Generative AI (GenAI) is no longer a novelty or optional experiment for financial institutions. From global banks to asset managers and fintech upstarts, it is rapidly becoming the cornerstone of next-gen transformation. But unlocking real value—and avoiding headline-making risks—requires more than plugging into a model. It demands a new data strategy, operating model, and mindset.
We’re Past the Pilot Phase
According to McKinsey’s review of 16 major financial institutions, over 90% have now established some form of centralized GenAI function. Those with highly centralized models are leading the pack—70% have put GenAI use cases into production.
McKinsey identifies four GenAI operating models being used by banks:
- Highly Centralized – Central teams manage strategy, data, and execution. (~20% adoption, ~70% in production)
- Centrally Led, Business Unit (BU) Executed – Strategy set centrally; BUs implement. (~30% adoption, ~50% in production)
- BU Led, Centrally Supported – BUs lead with central guidance. (~30% adoption, ~50% in production)
- Highly Decentralized – Each BU acts independently. (~20% adoption, ~30% in production)
The highest performing institutions have adopted centralized models that enable faster scaling, governance, and data access.
Meanwhile, GenAI is also a disruptor of the asset management industry. Investment strategies are being reimagined, client reporting is real-time and contextual, and entirely new asset classes are emerging.
Those who hesitate risk falling into irrelevance.
The Data Activation Gap (a.k.a. The Data Infinity Loop)
Yet many financial institutions are struggling to move beyond isolated pilots, and a significant bottleneck is data activation. Success with GenAI isn’t about model selection—it’s about creating a continuous data-to-insight-to-action cycle:
- Data Engineering and Integration — Unifying real-time data from distributed systems
- Artificial Intelligence/Machine Learning (AI/ML) Operations — Feeding relevant, timely, trusted data to GenAI
- Workflow Embedding — Using insights to support decisions
- Feedback Loops — Closing the loop with real-world results and refinement
Those who can’t activate data across this cycle will struggle to scale.
What the Leaders Are Doing Differently
- Centralizing to Scale
Highly centralized models succeed because they concentrate talent, unify architectures, and reduce friction.
- Rewiring Workflows
GenAI is being embedded into customer service, risk, compliance, and investment workflows—in addition to being deployed as a chatbot.
- Treating Data as a Product
Winners focus on AI-ready data products like Bronze (raw), Silver (curated), and Gold (decision-grade) datasets tailored to specific GenAI use cases.
- Prioritizing Trust and Governance
Top firms are building explainability, auditability, and ethical use into the foundation of GenAI programs.
The Last Mile: Why Denodo Matters
From “Try Everything” to “Prove It Works”
The GenAI wave is maturing. Success now requires:
- Real-time, trusted data inputs
- Cross-functional orchestration
- Insight-to-action delivery
- KPI tracking and performance tuning
Firms that master the data-insight-action loop will outpace their peers.
But getting there means overcoming the growing data tide. Financial institutions are facing a deluge of structured and unstructured data from transactions, logs, apps, cloud services, regulators, and third parties. Trying to centralize all of it into a single source of truth is increasingly impractical.
Instead, leading firms are leveraging logical data management platforms to provide a real-time virtual view across distributed sources. This lets them activate AI-ready data without physically moving or replicating it—dramatically accelerating time-to-insight.
This is also the critical last mile in which most GenAI initiatives falter: At this point, data must be context-rich, compliant, and fresh enough to feed the AI engine at just the right moment – in other words, data must be in the form of Gold Standard Data Products. When it is not, hallucinations rise, outcomes degrade, and trust is lost.
The Denodo Platform, and other logical data management solutions, are now essential for traversing that last mile.
The Rise of Agentic AI in Financial Services: Promise Meets Caution
As financial institutions accelerate their GenAI deployments, a new development is unfolding: the rise of Agentic AI—autonomous or semi-autonomous agents capable of executing tasks, making decisions, and interacting across systems with minimal human intervention.
What’s Promising
Agentic AI holds transformative potential:
- Portfolio rebalancing bots that monitor markets and adjust positions in real time
- Customer service agents that resolve multi-step issues across channels
- Fraud detection agents that monitor anomalies, trigger alerts, and execute interventions
- Regulatory surveillance bots that flag and explain compliance issues on-the-fly
These agents are being trained to reason, learn from outcomes, and interact with systems in increasingly human-like ways. Unlike earlier robotic process automation (RPA) or rule-based engines, they can contextualize, adapt, and prioritize actions dynamically.
A Cautionary Tale
But with this autonomy comes risk, and agentic AI surfaces a new class of challenges:
- Misalignment with business intent – Agents may interpret objectives in unintended ways, triggering incorrect actions or financial exposure.
- Model drift and overconfidence – Agents acting on outdated, biased, or partial data can make poor or risky decisions at scale.
- Hidden dependencies and chain reactions – Agents triggering other agents, in loops or across departments, could magnify errors.
- Auditability and accountability gaps – Who is responsible when an autonomous agent acts outside approved bounds?
A striking example: A trading desk prototype enabled a GenAI agent to monitor sentiment, adjust positions, and suggest hedges. But without adequate throttling, the agent executed a strategy that nearly exposed the firm to regulatory reporting violations—because it wasn’t constrained by compliance boundaries embedded in other systems.
What Financial Institutions Must Do:
- Limit autonomy to well-scoped, high-trust use cases
- Embed compliance, ethics, and explainability-by-design
- Continuously monitor and retrain agents based on outcomes
- Enable agent output to be traceable and overrideable
Agentic AI represents the bleeding edge of GenAI in finance. When paired with real-time, governed, and explainable data—Gold Standard Data Products—it could be game-changing.
But as with all powerful tools in finance: Trust must be earned, not assumed.
What’s Next: The AI-Native Financial Institution
The future belongs to AI-native financial institutions that can continuously learn, personalize, and adapt in real time.
They won’t rely on brittle pipelines or replicate data endlessly. Instead, they’ll leverage logical data management platforms to virtualize access, build composable data products, and deliver GenAI at scale with trust and governance.
This is not a technology upgrade. It’s an operating model revolution.
Call to Action: Make Your Data GenAI-Ready
The difference between pilots and production at scale comes down to one thing: the ability to activate trusted, real-time, governed data when and where it’s needed.
At Denodo, we help financial institutions to:
- Virtualize access to siloed data—without replication or latency
- Build and manage AI-ready data products
- Deliver insights to GenAI systems with trust, context, and control
Talk to us about how we can help you to make your GenAI ambitions a reality.
