The Risk team for the North American holding of a leading global bank, headquartered in New York, oversees market and liquidity risk across corporate and investment banking entities. Key processes still ran on EUCs—End-User Computing tools (e.g., Excel/Access with VBA) built and maintained outside central IT—tied to on-prem SQL. This created regulatory risk (limited controls, lineage, and audit) and business risk (fragility, inconsistent definitions, manual effort). Across the banking industry, there is a clear push to reduce EUC dependence in reporting to meet governance expectations, strengthen resilience, accelerate auditability, and lower manual cost—shifting logic from desktop files to governed data pipelines and semantic models.

This shift also paves the way for scalable AI: governed pipelines and unified, well-defined metrics make it far easier to deploy, monitor, and trust AI/ML models and copilots as they mature. EUCs are often a hidden problem—but the remedy is clear: a modern BI data stack and governed semantic models. At the same time, each EUC captures hours (often years) of expert business knowledge: proven queries, vetted filters, and executive-level visuals that have stood the test of time. These assets were treated as built-in technical requirements and used to multiply their value—turning them into a stronger launchpad that makes data far more powerful for the business and for risk management.

Starting Point (EUC Reality)

Across teams, more than 20 EUCs were identified. Metric definitions were scattered across pivot filters, VBA macros, Access SQL, and source queries. The work ranged from slide-ready visuals and large recurring packs to one-off ad hoc pulls copied into Excel. Without a shared model, teams repeatedly queried the same data in different ways, leading to logic drift and inconsistent numbers. Reporting ran on daily/weekly/monthly cycles with manual email blasts and minimal lineage or audit.

What Was Done — From Inventory to Trusted Models, Delivered in Two Passes

Pass 1: Fast front end on raw tables. To deliver value immediately, the highest-priority EUCs were rebuilt in Power BI directly on the same landing tables users already trusted. Familiar visuals were recreated one-for-one to anchor validation, then enhanced with interactive analysis—drillthrough, conditional tooltips, trend pivots, and quick filters—that revealed far more than static tables. Paginated Reports proved effective for regulated packs and archiving, and email subscriptions kept daily/weekly consumers current. Excel and PowerPoint add-ins were also enabled, allowing teams to refresh live numbers in workbooks and decks without copy-paste.

Pass 2: Model & pipelines. While users tested the front end, the hidden rules buried in pivot filters, VBA, Access queries, and upstream SQL were reverse-engineered. Data was curated in Azure Databricks and logic consolidated into two shared semantic models with conformed dimensions and business-friendly definitions. Row-level security, lineage, and versioned logic were implemented. Once the star schema was ready, reports were switched from raw tables to the model with minimal rewiring—bookmarks and layouts stayed intact—so users continued working without a long “backend wait,” saving weeks of test cycles.

Handling gaps & constraints. Not all data was ready on day one. Missing reference sets (e.g., calendars) were stitched in, compact ingestion jobs added, histories backfilled where needed, and edge cases modeled pragmatically. The result was a sturdier backbone and faster, more reliable refresh. Analytical improvements included patterns that surfaced total breach counts by month/quarter/year and across key dimensions (desk, entity, risk class, product). Teams moved from seeing “today’s breaches” to scanning trends and hotspots, enabling faster root-cause analysis and better prioritization.

Run & adopt. The solution was rounded out with alerts, subscriptions, and Power ON writeback for commentary, plus a clean handover and training so teams could explore confidently.

Target Architecture (High-Level)

Sources were ingested into Databricks via ELT, curated into golden tables, and exposed through a governed Power BI semantic model feeding both interactive reports and paginated packs, including subscriptions and writeback for commentary. After UAT and sign-off, the solution was handed to the Bank’s BI team under production change control—no end-user edits—so the outputs are no longer EUCs.

Results (Business Impact)

The fragile EUC estate was converted into a governed, production-supported analytics platform that raises decision speed, consistency, and assurance across Risk and Finance.

The story began with replication for trust and acceleration for impact. The highest-value EUCs were rebuilt on the same landing tables to prove parity and eliminate noise from the change, while in parallel engineering a conformed star schema in Databricks. Once the model was ready, reports were switched over without disrupting users—keeping layouts and bookmarks intact—and shortening testing cycles materially.

Within a few quarters, EUC sprawl collapsed into conformed semantic models used across desks and entities, giving committees a single language and far fewer reconciliation loops. Recurring packs that previously took two days to assemble now publish in minutes with automatic refresh, and ad hoc questions are answered directly in the tool. Interactive views turned snapshots into signals—exposing breach trends by month, quarter, and year across desk, entity, product, and risk class—so hotspots are flagged earlier and root causes addressed faster.

Executive packs are standardized via paginated reports and subscriptions, while Excel and PowerPoint add-ins keep board materials current without copy-paste. With lineage, RLS, and change control in place, the solution runs in production workspaces with no end-user file edits; audits are smoother and ownership is clear. Because metrics and dimensions are conformed, each new use case lands faster and strengthens the shared model. A formal handover was completed to the Bank’s BI team with runbooks and SLAs, reducing key-person risk. Most importantly, the governed pipelines and unified metrics form an AI-ready foundation, allowing future models and copilots to plug into trusted data with minimal rework.

What Made It Work

Success followed a simple playbook: begin with a clean inventory and frequency-based triage, then run each EUC as a contained mini-project—trace the source, surface the business logic, shape the model, rebuild the report, validate side by side, and retire the file. By designing for reuse with conformed dimensions and shared metrics, each win strengthened the foundation and drove the marginal cost of future EUCs toward zero.

Deliverables

• Live EUC Register with ownership, controls, and SLAs
• Power BI Semantic Model (business-friendly names, KPIs, perspectives)
• Power BI Reports + Paginated Reports (archive-ready)
• Training & handover (export to Excel, bookmarks, subscriptions)
• Writeback for commentary (Power ON)