Agentic AI refers to intelligent systems designed to perform complex, multi-step tasks autonomously. In the context of open banking and financial services, these systems move beyond simple data sharing. They can proactively manage and optimize personal or business finances. This technology enables automated financial workflows that previously needed human intervention.
The Foundation of Autonomous Finance
Deploying autonomous systems in finance demands a strong and secure framework. Three structural pillars are essential for this: enterprise data management and governance (EDMG), the domain agent model (DAM), and the prompt engineering architecture cards (PACE) framework. This three-tier stack ensures that as financial institutions scale their automation, every action remains grounded in verifiable data and operational trust. It also helps meet strict regulatory requirements.
Enterprise Data Management and Governance (EDMG)
Intelligent systems rely heavily on the quality and integrity of the data they use. Enterprise data management and governance (EDMG) forms the mandatory foundation for any AI deployment in finance. This layer acts as an active filter. It validates data and rejects unauthorized attempts to access or use information. EDMG verifies consent and establishes clear data lineage. This means every piece of information is traceable to its legal origin. This data-first approach provides the auditability needed to manage liability and verify compliance. It ensures these checks happen before the AI acts on any data.
Structuring AI for Financial Tasks: The Domain Agent Model
Using a single, large AI model for all banking operations can create significant auditability risks. The domain agent model (DAM) addresses this challenge by breaking down complex workflows into specialized teams of agents. Each agent is assigned atomic roles, meaning they handle specific, small tasks. For example, in a Know Your Customer (KYC) process, one agent might process documents. Another agent could check sanctions lists. A third agent would then verify overall compliance.
These specialized agents collaborate in a sequenced loop, using open APIs to process necessary data. This allows for autonomous customer onboarding in real time. Isolating banking tasks into these microservices makes the entire operation more secure. It also makes it much easier to audit. This modular approach helps ensure that each step in a financial process is transparent and accountable.
Guiding AI Behavior with PACE
Generic language models are often unconstrained by default. This lack of boundaries is a liability in a regulated financial environment. To build clear boundaries around these financial agents, the prompt engineering architecture cards (PACE) framework is used. A PACE card slots into an agent’s logic. It explicitly defines the agent’s role and the rules it is prohibited from breaking.
For instance, an affordability agent’s PACE card would include hard-coded rules. These rules prevent it from offering a loan that exceeds thresholds defined by banking regulations. PACE serves as a translation layer. It turns subjective regulations into objective, enforceable code for the AI. This ensures that AI agents operate within legal and ethical limits.
Agentic AI in Open Banking and Beyond
Once the internal architecture is governed and secure, the focus can shift to external growth. Agentic AI, combined with open banking, allows for embedded finance. This means projecting banking services directly into non-financial platforms. An example is a loan module appearing inside a ride-share application. This brings financial services to consumers where they are already active.
Scaling these offers requires access to broader open finance data sets. These include information from insurance, pensions, and investment accounts. By embedding services, banks can remain central to transactions. This helps prevent them from being displaced by new fintech competitors.
Different regions have adopted open banking with varying approaches. In the UK, the CMA mandate created a standardized environment. This environment now supports over 12 million active open banking users. Brazil followed a similar mandatory path. It integrated open banking with its Pix payment system. This drove massive transaction volumes and expanded financial access. Singapore, on the other hand, used a market-led approach. It built the SG FinDex platform using a national digital identity system. This facilitated secure data sharing. A resilient architecture must be agile enough to handle both strict mandates and voluntary market standards.
Scaling Agentic AI for the Future of Finance
Successfully integrating the EDMG, DAM, and PACE layers allows a bank to operate as an AI-native engine. Scaling this architecture requires addressing six strategic imperatives. These range from engaging with regulators to strengthening data governance and building API-first infrastructure. These steps create a structured path. They transform raw open data into personalized, automated services for the consumer. The open data economy is accelerating. Banking leaders who succeed will be those who use this architecture to define its future.
