If you’re a midsize bank considering AI strategy and implementation, welcome to the club.
Banks of all sizes are working to bring AI to their business. Mid-sized banks, which lack the R&D and technical resources of larger brands, face particular challenges because they don’t have the financial or bench depth to withstand a lot of volatility and failure.
Cutting to the chase, midsize banks should approach agent AI by focusing on three priorities: Understand and map 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 mean it’s easy to implement.
There is no debate whether generative AI and emerging agentic AI can deliver real value to financial services. These technologies can improve efficiency through automation, document processing, and analytics. Reduce risk through better fraud detection, credit scoring, and compliance. And you can drive growth through better customer service, personalization, and relationship management tools.
This is one of the reasons why implementing AI in banking is so difficult, and one of the reasons why leaders in various departments across banking are so keen to implement AI. Vendors fascinate them with captivating demos spanning origination, relationship management, underwriting, loan repayment, portfolio monitoring, risk management, compliance, and product development.
But remember: Long before the advent of AI, a drizzle of discrete point solutions often led to a storm of costly problems. Here are the questions you should ask to avoid them:
How should midsize banks manage AI development and deployment?
AI is the most strategically impactful banking technology since the calculator. Successfully navigating an AI strategy requires executive attention and messaging, and leadership must be aligned on the necessary infrastructure upgrades, build-or-buy decisions, and deployment of AI capabilities.
The key is for executives to be careful to prevent specific business units from driving AI strategies. Worse, it prevents multiple business units from embarking on their own initiatives, which can become redundant or locked in with niche vendors, slowing overall AI adoption progress.
In addition to executive-level support and collaboration, more and more companies are implementing organization-wide PMO structures dedicated to strategic AI development. Some of their most important tasks include addressing the following questions:
How are banks currently doing what AI should enhance? Where is the data being captured and stored?
For example, a midsize bank may be turning to AI to enhance its marketing with personalized offers. The first step is to understand at an operational level what you are already doing to attract business. What we hear from banks of all sizes is that there’s a lot they don’t know. If you don’t know how it works, how can you improve it?
Perhaps the most important aspect of strategic AI development for midsize banks is understanding what the bank actually does, process by process, function by function, and bringing together a vision of how those functions interact as a whole for the banking business. This understanding is also fundamental to the digital transformation that AI is driving across this and other industries.
In the marketing example, what systems are involved? Where is the data captured? Where is it stored? Is there a disconnect between your processes and where the data resides?There probably is, as useful marketing tools probably pull from CRM systems, financial systems, and other systems to gather customer profiles and match them to your offerings.
Part of establishing the bank’s process flows and data architecture involves interacting with people. But business transformation management software also plays an important role in really understanding what’s under the hood. There are two main categories here.
- Enterprise architecture management software Determine and visualize what you’re doing to help plan AI-driven upgrades.
- business process management software It provides a way to model and mine processes, show and analyze how business processes work and work together, identify inefficiencies, and model how AI capabilities can best fit together.
Existing processes and systems rely on data, which brings us to the third big question that midsize banks must answer to strategically implement AI.
How can banks make their data AI-ready?
AI requires relevant, reliable, and accountable data to avoid the universal truth that “garbage in, garbage out.” Agent AI puts even more emphasis on data quality. Take AI as an example to help evaluate potential loans. An AI agent might focus on the value of a property and the business plan developed by a potential borrower. The second agent may investigate the applicant’s financial situation and assets. Agents utilize different types of data, and when this data is spread across different systems (as is often the case), it is difficult for agents to run well independently, much less as an agent AI team.
There is a strong interest in enterprise data and analytics platforms to consolidate data with minimal disruption. These data fabric platforms overlay existing systems, unify fragmented data, standardize how businesses interpret that data, and impose a semantic layer that makes that data available to AI agents.
This has interesting implications for HR in terms of data science. Data architecture remains at the core of a data scientist’s job. But their focus on analytics-centric data mining may soon be replaced by data management and data compliance. What data should an AI agent know? How can you make its decision-making process auditable?
What midsize banks are really preparing for is tomorrow’s AI
AI applications are already transforming banking. To prepare for the complexity of agent AI adoption, midsize banks should:
- Establish a centralized AI management and governance structure led by top executives.
- Identify your existing processes to understand where you can best leverage AI.
- Either modernize the data architecture that impacts essentially every application in your bank, or implement an enterprise data and analytics platform to deliver the relevant, reliable, and accountable data that AI requires.
By following these steps, midsize banks can reap the best that AI has to offer today, while positioning themselves to take advantage of tomorrow’s AI tools.
