Mambu launches Intelligent Core for agentic banking AI

Mambu has launched its Intelligent Core architecture, combining its core banking, payments and AI software on a single platform.

The rollout includes a new product, Mambu Agentic, linked with Mambu Core and Mambu Payments. It is intended to give financial institutions a way to connect AI agents directly to banking ledgers and payments systems.

The architecture is designed to let AI agents work with real-time banking data, operate within predefined guardrails and take authorised actions. Mambu said the approach is intended to help banks and other financial firms automate routine operational tasks and speed up decision-making.

The launch introduces three elements within the agentic layer. The first is Model Context Protocol, or MCP, designed to connect enterprise AI agents to core banking and payments systems without extensive integration work.

The second is AI Insights, which Mambu described as a curated data layer intended to ensure the information used by AI systems is trustworthy. It also aims to identify growth opportunities, risks and anomalies across portfolios and operations.

The third is a set of agents that can gather data, assess it against configurable guardrails, act where permitted and explain decisions and actions. Mambu said these workflows are auditable and can be configured around internal policies, approval thresholds and business rules.

Banking stack

Mambu has built its business around what it calls composable banking, an approach that allows financial institutions to assemble separate technology components rather than rely on a single monolithic core platform. According to the company, the new architecture extends that model by embedding AI tools more deeply into the operating layer of banking systems.

Mambu said many banks are still trying to add AI services on top of older systems that were not designed for real-time intelligence or autonomous action. By connecting AI to both the ledger and payments layer, the company is seeking to position its platform as a way to embed automation into day-to-day banking processes rather than limit AI to customer-facing tools or standalone analytics.

Over the past 15 years, Mambu said its platform has been used by financial institutions building or modernising deposit, lending and Islamic banking products. It says it has more than 260 customers in over 65 countries, including Western Union, Commonwealth Bank of Australia, N26, BancoEstado, Raiffeisen Bank, ABN AMRO and Bank Islam.

The move reflects a wider push across financial services technology to turn recent advances in AI into tools that can be used inside regulated operational systems. Banks have been testing AI in areas such as customer service, fraud monitoring and internal support, but connecting those systems directly to transaction ledgers and payment rails raises questions about control, auditability and risk management.

Mambu said its design addresses those concerns by requiring AI agents to operate within strict guardrails and explain their actions and decisions. The system is also configurable, allowing institutions to set their own policies and approval requirements.

Fernando Zandona, chief executive officer of Mambu, outlined the company's position on the shift. "For 15 years, Mambu has freed hundreds of global financial institutions from the constraints of monolithic legacy cores through composable banking," Zandona said.

"That composable foundation now evolves to become intelligent. Intelligent Core is our vision for the future of financial services: core, payments, and agentic AI working as one. Mambu Agentic brings this to life with AI agents that connect directly into the ledger, reason against strict guardrails, act where authorised, and explain every decision. Fully configurable, auditable, and scalable," he said.

He said client conversations are now focused less on whether to adopt AI and more on whether it can be deployed safely and reliably in live operations. "Every financial institution we talk to is not asking if they should adopt AI, but whether it can run safely, reliably, and deliver on its promise at scale. Today with Mambu, it can," Zandona said.