Mitra Innovation, a cloud-native software and digital transformation company, has built its practice around a conviction that understanding legacy systems must come before transforming them. The firm positions itself as the category creator for legacy intelligence, the idea that decades of accumulated business logic trapped in aging systems represents more than technical debt; it is institutional knowledge that enterprises cannot afford to lose.
In a conversation with Echelon, three leaders from Mitra Innovation discuss what that means in practice.
Dr Ashok Suppiah, Co-founder and Chairman at Mitra Innovation, suggests that the industry’s fixation on migration overlooked a deeper knowledge crisis, and that AI was always the only viable tool to decode it. Thilina Herath, Chief Technology Officer at Mitra Innovation, explains how the company’s novolingo platform approaches legacy systems as a forensic investigation, rebuilding institutional intelligence through an agentic AI architecture. Dammika Ganegama, Co-founder and President at Mitra Innovation, makes the case for why enterprise leaders should understand their operational reality before committing to transformation, and why the hidden knowledge problem is often discovered too late.
Mitra has argued for understanding technology before transforming it. How early did you see AI-driven legacy intelligence as inevitable, and what were others getting wrong?
Dr Ashok Suppiah: We identified the global retirement of the engineers who built our world’s core infrastructure, over a decade ago. While the market was obsessed with simply moving old code to a new cloud, we knew this was a recipe for expensive failure.
Others were treating AI as a bolt-on feature; we saw it as the only way to decode decades of business logic trapped in forgotten languages like UniBasic, COBOL, and RPG. We built our first machine learning product in 2015 because we knew that intelligence must precede transformation.
Today, we provide the “legacy intelligence layer” that helps enterprises understand how their existing technology actually works. The world’s largest enterprises, from the NHS to Aetna and major financial institutions, rely on us to understand their past and own their future.
Is the legacy problem fundamentally about technology, or about knowledge? And how does your answer shape the way Mitra innovates?
Dr Ashok: It is fundamentally a knowledge crisis. Legacy systems are not just technical debt; they are the undocumented knowledge of an organisation. This is decades of business rules, edge cases, and logic that no living employee can fully articulate.
At Mitra AI, we innovate to solve this through our proprietary novolingo platform. We’ve moved beyond simple code translation to creating the Legacy Knowledge Graph (LKG), a multi-layered data structure that captures the operational core of a business.
This allows enterprises like Travis Perkins and Biffa to innovate without breaking the system. By turning old, forgotten code into a machine-readable asset, we empower the C-Suite to drive unparalleled financial outcomes.
For the CRO, this accelerates time-to-market for new revenue streams by 20-30%. For the CFO, this slashes operating costs and technical debt by shifting from expensive legacy maintenance to efficient, AI-driven automation. For the COO, it maximises productivity over 30% by exposing legacy logic to modern AI agents and systems.
You have built a global tech company with proprietary IP, partly out of Sri Lanka. What does that say about where world-class innovation can come from?
Dr Ashok: It proves that world-class innovation is no longer a zip code. It is a mindset. We have built a profitable, global firm with 80%+ recurring revenue by solving the hard cases that Silicon Valley and Systems Integrator giants often ignore.
Our global team, including our high-calibre engineering talent in Sri Lanka, has built IP that is now the central storage system for some of the world’s most critical infrastructure. Whether it’s leveraging healthcare data for the NHS or driving digital agility for Sampath Bank, we show that the most defensible technology comes from deep domain expertise and a 14-year track record of zero-incident delivery.
We aren’t just participating in the AI revolution; we are leading the global charge to decode the past and accelerate a more productive, sustainable future for every enterprise.
What is the hardest technical challenge in teaching AI to interpret code that was never meant to be read by anyone but its original authors?
Thilina Herath: The real challenge is not syntax, it is reconstructing intent at enterprise scale. Many of the systems we work with contain over 20 million lines of code accumulated across decades of undocumented changes, hidden dependencies, and evolving business rules.
Novolingo decodes relationships across code, databases, APIs, batch flows, and operational behaviour to rebuild the institutional intelligence behind the platform.
What does a discovery actually look like on novolingo when the AI encounters a legacy system for the first time?
Thilina: The first interaction feels less like a code scan and more like a forensic investigation. Within hours, novolingo analyses millions of lines of code, databases, APIs, configurations, logs, and inter-system relationships to understand how the enterprise truly operates.
The platform builds a living intelligence layer that exposes hidden dependencies, business rules, operational risks, and undocumented workflows, while its governance layer highlights technical debt, fragile integrations, security exposures, and unused code across the legacy estate.
