Underwriting the unknown: The case for living with legacy

Insurers price risk for a living. They build actuarial models, measure exposure and dedicate entire business functions to understanding liability before accepting it. Yet the most consequential risk inside their own organizations is the slow erosion of institutional knowledge that keeps the business running. 

The same 30- and 40-year-old applications still underpin the industry’s core operations, from policy administration and claims processing to actuarial modeling.

This isn’t because replacing them isn’t on the agenda; it almost always is. The problem is that the industry has been sold two fundamentally different things under the same label.

Modernization is the destination: a reimagined architecture, a cloud-native core and a decommissioned mainframe. Living with legacy is the journey. It’s the practical approach that allows you to move forward, preserve the knowledge embedded in your existing systems and avoid treating transformation as a blank-check exercise.

The risk your actuaries can’t model

For most insurers, the second conversation is now the more urgent one. Yet it’s also the one they’re least likely to have.

The engineers who encoded your underwriting rules, claims logic and product structures are retiring. Much of that logic is expressed in COBOL, which few younger developers can read. As those engineers leave, decades of expertise risk leaving with them.

Call it institutional knowledge decay. It doesn’t appear on a balance sheet, carry a sum insured or fit neatly into an actuarial model. But once that knowledge is lost, it’s gone for good.

In our experience, much of the global insurance industry’s IT spending flows toward replacement decisions made without a clear picture of what is being replaced. Urgency without understanding is speculation dressed as strategy.

Don’t confuse the product you’re buying

Wholesale modernization may well be the right answer — in time. But committing to that level of exposure before you’ve assessed the risk is the IT equivalent of underwriting a policy you haven’t read.

And the cost of inaction keeps mounting.

As Gartner® also states in their 2026 report AI Is Accelerating the Technical Debt You’re Not Tracking: “What makes technical debt strategically dangerous is not its existence. In fact, all systems and organizations carry many forms of debt. The most damaging forms of debt do not appear as defects, backlogs or failed builds. They accumulate quietly in architecture, integration seams, data assumptions and operating models — only becoming visible when they constrain speed, raise transaction costs or block strategic options for change.

“Teams then compensate locally through architecture exceptions, process workarounds and applying undocumented team knowledge. Over time, these compensations compound into systemic drag and complexity that no amount of cleanup can safely resolve.”*

That is an unpriced liability.

• ALSO READ → Why mainframe modernization in insurance is a strategic advantage

Make AI your expert underwriter

This is where AI is fundamentally changing the conversation about legacy systems in insurance.

It can now parse tens of millions of undocumented COBOL lines and produce a living map of what a system actually does, revealing the underwriting rules, claims logic and eligibility criteria that have remained buried in code for decades. It does so with a speed and accuracy that changes the economics of managing legacy technology.

This is the application digital twin: not a replacement system or a migration project, but an AI-generated, continuously maintained representation of your most critical legacy applications and the business rules they enforce.

For the first time, business leaders can see precisely how a policy administration platform behaves, which product rules are hidden in long-forgotten subroutines and where changes can be made safely without triggering unintended consequences elsewhere. The application is no longer a black box and becomes a fully governable asset.

But understanding the system is only the first step. The real value comes from being able to act on that knowledge with confidence, and that’s where the data digital twin comes in.

Then understand where the information goes, and why

If the application digital twin explains how the system works, the data digital twin explains how information moves through it.

It captures the data flows, transformation rules and lineage that give the application its operational meaning — in effect, acting as the COBOL expert who spent decades learning the system from the inside out and converting that knowledge into a structured, reusable form the next generation can access and apply.

Together, the two digital twins close a knowledge gap that has been widening for years. Changes to a rating algorithm that once required months of cautious investigation can now be completed in a fraction of the time, with far greater confidence. They also create a documented audit trail that meets both internal governance requirements and regulatory expectations.

Forrester projects that “industry technology spending will increase by $173 billion in 2026 — up 7.8% relative to last year.”**

The insurers that get the greatest return from those investments build an intelligent foundation for change in this way instead of just replacing systems based on incomplete assessments.

Aim for loss prevention, not emergency response

Insurers build entire product lines around one principle: preventing a loss is almost always cheaper than paying a claim.

The same logic applies to legacy technology. Wholesale core-system replacement is the claims event: disruptive, expensive and typically triggered by years of accumulated technical debt. Living with legacy is the loss-prevention strategy. It uses AI to understand, document and selectively evolve existing systems so that, when modernization does happen, it’s guided by evidence rather than optimism.

We’ve applied this approach in some of the insurance industry’s most mature and business-critical environments, using AI-powered assessment tools that map legacy estates and deliver modernization roadmaps 45% to 55% faster than traditional rationalization methods, according to our own data. That capability is built on more than 35 years of mainframe experience and work across some of the industry’s most complex legacy environments.

Our teams have helped global insurers navigate decades of accumulated policy logic spanning multiple lines of business, where the cost of a misunderstood dependency is measured not in development hours but in regulatory exposure, operational disruption and customer impact. The difference is simple: modernization begins with a clear understanding of what you’ve inherited, not a series of surprises down the line.

It’s time to review your coverage

Once you can see your systems clearly through a digital twin — once the business logic is visible, dependencies are understood and the AI-generated map is established as a trusted reference — the modernization conversation changes completely.

It becomes a structured, evidence-based process rather than a leap into the unknown. You can selectively rearchitect the parts of your estate that will deliver the greatest value for underwriting, claims or customer experience while preserving the institutional knowledge that has kept the business running for decades.

The result is a sustainable path forward, with investment decisions backed by evidence, modernization sequenced according to business impact, and the expertise embedded in legacy systems retained rather than discarded.

Every insurer is living with legacy. The question isn’t whether you have exposure; it’s whether you’ll understand it before the claim arrives.

• ALSO READ → Why insurers’ AI ambitions are exposing the cloud execution gap

* Gartner. AI Is Accelerating the Technical Debt You’re Not Tracking. Howard Dodd. 24 February 2026.

GARTNER is a registered trademark and service mark of Gartner, Inc. and/or its affiliates in the U.S. and internationally and is used herein with permission. All rights reserved.

** Forrester. US Insurance Tech Spending 2026: From Modernization To Intelligence. David Hoffman. 11 February 2026.