From technical debt to AI-first platforms: Why insurers must rewrite legacy systems

Despite all the excitement around AI, most insurers are still far from realizing its full value. It's not because of a lack of ambition but because legacy systems, poor data quality, regulatory complexity, and disconnected architectures are getting in the way. As the pressure to modernize grows, it's clear: what's needed is a platform rethink.

Most insurance companies still rely on old technology for core business functions. What's more concerning is that a large portion of IT budgets is being spent just to maintain these systems. That's capital locked in the past instead of fueling the future. Leaders recognize that their legacy tech stack is both expensive and ill-suited for today's application demands.

For instance, years of acquisitions have left behind a maze of siloed policy administration systems, often duplicative and rarely integrated. Migration to the cloud is lagging; even when it happens, insurers struggle to unlock its full value. In many cases, they've migrated to the cloud but haven't unlocked its full potential, constraining their agility and scalability.

AI runs on data, but for many insurers, data is still fragmented, inconsistent, and often inaccessible. Most insurers struggle to extract sustained value from their tech and data spend.

The issue came through loud and clear in Genpact's recent survey with over 200 insurance executives.

• 37% of insurance leaders said poor data quality and availability were the biggest roadblock to AI
• 27% pointed to compliance hurdles, especially relevant in a highly regulated industry dealing with protected health information (PHI) and personally identifiable information (PII)
• 27% cited a lack of tech maturity as a core challenge

Insurers often don't have a single source of truth. It's a data problem that needs to be fixed.

Here's where many insurers go wrong: they experiment with AI at the edge – small pilots, narrow tools, isolated innovations – while the legacy core remains untouched.

This might deliver quick wins, but it won't scale. Without modern foundational platforms, AI systems are forced to work around brittle processes, messy data, and slow feedback loops. That limits impact and often leads to disillusionment.

Real transformation only happens when the core evolves.

AI-first platforms are not just about embedding machine learning models into legacy systems. They're about architecting systems designed to be intelligent from the ground up.

This means shifting toward agentic architectures – systems where autonomous agents can reason, act, and adapt in complex, dynamic environments. These agents aren't just bots – they're intelligent collaborators that operate across workflows, making real-time decisions, orchestrating data, and continuously learning.

We're already seeing this in pockets of insurance:

• Agents triage claims in real time, flagging fraud or routing cases for fast-track processing
• Underwriting agents evaluate submission viability, surface the right deals, and reduce manual waste
• Customer-facing agents personalize interactions, recommend policies, and escalate when needed

But here's the key: these agents only thrive in environments built for them – platforms that are cloud-native, modular, data-rich, and extensible.

Insurers must stop thinking of AI as an add-on and start thinking of it as a catalyst for platform redesign. That means moving:

• From legacy monoliths to composable architectures with APIs, microservices, and interoperability baked in
• From fragmented systems to unified platforms where data flows seamlessly and AI can operate across functions
• From siloed functions to fusion teams, blending IT, data science, product, and compliance into agile delivery units
• From retrofitting compliance to embedding governance, so that transparency and trust are built into every model

Insurance players must let go of legacy and architect for intelligence. Those who move early will reap compounding returns in agility, efficiency, and growth.