Why Pricing is the Perfect Use Case for AI in Banking

Every major technology wave eventually finds its “killer app,” the use case that truly exemplifies the power of the new capabilities. For AI in banking, that use case just might be pricing. But to make that case, I first need to acknowledge something the industry isn’t being
honest enough about: most AI projects at financial institutions aren’t working.

The Dirty Secret of Enterprise AI

The enthusiasm around AI in financial services is genuine and the investment is enormous. But look beyond the press releases and vendor demos, and a different picture emerges. According to McKinsey, while 78% of organizations now use AI in at least one business function, roughly two-thirds remain stuck in what has become known as “pilot purgatory” — running experiments that never graduate to real production deployments. BCG puts the problem in even starker terms, finding that 74% of companies are struggling to achieve any meaningful, scalable value from their AI investments.

What financial institutions need are AI use cases that hit the sweet spot: meaningful enough to justify real organizational commitment, but tractable enough to implement with current systems and current data. Pricing is exactly that use case.

The Problem Has Gotten Much Harder

Not long ago, a bank’s pricing committee could meet monthly, review a tidy packet of rate surveys and margin reports, make a few adjustments, and call it a day. That world is gone.

Markets move fast now. Rate environments shift on headlines. And the competitive landscape has been fundamentally reshaped by a new class of rivals with a very different operating model. According to research from The Financial Brand, fintechs captured roughly 44% of new checking account openings in 2024 — and their momentum accelerated by six percentage points in a single year. These competitors don’t have legacy systems or branch overhead weighing on their cost structure. They have clean data architectures, modern pricing engines, and the ability to make and deploy decisions in hours, not weeks.

According to research from The Harris Poll, commissioned by Q2, 74% of consumers across all generations are now actively asking for more personalized banking experiences, and they’re comfortable with their financial institution using their data to provide them. They’ve been trained by Amazon and Netflix to expect that a company which knows them will offer them something relevant, not a rate card that applies to everyone. When a customer applies for a home equity loan or a certificate of deposit, they’re not comparing your offer to a generic market average anymore. They’re comparing it to an offer that feels like it was made for them.

Banks that cannot deliver that experience are going to lose. The only question is how fast.

Why Pricing Is Uniquely Hard

Before we get to how AI helps, it’s worth understanding why pricing is so difficult in the first place.

Pricing a financial product requires synthesizing an enormous volume of data from systems that don’t naturally talk to each other. A robust pricing decision incorporates instrument-level performance data from the core system, cost of funds from treasury, credit risk parameters from the risk team, competitive rate intelligence from market data providers, customer relationship and profitability data from the CRM, regulatory guidance from compliance, and strategic intent from the line of business and executive leadership. In most institutions, pulling all of that together requires a chain of meetings, manual spreadsheet work, and a dozen email threads just to establish the inputs, before anyone has even begun the analytical work.

Then there’s the decisioning itself. Pricing isn’t a single answer; it’s a matrix of answers. A bank doesn’t set
one CD rate. It sets rates by term, by tier, by channel, by geography, and increasingly by individual customer segment. The number of permutations that need to be analyzed and monitored is staggering, and the interdependencies between them are genuinely complex. Lower one rate, and you shift volume in ways that affect funding costs, which affect the rates you can offer elsewhere.

“That’s a behavioral science problem wrapped in a statistics problem, and the answers live in datasets that are far too large and complex for human analysts to comb through manually with any thoroughness or speed.”

Where AI Changes the Game

This is where the fit between AI and pricing becomes almost obvious, once you see it.

You Don’t Need Another Transformation Project

Here’s where I want to address the concern I hear most often from executives when we start talking about AI-powered pricing: “We’d love to do this, but our data isn’t clean enough” or “We’d need a real-time integration to our core system first.” In twenty-plus years of delivering pricing solutions to financial institutions, we’ve never once worked with an institution that felt its data was fully clean or its systems were fully modernized. If you’re waiting for that moment, you’ll be waiting forever.

The good news is that improving pricing doesn’t require it. Unlike AI use cases that depend on a unified, pristine data environment — full customer 360 views, perfectly harmonized data lakes — pricing can deliver real value with the imperfect, fragmented data infrastructure that actually exists at most institutions today.

Why? Because pricing performance is fundamentally about finding relative patterns, not absolute perfection. A model that can identify that a specific customer segment in a specific product category is systematically underpriced relative to their willingness to pay (even with imperfect data) will generate measurable margin improvement.

On the integration side, the picture is similarly encouraging. Many of the data connections required for an AI-powered pricing workflow already exist or can be established with relatively light technical lift. At Nomis, we’ve spent two decades building and refining these integrations with the systems financial institutions actually run on. The infrastructure required to get started is easier and less expensive than you might expect.

This is what separates pricing from the AI projects that die in pilot purgatory. The path to production is clear, the required infrastructure is largely already in place, and the business case is measurable from the start: customer satisfaction, margin improvement, volume impact, and campaign efficiency are all concrete, quantifiable outcomes that pricing AI can influence in a relatively short timeframe.

The Limits of Automation — And Why They Matter

I also want to be clear about something that the financial services industry needs to discuss more honestly: pricing is not a process that should be fully automated by an all-knowing “banker-bot” agent. It involves too much regulatory exposure, too much P&L impact, and too many dimensions of judgment that genuinely require human expertise. When a pricing decision has fair lending implications, or when it requires a call about strategic direction that affects the institution’s competitive positioning for the next year, that is not a decision to delegate to an algorithm.

“What AI should do is augment human decision makers. It should ensure that when your pricing committee makes a call, they’re making it with better information, faster preparation, and clearer visibility into the downstream implications of their choices. The humans remain in the loop, accountable, and empowered.”

This distinction matters not just philosophically but practically. Institutions that try to automate pricing decisions entirely will eventually face a regulatory examination or a market dislocation that exposes the risks of removing human judgment from the process. Institutions that use AI to make human judgment sharper and faster will outperform on both dimensions that matter: compliance and profitability.

The Downstream Benefits Are Bigger Than They Look

I’d also argue that the business case for AI-powered pricing is underestimated because most institutions are only thinking about the direct margin impact, which is real and significant. But there are second-order effects that are equally compelling.

Pricing is the single most direct lever a bank has for influencing customer behavior. Marketing can create awareness. Branding can create affinity. But pricing is what actually moves the needle on what a customer does. A better offer to the right customer at the right moment drives acquisition, deepens relationships, reduces attrition, and builds the kind of loyalty that survives competitive pressure.

There’s also a significant efficiency argument on the campaign side. When you can identify the specific customer segments that will respond to an offer and price for them precisely, you eliminate the cost of broad campaigns that are accepted by a far wider audience than you intended. Every time a bank runs a promotional rate that attracts customers outside the target cohort, it pays an unnecessary cost. More precise targeting, enabled by better behavioral analytics, makes marketing spending dramatically more efficient while reducing the marginal cost of campaigns that land too broadly.

The Moment Is Now

Financial institutions are under more pricing pressure than at any time in recent memory: compressing margins, rate volatility, a customer base with rising expectations, and a class of fintech competitors that has proven it can win accounts at scale. The tools to respond to all of this exist today.

The institutions that will lead the next decade are the ones that figure out now how to make AI a genuine part of their pricing capability — not as a novelty or a pilot, but as foundational infrastructure for how they make decisions. It doesn’t require a transformation. It requires a decision.

Dallas Wells is the Chief Product Officer at Nomis Solutions, a financial technology company that has delivered pricing optimization and customer segmentation solutions to banks, credit unions, and auto lenders for more than two decades.