Amit Kumar Tyagi describes modernizing the zTPF mainframe not as an escape from legacy systems, but as a disciplined translation of their unmatched reliability into cloud-native, AI-enabled architectures. By using Generative AI to deconstruct decades-old code and applying patterns like strangler migrations, his team preserves mission-critical stability while unlocking agility and faster innovation. Tyagi argues that true success lies beyond migration itself—measured instead by cultural alignment, accelerated time-to-market, and an organization’s ability to continuously reinvent at scale.
The travel industry’s core infrastructure, responsible for processing millions of transactions with sub-millisecond performance, is undergoing a monumental transformation. At the heart of this shift are mainframe systems, engineering marvels of reliability that now present significant modernization challenges. This evolution is not merely about adopting new technology but about strategically deconstructing complex legacy systems to build a more agile, AI-driven future.
Leading one of the largest initiatives in this space is Amit Kumar Tyagi, Vice President of Delivery & Engineering at Coforge-Sabre. With over two decades of experience in technology management across the telecom and aviation sectors, Tyagi is orchestrating a $2 billion program to modernize the industry’s foundational zTPF mainframe system. His work provides critical insights into how enterprises can bridge the gap between legacy stability and cloud-native innovation, using Generative AI as a pivotal tool.
The zTPF mainframe is a cornerstone of global travel, known for its high-speed transaction processing and robust reliability. However, its strengths are coupled with immense complexity. As Tyagi explains, its rigid architecture and dependence on legacy languages make even minor functional changes a significant undertaking. This complexity creates a strong case for modernization, driven by both risk and opportunity.
“A z/TPF system delivers unmatched sub-millisecond performance and extreme reliability, but it comes with significant complexity due to its rigid architecture, tightly coupled business logic, and dependence on legacy languages like C and Assembly,” notes Tyagi. The high cost of maintaining these systems, with some estimates suggesting up to 80% of IT budgets dedicated to legacy maintenance, further fuels the drive for change. This environment makes modernization essential for future innovation.
Integrating these systems with modern platforms adds layers of difficulty. “Integration with cloud-native systems, NDC/REST APIs, and modern DevOps pipelines requires additional adapters and orchestration layers, adding to technical debt,” he adds. According to McKinsey, using AI can accelerate modernization timelines by up to 50%, making it a strategic advantage in tackling these deeply embedded challenges.
Aligning a 2,000-person global team on a multi-year transformation requires more than a project plan; it demands a set of foundational principles. When the journey began, Tyagi and his leadership team established what he calls a “playbook” of non-negotiable principles. The first was anchoring everything on clarity of purpose: “Why we’re transforming, what success looks like, and how every team contributes to long-term architecture.”
This was followed by a shared operating model built on common tooling and a unified cadence to ensure every team operated with the same rhythm. He also notes, “I maintain a commitment to architectural integrity—prioritizing modularity over shortcuts, backward compatibility over quick fixes, and a disciplined, intentional approach to managing technical debt.”
This aligns with structured methodologies like the Cognitive Enterprise Transition Model, which provide a systematic approach for moving from legacy systems to AI-native architectures. This transformation also requires investment in human capital, a sentiment echoed by industry calls to upskill mainframe professionals in cloud computing and AI.
Generative AI serves a dual role in this modernization effort, acting as both an analytical and a creative tool. On the reverse-engineering side, it functions like an X-ray, dissecting monolithic legacy codebases to map their underlying business capabilities. This process is crucial for understanding systems that have been in place for decades, where institutional knowledge may be scarce.
“We point GenAI at large, monolithic legacy codebases—say a 20-year-old C/COBOL or z/TPF module—and have it ingest source, copybooks, config files, and sample logs,” Tyagi explains. The model then produces human-readable artifacts like data models and API candidates. This capability is reflected in industry tools like Google’s Mainframe Assessment Tool, which analyzes languages such as COBOL and JCL to map system dependencies.
For forward engineering, the same AI stack accelerates the creation of new, cloud-native services. “Once a legacy ‘pricing’ flow is understood, we translate it into a set of GCP microservices,” he says. Case studies have shown GenAI can process millions of lines of code in days, a task that would take human teams months, vastly accelerating the transition, as noted in a BCG analysis of a public sector modernization project.
Managing a large, distributed team through a complex, 24/7 modernization program requires a globally cohesive culture yet allows for local ownership. Tyagi’s approach is anchored in a unified engineering operating system with one set of tools, one cadence, and a single definition of “done.” “Whether someone sits in Dallas, Kraków, Montevideo, or Bangalore, the engineering rhythm is identical, which removes friction and builds muscle memory,” he states.
