Mambu wires agentic AI directly into core banking ledger
Amsterdam-based cloud banking platform Mambu has launched what it calls Intelligent Core, a technology architecture that connects agentic AI directly to the core banking ledger and payments layer. The move marks a shift from the dominant pattern of the past decade, in which banks bolted AI tooling onto legacy cores as an external layer, and instead attempts to make intelligence a native property of the banking stack itself.
Mambu's platform already serves more than 260 financial institutions across 65 countries, including Western Union, Commonwealth Bank of Australia, N26 and ABN AMRO. The new architecture unifies three existing product lines, Mambu Core, Mambu Payments, and the newly launched Mambu Agentic, into a single composable platform.
What Mambu Agentic actually does
The Intelligent Core rests on three new capability layers. The first is a Model Context Protocol (MCP), which gives enterprise AI agents a direct data channel into core banking and payments systems without requiring bespoke integration work. The second is AI Insights, a curated data layer designed to ensure the information an agent reasons against is clean and auditable, while surfacing portfolio-level anomalies and growth signals automatically. The third is a pre-built Agents library: configurable action agents that can gather data, apply business-logic guardrails, execute authorised transactions, and generate a full audit trail of every decision.
"Intelligent Core is our vision for the future of financial services: core, payments, and agentic AI working as one," said Fernando Zandona, Mambu's Chief Executive. "AI agents connect directly into the ledger, reason against strict guardrails, act where authorised, and explain every decision. Fully configurable, auditable, and scalable."
The emphasis on auditability and guardrails is deliberate. Regulatory scrutiny of autonomous systems in financial services is intensifying across the EU, the UK and Singapore, and any bank deploying agents that can initiate transactions needs a defensible audit log. Mambu's design, agents that explain every action, is positioned as a compliance-first response to that environment.
The convergence read-across: fintech infrastructure meets the agentic era
The broader significance of this launch sits at the intersection of two accelerating forces. First, the agentic AI wave: enterprise software vendors from Salesforce to ServiceNow are racing to embed autonomous agents into their platforms, and the financial-services stack is a high-value battleground precisely because the decision-making surface is enormous. Loan approvals, fraud triage, KYC workflows and treasury operations are all high-volume, rules-governed processes that agents can, in theory, handle faster and more consistently than human queues.
Second, and less discussed, is the infrastructure dependency this creates. When AI agents are wired directly into a live ledger, the reliability and latency profile of that ledger becomes a bottleneck for AI performance. Banks that modernised onto cloud-native cores in the 2015–2023 wave, Mambu's own customer base being a prime example, are structurally better positioned to deploy agentic workloads than incumbents still running batch-processing mainframes. That gap is now a competitive moat, and it is measurable in terms of time-to-decision for credit and payments products.
For capital allocators watching the banking-technology landscape, the implication is that the composable-core vendors (Mambu, Thought Machine, Temenos on its cloud rebuild) are no longer competing only on core-replacement deals. They are now competing for the agent-infrastructure layer of the financial system, a market that sits upstream of both the AI-tooling vendors and the system integrators who deploy them. That repositioning, if it holds, expands the addressable opportunity considerably and changes the competitive set to include hyperscaler-native financial-services plays from Microsoft and Google Cloud.
Whether Mambu's specific implementation delivers on its promise at the scale of a tier-one bank remains an open question. The customer list leans toward digital challengers and mid-tier institutions rather than the largest global incumbents, and the complexity of deploying autonomous agents safely across a full retail banking operation is not trivial. The next proof points will be case studies from existing customers running Mambu Agentic in production.