Originally published March 2022 | Updated March 2026
Becoming a data-first bank or credit union has always been important. Now, it’s existential.
The institutions that thrived through the last few years of margin compression, digital disruption, and rising member expectations weren’t necessarily the biggest – they were the ones that could act on their data faster than the competition. “Data-First” still means what it always has: treating data as a raw material with inherent value, not a byproduct of the business cycle. But the stakes and the tools have changed considerably since we first published this piece.
The urgency is new. Generative AI, agentic workflows, and real-time personalization are no longer future-state aspirations – your competitors are piloting them now. And every one of those capabilities runs on a data foundation. If yours isn’t solid, you can’t participate. That’s the updated case for this journey.
Here’s where to start, and what to measure along the way.
Milestone 1: Establish Data Governance
Nothing downstream works without this. Data governance in 2025 means the same three pillars it always has – quality, collection, and comprehension – but the urgency around each has compounded, because AI models are only as reliable as the data you feed them.
Ensuring Data Quality
Garbage in, garbage out has never been more consequential. When an AI model surfaces a next-best-action recommendation or flags a member at risk of attrition, that output is only trustworthy if the underlying data is. A data stewardship team – people embedded across business lines who own the accuracy of data in their domain – remains the foundation. Their job: eliminate duplicates, reconcile inconsistencies, and keep records compliant and current.
Visualization tools like Tableau, Microsoft Power BI, or Salesforce Data Cloud’s native dashboards make this manageable at scale, surfacing exceptions before they propagate downstream.
Collecting and Storing Your Data
A data warehouse (or modern data lakehouse) is still your single source of truth. Start with the use cases that matter most – typically a customer or member data mart – and build from there. What’s changed is the expectation: a unified data model that connects your core, your CRM, your lending platform, and your digital channels is now a prerequisite for AI, not a nice-to-have.
Salesforce Financial Services Cloud paired with a real-time integration layer (MuleSoft is the architecture we see working at scale) is one proven path. We’ve seen institutions like Ponce Bank connect FSC, Fiserv Cleartouch, and nCino through API-led integration to achieve a true 360-degree customer view – and institutions like Customers Bank use that same foundation to digitize 140+ processes and post 50% asset growth. See how that architecture works →
Understanding Your Data
Your data dictionary and data catalog are table stakes. What’s evolved is the analytics maturity model. In 2022, most banks were still working on descriptive and diagnostic analytics. In 2025, the bar has moved: institutions that haven’t moved toward predictive and prescriptive analytics are already behind, because that’s the layer where AI and machine learning do their most valuable work.
Get the documentation right, master your descriptive baseline, and then build toward the analytics layers that actually drive decisions.
Milestone 2: Build Your AI-Ready Data Model
This milestone didn’t exist in our original post. It does now.
Generative AI and agentic AI have introduced a new technical requirement: your data model has to be structured in a way that AI can actually use. That’s not a given with most legacy core systems or homegrown CRM setups. Disconnected fields, inconsistent schemas, and siloed household data create the kind of ambiguity that breaks AI outputs – and erodes trust in the technology fast.
Salesforce has introduced a new FSC Data Model specifically designed for AI readiness: unified household structures, cleaner relationship graphs, and the connective tissue that lets Data Cloud and Einstein AI surface insights that are actually actionable. If your institution is on FSC – or evaluating it – understanding this model isn’t optional. We wrote the whitepaper on it. →
The practical question for any Data-First institution right now: if you turned on an AI feature tomorrow, could your data support it? If the answer is “maybe” or “I don’t know,” this milestone is where to focus.
Milestone 3: Apply AI and Machine Learning
Once your foundation is solid and your data model is AI-ready, you can unlock the capabilities that make Data-First institutions measurably different.
Traditional ML use cases – churn prediction, default risk scoring, next-best product – are well-established and worth pursuing if you haven’t already. But the newer surface area is generative and agentic AI: tools that don’t just flag a risk or recommend an action, but can draft the follow-up communication, route the case, and log the outcome automatically.
A few examples worth paying attention to:
- Intelligent Document Processing (IDP): Automating the extraction and classification of loan documents, statements, and forms. The ROI here is tangible and fast – we’ve calculated it out to roughly $1,500 saved per loan when IDP is implemented correctly. See the math →
- Agentic servicing workflows: AI that can handle tier-1 member inquiries end-to-end without human handoff – freeing your team for relationship-intensive work.
- Predictive attrition and cross-sell: Using behavioral signals across your unified data model to surface the right offer to the right member at the right moment, automatically.
None of these work without Milestones 1 and 2 in place. That sequencing is intentional.
Milestone 4: Measure ROI to Your Organization
Data strategy without ROI accountability is a cost center. With it, it’s a competitive advantage – and one you can defend to a board.
The five pillars we outlined in 2022 still hold. Here they are, updated for today’s environment:
- Innovation: Has your data foundation enabled you to launch new products, expand into new markets, or pilot AI capabilities ahead of peers? This is increasingly a valuation driver for acquirers and investors.
- Revenue: Are you using unified data to cross-sell and upsell more intelligently? Digital account opening – the build vs. buy calculus matters here – is one area where data-driven personalization is closing the conversion gap.
- Operational efficiency: How many hours are your teams spending on manual data reconciliation, document handling, or swivel-chairing between systems? IDP, workflow automation, and AI-assisted servicing all show up directly here.
- Risk avoidance: Are your compliance and data quality controls proactive or reactive? A strong data foundation reduces regulatory exposure and audit burden.
- Direct costs vs. value created: What does it cost to run your data strategy, and what are you getting back? Model it out quarterly. The institutions that make Data-First a board-level priority tend to be the ones that can show a live ROI figure – not just a project status.
A Journey, Not a Sprint – But the Map Has Changed
Becoming a Data-First institution still isn’t built in a day. But the urgency has accelerated, and the milestones have multiplied. AI readiness is no longer a future milestone – it’s the thing that determines whether the investments you’re making in data infrastructure pay off at the pace you need them to.
If you’re not sure where your institution stands on this journey – or whether your current CRM and data environment can support where you’re trying to go – let’s talk. Our Financial Services Cloud LaunchPad is designed specifically for institutions that are ready to move fast without starting from scratch.
The question “where does our customer data actually live?” is a good one. The better question is: what can you do with it once you know?
Related reading:
