How AI is changing personalization in financial services

Personalization in financial services has been on the agenda for years. What changed is user expectations and the technology’s capacity to respond in real time.

Platforms like Uber, Netflix, or Mercado Libre don’t compete with banks. But they do set the standard for digital experience that users expect from any product. Every personalized interaction on those platforms raises the bar for everything else, including their bank.

That puts real pressure on product and marketing teams: generic campaigns don’t move metrics anymore. And static segmentation, updated weekly or monthly, arrives too late.

Experience beats price

A study of 5,802 borrowers ranks the factors that drive satisfaction with personal loans. The top three: that the loan met the borrower’s needs, trust in the lender, and the experience of obtaining the loan. Ease of doing business ranks fourth. Interest rate doesn’t appear as its own factor: it’s embedded in whether the loan met needs. The takeaway is clear: how the process feels matters as much as what it costs.

According to Braze’s 2025 Financial Services Customer Engagement Review, 66% of financial services leaders say they understand their customers’ preferences well. Yet only 41% personalize messages based on real-time behavior, the lowest figure across all industries surveyed.

The real problem is fragmentation

Financial institutions have transaction history, demographic profiles, and account data from day one. But that data lives in different systems, channels are managed separately, and teams operate on different decision logic.

As a result, coordinating messages in real time becomes complex. The information exists, but the infrastructure wasn’t designed to act on it the moment a user makes a decision.

In Latin America, this is amplified. 40% of financial services companies plan to use WhatsApp as a marketing channel, but only 20% manage multiple channels from a single coordinated interface. What’s the result? Duplicate messages, out-of-sync timing, and lost context.

Meanwhile, 28% of financial services leaders say IT or product teams own customer engagement, the highest of any sector. Every new campaign depends on a technical ticket. Execution speed drops.

Open Finance is starting to unlock some of these data silos across the region. But having more data only amplifies the coordination problem if you can’t act on it in real time. We wrote about this transition and its relationship to growth here.

How real-time personalization works

Real-time personalization is built on events: concrete actions a user takes inside the product. Each event (opening the app, simulating a loan, abandoning a flow) becomes a signal that can trigger a decision.

Traditional segmentation groups users by similar traits and sends them the same message. But two users in the same segment can behave in completely different ways.

Hyperpersonalization works with each person’s actual behavior instead. A user who primarily uses the app for travel gets communications about international benefits. One with recurring savings patterns gets investment suggestions based on what they already do, not their age bracket.

An important nuance here: the best implementations don’t make users feel like the system is deciding for them. According to Fiserv research, consumers are comfortable receiving AI-powered recommendations, but they want to make the final call themselves. In other words, the most effective hyperpersonalization doesn’t replace user agency: it amplifies it.

A concrete example

A user opens their bank app on a Tuesday at 10pm. They simulate a USD 5,000 personal loan and drop off at the documentation step. In a static segmentation model, that user gets nothing, or a generic loan email three days later.

With an event-based model, the system logs the simulation, the amount, the drop-off point, and the time. Within 24 hours, it can trigger a push notification with a specific message:

“Your USD 5,000 loan is ready. Just one document left.”

The channel, message, and timing are defined by what the user did, not which segment they belong to.

The same principle applies to fraud detection. When the system has a baseline of each user’s typical behavior, any anomalous transaction gets flagged faster, against that person’s individual pattern rather than a segment average.

Where AI comes in

Back to the example. The user abandoned the loan simulation at 10pm. The team needs to decide: what message to send? Through which channel? When? With what offer?

If they also want to vary tone, frequency, and format, the possible combinations run into the thousands. In practice, most teams solve this with manual rules: if segment A, then message B, at time C. Each new combination means a new campaign, a new test, and weeks waiting for results. That approach doesn’t scale.

This is where AI-based decision systems come in.

Amplitude AI answers a question many product teams ask but few can answer with data: what’s happening inside the product, and why do some users succeed while others don’t? Which actions during the first 7 days predict a user will still be active at month 3. Which abandoned flows correlate with churn. Which features the highest-value users engage with. Amplitude surfaces those patterns automatically, before anyone on the team has formed the hypothesis.

Braze AI Decisioning answers a different but complementary question: for each individual user, what’s the best combination of message, channel, timing, frequency, and offer to achieve the goal we’ve defined? It uses reinforcement learning to continuously experiment and optimize communication decisions at a 1:1 level. Every interaction feeds the model, which adjusts on its own. The team defines the outcome they want (retention, activation, conversion) and the system finds the path.

The measured impact

Each tool is powerful on its own. Combined, they create a closed loop: product intelligence (Amplitude) informs the communication strategy (Braze), and in-app behavior determines the experience users receive outside the app. As a result, teams move from defining every rule by hand to defining objectives (retention, activation, lifetime value) and letting the system optimize continuously.

The impact is measurable. According to Braze data, financial services companies that go from zero channels to one messaging channel see a 2.9x increase in 90-day retention. Those that add a second channel see an additional 69% increase. And when cross-channel messaging is coordinated with AI-based personalization, Braze reports an 86% uplift in 6-month retention and 2x average user lifetime.

Product adoption: 37%.Retention: 29%.Lifetime value: 28%.

Those are the top three KPIs for financial services leaders according to Braze. When personalization works well, all three move together.

For a deeper look at how AI Decisioning works and how it differs from traditional predictive models (Next Best Action vs. Next Best Everything), we published a detailed analysis here.

A concrete example from the region

Deuna grew from 50,000 to 2.6 million active users in a market with strong cash preference.

From a focus on real behavior, to the use of data, experimentation, and personalization with Braze AI Decisioning, on March 25 at Fintech Americas they’ll explore how Deuna transformed their wallet into an everyday tool that drives adoption, retention, and growth.

If your team is still running static segmentation and campaigns that depend on technical tickets for every variant, the gap between what your users expect and what they get is widening. The good news: you don’t need to replace your entire infrastructure. You need to connect what you already have (product data, engagement platform, teams) with a decision layer that operates in real time.

Want to see how it applies to your product? Book a demo of Amplitude or Braze with Minders and visit our booth at Fintech Americas.