If you’re a midsize bank trying to figure out AI strategy and implementation, welcome to the club.
Banks of all sizes are wrestling with bringing AI into their businesses. Midsize banks lacking the R&D and technical resources of the biggest brands face a particular challenge, because they don’t have the financial or bench depth to afford many swings and misses.
To cut to the chase, midsize banks should approach agentic AI by focusing on three priorities: establishing centralized AI governance; understanding and mapping internal processes and data flows; and establishing a unified data architecture. Without these foundations, AI implementations risk fragmentation, low adoption, and regulatory exposure.
Banking is made for AI. That doesn’t make implementing it any easier.
It’s beyond debate that generative AI and emerging agentic AI can deliver real value in financial services. These technologies can boost efficiency through automation, document processing, and analytics; they can reduce risk through better fraud detection, credit scoring, and compliance; and they can stoke growth through better customer service, personalization, and relationship management tools.
That’s part of why AI is so hard in banking, and why leaders of various functions across the bank’s operations are eager to get their hands on it. Vendors are serenading them with seductive demos across origination, relationship management, underwriting, loan servicing, portfolio monitoring, risk management, compliance, and product development.
Remember, though: Long before AI, a drizzle of discrete point solutions often brewed into storms of costly problems. Here are the questions you should be asking to avoid them.
How should a midsize bank manage AI development and rollout?
AI is the most strategically impactful banking technology since the calculator. Getting AI strategy right takes C-level attention and messaging, and top management must be aligned on the necessary infrastructural upgrades, build-or-buy decisions, and rollout of AI capabilities.
Crucially, C-level attention prevents a particular line of business driving AI strategy – or, worse, multiple lines of business embarking on their own initiatives that may be redundant or invite lock-in with niche vendors, slowing overall AI-implementation progress.
In addition to C-level executive sponsorship and alignment, we’re seeing companies implement a dedicated, organization-wide PMO organization dedicated to strategic AI development. Some of their most vital work involves addressing the following question.
How is the bank now doing what AI is supposed to augment, and where is the data captured and stored?
For example, let’s say a midsize bank is looking to AI to sharpen its marketing with personalized offers. The first step is to understand at an operational level what it’s doing to attract business already. What we’re hearing from banks of all sizes is that they often don’t know. How do you improve something if you don’t know how it works?
Perhaps the most important facet of a midsize bank’s strategic AI development is understanding what the house is actually doing, process by process, function by function, and then pulling together a vision of how those functions interact as the whole of a banking business. This understanding also happens to be fundamental to the digital transformation that AI is catalyzing across this and other industries.
With our marketing example, what systems are involved? Where are you capturing the data? Where are you storing it? Is there a disconnect between the process and where the data resides? There may well be, because a useful marketing tool probably pulls from a CRM system, a financial system, and other systems to assemble a customer profile and match it with offerings.
Part of establishing a bank’s process flows and data architecture involves talking to people. But business transformation management software also plays an important role in really getting a sense of what’s under the hood. There are two broad categories here.
- Enterprise architecture management software determines and visualizes what you’re running and helps plan an AI-driven upgrade.
- Business process management software does process modeling and mining, showing and analyzing how your business processes work and fit together, helping spot inefficiencies, and providing a way to model how an AI capability might best fit in.
Your existing processes and systems hinge on data, which brings us to the third big question a midsize bank must answer to deploy AI strategically.
How do we as a bank get our data ready for AI?
AI needs relevant, reliable, responsible data to avoid the universal truth of “garbage-in, garbage-out.” Agentic AI puts even more stress on data quality. Take the example of AI to help assess a potential loan. One AI agent might focus on the property’s value and the business plan the potential borrower has laid out. A second agent might look at the applicant’s financials and assets. The agents are tapping into different types of data, and if those data are scattered across different systems, which is often the case, it’s hard to get the agents to run well independently, much less as an agentic-AI team.
To bring data together with minimal disruption, we’re seeing keen interest in enterprise data and analytics platforms. These data fabric platforms overlay existing systems, integrating fragmented data and imposing a semantic layer to standardize how the business interprets that data and lets AI Agents loose on it.
This has interesting personnel implications on the data-science front. Data architecture remains a mainstay of the data scientist’s work. But the analytics-focused data mining they’ve also focused on may soon be supplanted by data stewardship and data compliance: What data should an AI agent be privy to? How can we make sure that AI agents’ decision-making processes are auditable?
Tomorrow’s AI is what a midsize bank is really preparing for
AI applications are already making a difference in banking operations. To prepare for the complexities of agentic AI’s proliferation, midsize banks should:
- Establish centralized AI management and governance regimes led by top management.
- Nail down existing processes to understand the best leverage points for AI.
- Either revamp the data architecture, which impacts basically all a bank’s applications, or implement an enterprise data and analytics platform to provide for relevant, reliable, responsible data AI needs.
Following those steps can help midsize banks reap the best what AI can deliver today while positioning themselves to exploit the AI tools of tomorrow.