Agentic AI for claims management
Capco and Yarowa built an AI-powered claims workflow, cutting manual interactions by 90%, speeding provider vetting, and delivering production-ready solutions in weeks.
Client challenge
A Wealth Management division of a global Tier 1 bank was preparing to embark on a large-scale legacy modernization journey. Their existing infrastructure was showing its age – driving up operational costs, creating critical talent and continuity risks, and leaving them vulnerable to high project failure rates. On top of this, the organization struggled to gain a clear view of its true Total Cost of Ownership and Return on Investment.
Success would require untangling complex legacy code, extracting vital business logic, and shaping a high-quality, future-proof target architecture – all without losing sight of a human-centric developer experience.
Success would require untangling complex legacy code, extracting vital business logic, and shaping a high-quality, future-proof target architecture – all without losing sight of a human-centric developer experience.
How Capco helped
We began with a focused Proof of Value built on the client’s own mainframe code, allowing us to demonstrate our legacy modernization approach in action. Using GenAI-infused pipelines, we automated the reverse engineering process and extracted the business rules embedded in their systems.
From there, we generated structured documentation, semantic code models, and a target state architecture, drawing on configurable templates to accelerate and standardize the work. Our methodology kept humans firmly in the loop, with deterministic controls ensuring every step was explainable, high-quality, and safe.
Flexibility was built into the heart of the solution, with a pluggable architecture that could adapt to different legacy and target technology stacks, for both frontend and backend code, and that considered the customers architectural patterns and engineering standards and best practices. We developed a resilience-focused system anchored by a graph-based source of truth, reinforced with embedded AI governance to safeguard decision-making.
Logic rules harvested from the legacy code were turned into test scenarios that, in turn, were injected into the forward engineering processes towards creating Unit Tests that were designed to test all the nuanced behavioral aspects of the legacy code.
Outcomes achieved
The results speak for themselves. The PoV we delivered not only proved scalable but also set us apart from competitors, winning the client’s trust thanks to our emphasis on flexibility, explainability, human oversight, and the fact that the code that was generated did not appear to be machine generated, but closely resemble code that their bouest engineers would generate by hand.
Productivity gains were dramatic – up to 90% in reverse engineering and 75% in forward engineering – while fine-grained control over code migration became possible through sophisticated, testable automation ‘recipes’.
The client emerged from the engagement with a transparent, adaptable, and flexible approach to legacy modernization, positioned to move forward with confidence. By mitigating AI-related risks such as hallucinations and context drift through a semantic network and governance-driven LLM gateway, we reduced their dependency on scarce legacy expertise and paved the way for a faster, safer modernization journey.
AI Infused success stories
