, deal-making, liquidity management and treasury functions, which are all areas that are notoriously difficult to manage manually. With agentic AI, banks can leverage these programmable assets to streamline operations, enforce regulations more easily and create new, efficient business models,” he shares.
However, Barbosa warns that the path to agentic AI is not without its challenges. Banks are traditionally accustomed to closed, highly regulated environments, and the shift to open protocols and transparent systems can be daunting.
Barbosa explains that legacy technology presents a significant hurdle, as integrating new AI capabilities with old systems often requires building interfaces and orchestration layers to connect the two worlds. Security and fraud risks are also evolving, with open systems and programmable money introducing new attack vectors that demand advanced monitoring and prevention strategies.
While bullish about agentic AI, research firm Forrester paints a realistic picture of the challenges for early adopters. “We are still in the early stages of agentic AI’s market impact; companies must test, learn and iterate because these powerful systems can be misaligned, creating actions that are at best undesirable and at worst harmful to your customers and critical applications.”