Cultural cohesion is reinforced through cross-geo guilds and architecture councils to ensure decisions are collective, not siloed. “This creates a ‘one team, many locations’ mindset,” Tyagi adds.
A continuous feedback loop is a best practice for improving AI and human performance, involving tracking which AI suggestions are accepted versus rejected. This transparency and AI-enabled execution are critical for maintaining momentum in a 24/7 modernization program.
In any large-scale modernization, the tension between innovation and stability is constant, particularly when dealing with systems that cannot fail. Tyagi recalls a pivotal moment when modernizing a core workflow that had been unchanged for over a decade. The team was eager to introduce a new microservices-based pattern, but a single mistake could have disrupted live airline operations.
Instead of a high-risk “big-bang rewrite,” they adopted a safer approach. “We chose a strangler pattern—building the new service alongside the legacy flow and shifting traffic gradually behind feature flags,” Tyagi explains.
This method, detailed in architectural guides for Azure and AWS, allows for gradual migration while maintaining operational continuity. The team also created a “‘fallback in 60 seconds’ plan,” ensuring that if any issues arose, traffic would instantly revert to the legacy system.
The most misunderstood aspect of mainframes is the belief that they are simply “old technology.” Tyagi argues they are engineering marvels, optimized for a level of performance that modern cloud architectures still struggle to match. The key is understanding why they were built the way they were, not just seeing them as something to escape from.
“What people often miss is that mainframes aren’t monoliths because they’re outdated—they’re monoliths because they were engineered for zero-latency, zero-failure, global-scale transaction processing long before distributed computing existed,” he asserts.
These systems rely on robust transaction software like IBM’s CICS and IMS to manage high-volume, critical operations. The necessary mindset shift is to treat modernization as a translation of proven mainframe principles into cloud-native patterns.
“The best engineers aren’t anti-mainframe or pro-cloud—they’re fluent in both worlds and know how to merge reliability with agility,” Tyagi continues. This human-centric view aligns with best practices for deploying human-in-the-loop AI, where human expertise is essential to guide and validate technological outcomes.
The ultimate measure of success for this transformation extends far beyond technical metrics. For Tyagi, the real indicators are tangible business value, improved economics, and stronger competitive positioning for airline partners. The key performance indicators are outcome-focused, tracking everything from time-to-market for new features to revenue uplift from dynamic offers.
“If airlines can launch offers, ancillaries, or pricing updates in weeks instead of quarters, we know the platform is unlocking value,” he notes. This goal is supported by industry standards like the IATA New Distribution Capability (NDC), designed to give airlines greater control over offer generation. Another key metric is partner confidence, measured by the adoption of new APIs and participation in co-creation sessions.
Tyagi also emphasizes innovation capacity, highlighting: “The ability to launch new products—AI-driven pricing, next-gen shopping, operational optimization—without touching the legacy core.” Modernizing legacy systems is a significant challenge for airlines seeking to implement real-time dynamic pricing, a hurdle this project aims to overcome. A case study of an airline that implemented a dynamic pricing engine saw an 18% revenue increase.
Looking back on this massive undertaking, Tyagi identifies a single, crucial lesson that he believes will define the next decade of enterprise technology transformation. It is the understanding that technology, culture, and execution must evolve together for any large-scale modernization to succeed.
“You can modernize code, move to the cloud, or deploy the best AI engines—but if the organization doesn’t embrace new ways of thinking, working, and collaborating, the transformation stalls,” he states. This perspective reframes modernization not as a technical rewrite but as a systemic shift in a company’s operating model. The benefits of such a holistic approach, as seen in greenfield ERP implementations, include streamlined operations and improved data accuracy.
Lasting transformation is less about the specific technology stack and more about organizational agility. As Tyagi concludes, “The real differentiator isn’t the technology stack—it’s the organization’s ability to adapt, simplify, and continuously reinvent itself at scale.”
The journey to modernize the travel industry’s core systems reveals a path forward for other sectors grappling with legacy technology. By combining a deep respect for mainframe principles with the power of modern cloud and AI architectures, leaders can unlock new value while preserving the stability that underpins mission-critical operations. The insights from this multi-billion-dollar effort underscore that successful transformation is as much about people and principles as it is about platforms.
The expansion introduces agentic decision flows to bolster investigative capabilities for a national security agency in the EMEA region.
Zoho founder Sridhar Vembu warns that companies must prioritize data sovereignty by building internal AI models to protect sensitive intellectual property.
Cities are deploying artificial intelligence on sanitation vehicles to monitor private property, raising complex questions about the future of urban privacy and municipal oversight.
