Traditional automation strategies fail due to the complexity of established rule sets; agentic AI opens the path to true process autonomy for banks – through radical redesign rather than optimization of legacy processes. This strategic topic will also shape the Handelsblatt Banking Summit 2026.
The dead end of rule-based automation
Many banks are stuck in a rut: Historically evolved, human-optimized processes on core banking systems are being made more efficient with an additional layer of RPA or generative AI. This "add-on" approach falls short. Whether it's corporate client onboarding, sanctions screening, or credit decisions, processes run sequentially, in silos between front office, back office, and compliance – prone to errors, expensive, and riddled with media breaks. Strategic superiority doesn't arise from digitizing the status quo, but from redesigning core processes for AI-centric management.
Reducing complexity: the universal process grammar
The core of this approach is the decomposition of each process into standardized building blocks – a uniform "grammar" that is readable for both humans and AI systems. Instead of maintaining hundreds of product and process variations across lending, account, and payment transactions, banks work with a few modular elements that can be flexibly combined. This allows AI agents to independently execute, monitor, and optimize processes.
The „Process Operating System“: model-agnostic instead of vendor lock-in
The rapid pace of AI innovation works in the banks' favor: Today's language models represent the "worst-case scenario"—performance increases monthly while costs decrease. The answer is a model-agnostic architecture as a "Process Operating System": Flexible middleware allows new models to be tested and replaced directly with existing use cases. The central question is no longer which model to choose, but how processes can be continuously optimized based on reliable data.
Governance by Design: the „Safe Envelope“
For banks, the "black box" problem is non-negotiable: MaRisk and the EU AI Act require traceability. An agentic-native architecture integrates governance directly into the execution layer: Hard-coded guardrails embed regulatory safeguards—from the four-eyes principle to approval authority—into the process components and guarantee deterministic results where necessary. The system reliably identifies uncertain cases and directs them to human review; automated monitoring fully documents every decision.
From practice: from the back office to the customer interface
At a major international bank in Germany, manual review steps and fragmented workflows limited throughput and scalability; early GenAI pilot projects failed due to legacy architectures. A redesign based on process grammar—with autonomously interacting GenAI agents, documented as "Process-as-Code" blueprints—delivered a ready-to-implement end-to-end process model, complete with a business case and scaling roadmap. The principle also works at the customer interface: at a Swiss direct bank, a multi-agent system, acting as a voicebot, handles typical service requests end-to-end—with a higher first-call resolution rate and increased customer satisfaction. In both cases, the key to success lies beneath the surface: a process design built for AI from the outset.
Conclusion: The new role of humans as supervisors
Agentic-Native Process Design fundamentally changes the role of humans: away from manual, case-by-case processing, towards a "human-on-the-loop" approach that controls AI agents and manages exceptions. Because the value potential of AI depends significantly on employees' willingness to adopt new ways of working, accompanying coaching remains essential. Banks that think about their core processes in an AI-native way today gain an advantage that grows with each generational leap of the models.
Further reading: Eraneos white paper „Rethinking processes for AI: The building blocks of successful business transformation“ – www.eraneos.com/de/whitepapers/prozesse-neu-denken-fuer-ki
authors
Claudia Schulze is a partner at Eraneos Analytics GmbH. She is responsible for Data & AI across the entire Eraneos Group. Email: claudia.schulze@eraneos.com
Matthias Reck is a Principal at Eraneos Strategy GmbH. His focus is on strategic AI transformation in the financial services sector. Email: matthias.reck@eraneos.com
Sebastian Frey is a Principal at Eraneos Germany GmbH. His expertise lies in large IT transformation projects for complex capital market platforms. Email: sebastian.frey@eraneos.com