For more than six decades, COBOL has quietly powered the world’s most critical systems. In Part 1 of this article, we examined the criticality of COBOL applications still running in organisations today, the key drivers for modernisation and the associated risks.
In Part 2, we will review common modernisation approaches and how AI can assist in addressing some of the risks associated with modernisation programs.
Common Modernisation Approaches
Organisations typically choose from several strategies, each with their own trade-offs. Recent research on application modernisation emphasises that successful programs employ structured assessment, phased execution and strong alignment with business outcomes, rather than relying on a single technical pattern (Chindanuru, 2025).
1. Portfolio Assessment and Sequencing
Modernisation initiatives increasingly begin with a formal assessment of the application portfolio. According to industry research summarised in Modernising Legacy Applications: Strategies and Emerging Trends, organisations that apply structured assessment frameworks experience significantly fewer migration incidents and faster return on investment (Chindanuru, 2025). These assessments evaluate business criticality, technical complexity, regulatory exposure and modernisation risk before deciding on an approach.
This portfolio-driven sequencing enables organisations to modernise selectively – prioritising high-value, lower-risk systems first rather than attempting extensive transformation.
2. Incremental Modernisation and the Strangler Pattern
Rather than full rewrites, many organisations adopt incremental modernisation patterns. The article highlights the effectiveness of gradual decomposition techniques, such as the Strangler pattern, which allow legacy functionality to be progressively replaced while keeping core services operational. This approach reduces business disruption and supports continuous delivery alongside legacy stability.
3. Cloud, Containers and Hybrid Architectures
Modernisation strategies increasingly combine cloud migration with containerisation and hybrid deployment models. Research cited by Chindanuru (2025) shows that organisations using container orchestration platforms achieve faster deployment cycles, improved scalability and lower infrastructure costs, while retaining control over sensitive or regulated workloads through hybrid and multi-cloud architectures.
4. AI-Assisted Modernisation
Artificial intelligence is emerging as a key enabler in large-scale legacy transformation. AI-driven code analysis, dependency mapping and business-rule extraction significantly reduce the time required to understand complex legacy systems. Studies referenced by Chindanuru (2025) demonstrate that AI-assisted approaches improve modernisation accuracy, shorten delivery timelines and reduce post-migration defects – particularly in very large codebases.
5. Governance, Risk, and Sustainability
Finally, the research stresses that the success of modernisation depends as much on governance and risk management as on technology. Organisations that embed continuous risk assessment, security controls and compliance checks into their modernisation processes report higher success rates and fewer operational incidents. Sustainable modernisation is framed as an ongoing capability and not a one-time project.
How AI Can Assist with Legacy Modernisation
Artificial intelligence is increasingly emerging as a practical enabler of legacy modernisation rather than a speculative future capability (Kansal & Siddharth, 2024). Recent research on adaptive AI models demonstrates that AI can significantly reduce the risk, cost and duration of modernising complex legacy systems.
1. Automated System Understanding
One of the greatest challenges in modernising legacy solutions – particularly COBOL systems – is understanding what the system does. Adaptive AI models can analyse large legacy codebases, extract structural and behavioural patterns, and map dependencies faster than manual techniques. Research shows that AI-driven analysis significantly improves the accuracy of identifying business logic, data flows and system interactions, even in poorly documented environments (Kansal & Siddharth, 2024).
2. AI-Assisted Code and Data Migration
AI techniques such as natural language processing and deep learning can assist in translating legacy code constructs into modern equivalents and in mapping legacy data structures to contemporary schemas. Empirical results from AI-driven migration experiments demonstrate higher accuracy and significantly lower error rates compared to traditional manual approaches, particularly in code translation and data mapping tasks (Kansal & Siddharth, 2024).
3. Optimising Migration Strategy with Adaptive Learning
Unlike static migration tools, adaptive AI models continuously learn from migration outcomes. Reinforcement learning techniques allow AI systems to optimise migration sequencing, tooling choices and execution paths based on real-time feedback. This adaptability enables organisations to respond dynamically to unexpected system behaviours, reducing downtime and minimising business disruption during modernisation programs.
4. Automated Testing, Validation and Risk Reduction
Testing and validation represent a major cost and risk factor in legacy modernisation. AI-driven validation models can automatically compare legacy and modernised system behaviour, identify discrepancies and reduce regression risks. Studies indicate that AI-assisted testing significantly shortens validation cycles while improving overall migration accuracy and system reliability (Kansal & Siddharth, 2024).
A Human–AI Collaboration Model
Importantly, the research emphasises that AI does not replace domain expertise. Instead, the most effective modernisation approaches combine AI automation with human oversight. AI handles repetitive, large-scale analysis and execution tasks, while experienced engineers and business experts validate outcomes and manage exceptions. This collaborative model delivers faster modernisation while preserving institutional knowledge and control.
Where AI Should Not Be Used
While AI can significantly accelerate and de-risk legacy modernisation, it is not a universal solution. Research and real-world experience highlight several areas where ungoverned AI use can introduce new risks rather than eliminate existing ones:
- Unsupervised business-rule decisions: AI should not independently redefine business logic, pricing rules or regulatory behaviour without human validation.
- Regulated decision accountability: In highly regulated environments, accountability for system behaviour must remain clearly assigned to human owners.
- One-click end-to-end migrations: Fully autonomous migrations increase the risk of hidden defects, data inconsistencies and operational surprises.
- Obscure models without explainability: AI models used in modernisation must be auditable and explainable to support trust, compliance and governance.
The most successful programs treat AI as a precision tool, not a replacement for architectural judgment, domain expertise or organisational responsibility.
Key Success Factors
Organisations that modernise COBOL systems successfully tend to:
- Treat modernisation as a business program, not just a technical project.
- Invest heavily in system understanding and documentation.
- Retain or augment COBOL expertise during transition.
- Prioritise incremental change over “big bang” migrations.
- Align modernisation goals with customer and regulatory needs.
Conclusion
COBOL applications remain among the most significant yet understated components of the global technology landscape. Their scale and importance mean that modernisation carries both enormous opportunity and substantial risk.
What is often overlooked is that COBOL’s durability was not an accident of history. From its inception, it was engineered to be readable, portable and resilient – qualities that naturally led to decades of continuous use (Sammet, 1981).
For organisations today, the challenge is not simply replacing old technology, but carefully evolving systems that continue to underpin trust, revenue and client service. Done thoughtfully, COBOL modernisation can protect the past while enabling the future.
Daleen Meinhardt
Solutions Architect
