From Context to Intelligence: Building Self-Improving Products in Banking and Insurance

Most banking and insurance product onboarding flows today consist of a series of decisions someone made once, probably a while ago. Maybe it's been revisited or tweaked occasionally, but maybe not. The screens, the pricing logic, and the question order are stale. A twenty-two-year-old opening their first account and a sixty-year-old renewing a policy see the same content. The product is static despite an ever-changing customer base.

This isn't a criticism of any single institution. It's simply how most financial product development teams have always worked. They design a journey, test it once before launch, ship it, and move on to the next project. The journey then runs largely unchanged for years, reviewed only when a regulatory change forces a redesign or a competitor's product visibly outperforms it.

That model made sense when the cost of really understanding customer behavior was high, and the tools to act on it quickly didn't exist. But neither of those constraints holds anymore.

The institutions that are pulling ahead in financial services have stopped accepting this traditional product development approach. They're building products that improve themselves. Systems where:

• every customer interaction produces a signal
• every signal changes the understanding of who the customer is, what's happening, and what they're trying to achieve
• every understanding changes what the next customer sees, without a team having to redesign the journey each time manually

In this world, context becomes intelligence, and intelligence becomes continuous action.

What does self-improving product actually mean?

The term personalization has been used loosely for over a decade to mean anything from a first-name greeting to genuinely adaptive logic. Most of the time, people are just overselling some version of basic segmentation: grouping customers by static traits and serving each group a fixed variant.

This is not a self-improving system.

A self-improving system is a live loop with three parts:

1. Context: the raw behavioral and transactional data a customer generates
2. Intelligence: the interpretation of what that behavior actually means
3. Action: a product, price, or message that changes as a direct result of that behavior

The test of whether you've built a self-improving product is simple. If the tenth customer on the journey experiences something informed by what the first nine customers did, the loop is working. If not, you still just have a well-designed, static product.

Consider two insurers responding to the same signal: a spike in customers abandoning a health insurance quote at the “family member details” step.

A segmented product might add a help tooltip to that screen for everyone.

A self-improving product detects the abandonment pattern within days. Tests whether the friction is the number of fields, the wording, or the order. Ships the version that performs best to new customers immediately. Keeps testing the next likely friction point without waiting for a review cycle to authorize it.

The output looks similar on the surface, but the operating rhythm underneath is entirely different, and that rhythm is the actual asset.

A proof point from insurance: ACKO

ACKO, the Indian insurtech that grew from a single-category motor insurer into a multi-line unicorn, provides a great example of this loop in financial services.

ACKO ran a cross-sell campaign offering existing car insurance customers a time-bound discount on health insurance: 25 percent for the first thirty days, tapering to 10 percent after. The goal was to deepen the relationship with customers who already trusted the brand for one product. Straightforward logic.

What the behavioral data revealed was less obvious. The team had priced the discount just below competitor rates rather than purely on margin, and tracked the full customer journey from visit to quote to payment as the campaign ran. The result was a 41 percent increase in the cross-sell rate, and customers who took the discount converted from quote to payment 200 percent faster than the baseline.

Critically, average order value didn't fall as you'd predict with a discount. Existing customers used the lower entry price not to buy the minimum, but to add broader family coverage. This brought their spend back up even as the unit price came down.

The discount, in other words, was read by customers as a signal of trust rather than a race to the bottom. But the ACKO team only knew that because it was watching the behavior closely enough to see it happen in real time. Combined with a separate Cricket World Cup marketing campaign, the effects compounded, delivering a 209 percent quarter-on-quarter revenue increase.

A second, quieter signal from the same business shows the same loop operating at a smaller scale. ACKO had only ever been able to see whether a customer clicked a product video, a binary yes or no. Once the team could see how many times a customer replayed a specific educational video, they discovered that repeated viewing was a strong indicator of purchase intent. That one piece of context is now a live input into how content is sequenced for the next customer every cycle. Video views rose almost tenfold during the same campaign period, and app downloads rose 25 percent on Android and doubled on iOS.

Neither of these outcomes came from a bigger discount or a flashier campaign. They came from a business willing to treat behavioral context as an input to the next decision, not just a report on the last one.

Why banking and insurance are a harder version of this problem

In retail or media, getting something wrong costs a wasted impression or a lost sale. In banking and insurance, the risks are greater. A product that adapts too aggressively toward what a customer clicks, rather than what genuinely serves them, risks steering someone toward a policy or a credit product that's wrong for their circumstances, damaging trust and inviting regulatory attention.

In this world, a self-improving success isn't a higher conversion rate. It's the right fit for the right customer. ACKO's example shows how a self-improving loop should operate within the constraints of a regulated market. The team was not experimenting with the product's terms. They were experimenting with timing, framing, sequencing, and channel without touching the underlying risk or compliance model. The loop should run fastest on the parts of the journey furthest from the regulated core, and slowest, and with the most oversight, closest to it.

Which parts of your customer journey are fixed by regulation or risk appetite, and which parts are simply the way you've always done it? In most institutions, the second category is far larger than the first, and it's where there's the greatest opportunity to build a self-improving loop.

Pricing floors, capital requirements, and disclosure obligations are fixed. The order of questions on a form, the wording of a reminder, and the timing of a cross-sell offer are not. They're fair game, yet so rarely revisited.

This is the true competitive advantage and frontier

Data isn't what separates leaders and laggards in financial services today. Almost every organization has more data than they use. The gap is between institutions that still treat customer intelligence as something a team consults sporadically and those that build it into the product itself. Organizations that do this well ensure that every customer interaction generates insights that make the experience marginally better for the next customer.

That's the real competitive frontier in banking and insurance. It's about who builds the shortest, most trustworthy loop between context and action, and who's disciplined enough to keep it running.

Retail banks are learning to compete on product experience rather than rate, with your product continuously improving, on its own, informed by every customer who passes through. The banks and insurers that get there first will have built the habit of listening and acting faster than anyone else in the category.