Deutsche Bank builds AI resilience platform on Google Cloud

Deutsche Bank has developed an AI-based resilience platform for regulatory tabletop exercises using Google Cloud tools.

The platform is designed to modernise how the bank prepares and runs operational resilience exercises. It uses operational data from architecture, data flows, logs, incident history, alerting signals and other telemetry to generate scenarios, simulated operational evidence, structured session records and material for regulatory review.

The move comes as financial institutions face tighter scrutiny over operational resilience, particularly under the European Union's Digital Operational Resilience Act, which requires firms to show that critical services and supporting digital systems can withstand disruption, support a coordinated response and recover in a controlled way.

At large banks, tabletop exercises have often relied on manual preparation and static assumptions. Deutsche Bank's platform shifts that process towards a model based on live enterprise context, with scenarios tied to actual system dependencies, failure patterns and business impact.

Google Cloud's Gemini Enterprise Agent Platform turns that operational context into structured scenarios with timelines, decision points and expected responses. This is intended to create a more consistent record for audit and compliance purposes.

Dual model

The platform uses a dual-orchestration design that separates regulated workflows from more adaptive investigative work. For regulator-aligned execution, it uses LangGraph to create scenarios through a traceable process that records how inputs lead to outputs.

For investigative scenarios, the platform uses Google Agent Development Kit to let agents analyse conditions and generate responses without a predefined execution path. This allows the bank to combine tighter control for supervisory review with more flexible analysis when it needs to examine operational issues in greater depth.

The same agents and tools can be applied across scenario generation and incident analysis, allowing teams to work from a shared intelligence layer spanning architecture, data-flow diagrams, logs and code artefacts. The approach supports a common resilience model for both planned exercises and live operational events.

Cloud Run supports execution of scenario and evidence-generation services, while Cloud SQL stores scenarios, session artefacts and review records, according to Deutsche Bank and Google Cloud. Together, those services underpin the generation, execution and retention of records used in resilience workflows.

Each scenario generated by the platform leads to a structured tabletop session for teams responsible for response, escalation and recovery. Because the exercises are based on enterprise operating context, they are intended to reflect actual conditions more closely than conventional simulations.

Regulatory focus

Operational resilience has become a central concern for banks as regulators seek more explicit evidence that institutions can manage outages and cyber incidents across complex technology estates. Large financial groups often depend on interconnected applications, internal data flows and third-party services, making it harder to map the effect of a disruption across the business.

The platform is now being applied across multiple portfolios within Deutsche Bank. The bank described the model as a way to create a more consistent approach to resilience testing at scale and as a template that could be replicated more widely across the sector.

Root-cause analysis from real production incidents also feeds back into future tabletop exercises. In turn, the results of those exercises can be used to refine response playbooks, escalation paths and recovery planning.

That feedback loop extends the platform's role beyond planned exercises into day-to-day operational analysis. It also reduces reliance on fragmented manual preparation by giving teams a single model for resilience work across different parts of the organisation.

One executive at the bank described the initiative as an effort to build a resilience model suited to critical financial systems in distributed environments.

"By linking dynamically generated scenarios to real business context and combining governed orchestration with adaptive analysis, the platform has given us an intelligent, continuously adaptive model for operational resilience," said Sanjay Tripathi, Managing Director, Global Head of Surveillance Technology & Compliance Cloud & AI Transformation Lead, Deutsche Bank.