Errors in AI-generated intelligence about mission-critical systems can be catastrophic. How have you built trust into Novolingo?
Thilina: We build for high-stakes industries such as banking, healthcare, and large-scale enterprise operations, where reliability, auditability, and observability are non-negotiable.
Novolingo uses an agentic AI architecture with multiple specialised LLMs validating and cross-checking intelligence before it becomes enterprise insight. Combined with traceability, confidence scoring, observability layers, and SME feedback loops, the platform delivers reliable and auditable intelligence for mission-critical systems.
L-R: Dammika Ganegama, Co-founder and President at Mitra Innovation; Thilina Herath, Chief Technology Officer at Mitra Innovation
How do you persuade enterprise leaders that the smartest first move is to understand what they already have?
Dammika Ganegama: Enterprise leaders are under pressure to modernise fast, reduce costs, adopt AI, improve customer experience, and deliver measurable outcomes. The instinct is often to jump directly into cloud, automation, or AI initiatives. For Mitra and the novolingo platform, our advice is to accelerate intelligently and modernise progressively.
Before investing millions into AI, cloud, or transformation, leaders need clarity on workflows, dependencies, bottlenecks, and customer journeys. Novolingo helps enterprises see their operational reality before redesigning their future.
Executives must understand that they cannot modernise what they cannot see. Most enterprises operate with fragmented systems, undocumented processes, hidden dependencies, duplicated workflows, and tribal knowledge.
The biggest transformation risk is not moving too slowly, it is modernising the wrong thing with incomplete visibility. Novolingo provides operational visibility, process intelligence, automation discovery, and AI-ready workflow mapping.
We also refer to novolingo as “the essential intelligence layer before transformation” because it’s not just another audit. It is a rapid operational intelligence capability that reduces risk, accelerates ROI, and enables confident prioritisation.
Many transformation failures happen because technology is implemented before understanding how work actually happens.
AI readiness has to start with operational understanding because AI is only as effective as the operational context behind it. Novolingo helps enterprises uncover AI opportunities, identify high-value automation, structure enterprise knowledge, and prepare operational foundations for AI adoption.
Across banking, insurance, and healthcare, what patterns do you see in how enterprises recognise, or miss, the scale of their hidden knowledge problem?
Dammika: Enterprises in these sectors often believe they have a technology problem, until transformation efforts reveal the deeper issue of a hidden knowledge problem. Transformation fails when organisations digitise complexity they have never fully understood.
Critical operational knowledge exists across systems, policies, emails, SOPs, ticketing platforms, vendor documents, and with long tenured employees. But it is fragmented, inconsistent, undocumented, and disconnected from operational context.
The challenge is not lack of information. It is a lack of operational understanding.
This becomes visible during AI adoption, automation, cloud migration, compliance modernisation, merger integration, or digital transformation initiatives.
Organisations suddenly realise they cannot modernise what they do not fully understand, and the same patterns emerge across sectors: undocumented workarounds, shadow processes, duplicated controls, manual exceptions, dependency on tribal knowledge, and critical decisions embedded in people rather than systems.
Generative AI has accelerated this exposure further. Many enterprises assumed their processes and data were AI-ready, only to discover fragmented knowledge, conflicting policies, and poor process traceability.
Novolingo helps enterprises move from disconnected information towards operational intelligence, institutional visibility, and AI-ready enterprise understanding.
Mitra invests in ventures and runs community programmes. How does that broader thinking shape the legacy intelligence platform?
Dammika: Novolingo is not positioned as just another legacy modernisation platform. Most platforms focus on systems, applications, and migration risk. Novolingo’s broader ecosystem mindset is shaped by enterprise transformation, venture building, workforce development, and AI-enabled operating models, allowing it to operate at a far more strategic level.
The platform is not simply interpreting systems. It is interpreting institutional intelligence.
Traditional vendors ask: “What applications do you have?”
The novolingo intelligence layer asks: “How does your institution actually function, govern, decide, and evolve?”
That shifts the conversation from system discovery to understanding operational behaviour, institutional memory, governance logic, regulatory adaptation, and hidden workflow intelligence.
This is especially powerful for banks, insurers, healthcare providers, and national-scale enterprises where critical knowledge often lives outside documentation, instead embedded in people, processes, and operational workarounds.
The result is a platform positioned not only for modernisation, but for institutional resilience, AI readiness, operational continuity, governance intelligence, and enterprise cognition.
The strategic distinction is simple:
“Legacy is not old technology. Legacy is accumulated institutional intelligence.”