Session

Sponsored by: Deloitte | Future-Proofing Finance: Modernizing Banking Data Platforms

Overview

Experience	In Person
Track	Data Sharing & Collaboration
Industry	Consulting & Services, Financial Services
Technologies	AI/BI, Unity Catalog
Skill Level	Intermediate

As financial institutions race to modernize their data infrastructure amid rising expectations for real-time insights, data lineage, and AI-ready systems, Navy Federal Credit Union (NFCU) sought to unify multiple data sources and strengthen governance to improve scalability, security, and regulatory confidence. Deloitte collaborated with NFCU to design and deliver an integrated data platform leveraging Databricks Lakehouse, establishing standardized ingestion and transformation patterns, lineage-enabled governance with Unity Catalog, and secure access controls to support enterprise reporting and advanced analytics. The team has enhanced data engineering delivery using Databricks Assistant and is reimagining Finance processes embedded with multi-agent solutions powered by Agent Bricks. Together, we enabled a scalable foundation that accelerates insight delivery while improving transparency, auditability, and confidence in data used for future Finance and AI use cases.

Session Speakers

Jonathan Weil

/Finance AI & Data Lead
Deloitte Consulting LLP

IMAGE COMING SOON

Tom Hughes

/SVP, Foundational Data & AI Platforms
Navy Federal Credit Union

Full Summary

Future-proofing finance: how Navy Federal Credit Union rebuilt its data foundation for AI

Many finance teams are data-rich yet insight-poor, hindered by fragmented systems, manual handoffs, and inconsistent definitions that erode trust in reporting. The conversation outlines how Navy Federal Credit Union modernized finance on a governed lakehouse foundation, moving from report chasing to a data product-centric model built for analytics and AI. The result is faster delivery, stronger controls, and pathways to generative and agentic AI that meet the bar for audit, compliance, and ICFR.

A clear playbook emerges: start with a finance data product blueprint, embed governance by design, use accelerators to speed delivery, and prove value with end-to-end use cases. Anchoring on a modern Databricks Platform created a common, trusted layer that now powers both analytics and AI across the office of the CFO. Organizations pursuing similar transformations can explore how leading companies are delivering trusted AI-powered self-service analytics to benchmark their own programs against proven patterns.

Reframing finance around interoperable data products

Finance's legacy sprawl included more than 80 applications, point-to-point interfaces, duplicated data, and siloed analytics. The team pivoted from building one-off reports to designing shared, interoperable data products that span controllership, FP&A, and tax. Business stakeholders took on product ownership and accountability for definitions, a cultural shift as important as the technical redesign.

Governance was built in from day one. Metadata and lineage were captured in Unity Catalog, and both technical and business data quality rules were codified alongside development. Industry-standard data models served as accelerators, giving finance a common vocabulary and shortening the path from design to delivery.

AI-accelerated build, enterprise-grade outcomes

Agents scanned metadata and profiled source systems to generate context-aware definitions and data quality thresholds, replacing slow manual hunts through legacy documentation. Databricks Assistant helped refactor Databricks Notebooks, speed defect fixes, and enable more frequent, reliable releases, so finance saw new data land in production on a predictable cadence.

A loan securitization MVP proved the model. In 12 weeks, data from six systems and 150 attributes flowed through the architecture into a working data product and dashboards. Given securitization supports 50 to 100 million dollars in monthly funding, a 2 to 5 percent improvement translated directly into new lending capacity. The tangible impact, visible to the CFO, became the benchmark for subsequent finance data products.

Earning trust for finance-grade AI

Finance adoption of AI demands trust, validity, and context. Trust came from transparent, industry-aligned definitions down to the attribute level, owned by the finance organization. Validity was preserved by maintaining explainability, lineage, and audit artifacts, even as probabilistic techniques accelerated building and discovery, so outputs remained traceable to deterministic, documentable logic.

Context was delivered through a semantic layer aligned to how the business operates. When AI surfaces a metric like gain on sale, it reflects the organization's own definition, not a generic one. Unity Catalog underpins the governed "digital storefront," enabling self-service discovery and conversational analytics, including early exploration of generative BI with Genie, without sacrificing control.

From digital storefront to agentic finance workflows

The roadmap expands governed self-service and turns manual interfaces into intelligent workflows. Agentic AI, including Agent Bricks, is being applied to replace point-to-point data movements and to redesign processes such as loan document audits. Operational data capabilities are growing alongside analytical ones, aligning finance's transformation with core banking reengineering and tightening the loop between operational events and finance outcomes.

Strong scaffolding before smart tools

Generative and agentic AI only deliver durable value when built on well-modeled, governed, business-owned data products. Investment in blueprint, governance, and semantics turns impressive demos into dependable, finance-grade operations.

FAQ

Upstream platforms that feed finance were reverse engineered first. Prioritizing the anchor systems built trust in data sources before constructing the finance data product layer on top of them.

A data product is a governed, reusable asset with shared definitions, owned by the business and designed to interoperate across functions like controllership, FP&A, and tax. Reports consume data for a single purpose, while data products serve many use cases and persist as managed assets.

Explainability, lineage, and audit artifacts were preserved so outputs remain traceable to deterministic logic suitable for controls and ICFR. AI accelerates discovery and build steps, while the end-to-end process meets audit and documentation standards.

Unity Catalog captures lineage, business rules, and governance metadata, then feeds the enterprise data catalog that acts as a digital storefront for finance. That governed layer enables safe rollout of self-service, generative BI, and agentic AI.

Over the next 6 to 12 months, the focus is on expanding governed self-service via the digital storefront, applying Agent Bricks to eliminate manual point-to-point interfaces, and redesigning workflows like loan document audits as agentic processes, in parallel with broader operational platform modernization.