AI-Powered COBOL Modernization: What Canadian Enterprises Need to Know

On February 23, 2026, Anthropic announced that Claude Code can now modernize COBOL codebases — mapping dependencies, documenting workflows, and identifying risks across thousands of lines of legacy code. IBM stock dropped over 10% on the news as markets priced in what this means for the legacy modernization industry. For Canadian enterprises still running COBOL in banking, government, and transportation, this is the most significant shift in legacy modernization economics in decades.

The scale of COBOL in production is staggering. An estimated 95% of ATM transactions still run through COBOL code. Hundreds of billions of lines remain in active production worldwide. And the talent pool capable of maintaining these systems is shrinking every year — the average COBOL developer is over 55, and Canadian universities stopped teaching the language decades ago.

Why Has COBOL Modernization Been Stuck for Decades?

Legacy code modernization has been stuck because understanding legacy code costs more than rewriting it. That single economic reality has kept billions of lines of COBOL running long past the point where anyone thought it made strategic sense.

The barriers are well documented:

• The analysis bottleneck. Before you can migrate a single line of COBOL, you need to understand what it does, what depends on it, and what breaks if you change it. For a large codebase, that analysis alone can take years of senior developer time.
• Tribal knowledge locked in retiring developers. Critical business logic often exists only in the heads of developers who wrote it 20-30 years ago. When they retire, that knowledge leaves with them — and no documentation exists to replace it.
• The risk of touching working systems. COBOL systems in Canadian banks process billions of dollars in transactions daily. The downside risk of a failed migration dwarfs the upside of modernization, so the default decision is always to leave it running.
• Failed rip-and-replace projects. The industry is littered with multi-year, multi-million-dollar COBOL migration projects that went over budget, over timeline, or were abandoned entirely. These failures make every subsequent business case harder to approve.

In Canada specifically, the stakes are enormous. The Big Five banks — RBC, TD, Scotiabank, BMO, and CIBC — all run significant COBOL workloads, as we explore in our deep dive on mainframe modernization for Canadian banks. Federal government systems including the CRA's legacy tax processing systems and immigration processing rely on COBOL. Air Canada's reservation and operations systems have deep COBOL roots. These are not niche legacy systems — they are the infrastructure that Canadian commerce and government run on.

How Does AI Change the Economics of Legacy Code Migration?

AI flips the fundamental equation that has kept COBOL modernization stalled. The biggest cost driver in any legacy migration is not the rewriting — it is the understanding. Developers spend 60-70% of total project time reading, tracing, and documenting existing code before they write a single line of the replacement.

AI automates exactly that phase. Here is what tools like Claude Code can now do with COBOL codebases:

• Dependency mapping across thousands of files. AI can trace execution paths, identify shared data structures, and map every dependency in a codebase that would take a human team months to catalog.
• Automated workflow documentation. The AI reads the code and produces human-readable documentation of what each module does, how data flows between components, and where business logic lives.
• Risk identification. AI flags tightly coupled components, global state dependencies, undocumented side effects, and other risks that would otherwise surface as surprises mid-migration.
• Hidden dependency discovery. COBOL codebases often have coupling patterns that are invisible to anyone who did not write the original code — shared copybooks, implicit data flows through JCL job streams, and undocumented inter-program dependencies. AI finds these systematically.

The key insight: legacy code modernization stalled because understanding legacy code cost more than rewriting it. AI compresses the understanding phase from years to weeks, making the total project economics viable for the first time.

What Does the AI-Assisted Modernization Process Look Like?

AI-assisted COBOL modernization follows a four-phase methodology that combines AI speed with human judgment. This is not a fully automated process — it is a structured approach where AI handles the labor-intensive analysis and humans make the strategic decisions.

Phase 1: Automated Exploration and Discovery

AI tools ingest the entire COBOL codebase and produce a comprehensive map of the system:

• Entry points and execution paths through every program
• Data flows — where information originates, how it transforms, and where it lands
• Dependencies across hundreds or thousands of files, including copybooks and JCL
• Dead code identification — modules that are compiled but never executed

Phase 2: Risk Analysis and Opportunity Mapping

With the discovery map in hand, AI performs a systematic risk assessment:

• Coupling analysis — which components are tightly bound and must migrate together
• Refactoring opportunities — where the codebase can be simplified before migration
• Component risk scoring — ranking each module by complexity, dependency count, and business criticality
• Technical debt inventory — cataloging workarounds, patches, and undocumented modifications accumulated over decades

Phase 3: Strategic Planning with Expert Oversight

This is where human expertise becomes critical. AI suggests a prioritized migration sequence, but experienced architects and business stakeholders validate and refine:

• Migration priority sequencing based on business value and technical risk
• Target architecture definition — what the modernized system should look like
• Code standards and patterns for the target language
• Testing strategy — how to verify that migrated code behaves identically to the original

Phase 4: Incremental Implementation

Migration happens one component at a time, with rigorous validation at every step:

• AI-assisted code translation with human review of every module
• Function-level testing — each translated function must produce identical outputs to the original
• Side-by-side execution scaffolding — old and new code run in parallel until the new version is validated
• Performance benchmarking — ensuring the migrated code meets or exceeds original throughput requirements

What Does This Mean for Canadian Enterprises?

The implications for Canadian enterprises are immediate and significant. Canada has a disproportionately large COBOL footprint relative to its developer population, which makes the talent crisis particularly acute.

Canadian banking. The Big Five banks collectively maintain millions of lines of COBOL powering core banking, payment processing, and regulatory reporting. AI-assisted modernization gives them a viable path off COBOL without the existential risk of a failed big-bang migration.

Federal government. The CRA, IRCC, and other federal agencies run critical systems on COBOL that directly affect millions of Canadians. The federal government has struggled to attract and retain COBOL talent, and AI-assisted modernization could break the staffing bottleneck that has delayed modernization for years.

Airlines and transportation. Air Canada and other Canadian carriers have deep COBOL dependencies in reservation, operations, and crew scheduling systems. These systems require 99.99% uptime, making the incremental, validated approach of AI-assisted migration particularly attractive.

Insurance. Canadian insurance companies — including major players like Manulife, Sun Life, and Intact — run policy administration and claims processing on COBOL. The regulatory environment adds complexity, but AI-assisted analysis can map compliance-critical logic before any code moves.

The practical implication for all of these sectors: Canada has a smaller developer pool than the US, which means the COBOL talent crisis hits harder here. AI tools do not just make modernization faster — they make it possible for organizations that could not find enough skilled developers to even start the project.

For a broader framework on building the business case for AI on legacy enterprise systems, see our guide on making the business case for AI in legacy ERP. For context on how Oracle and SAP are approaching AI in their own modernization roadmaps, see what Oracle's and SAP's AI roadmaps mean for your 2026 IT strategy.

What Should CIOs Do Right Now?

If your organization runs COBOL, the announcement from Anthropic changes your options materially. The COBOL developer shortage in Canada makes this particularly urgent, and understanding the realistic costs of COBOL to Java migration is the first step toward building a business case. Here are five concrete actions to take in the next 90 days:

1. Inventory your COBOL footprint. Map every COBOL system, its business criticality, upstream and downstream dependencies, and the teams that maintain it. You cannot plan a modernization you have not measured.
2. Assess your talent risk. How many of your COBOL developers are within five years of retirement? What is your plan if two of them leave in the same quarter? The talent risk is often more urgent than the technology risk.
3. Run a pilot analysis. Pick a non-critical COBOL subsystem and run AI-assisted analysis on it. The goal is not to migrate it — the goal is to see what AI tools produce and calibrate your expectations for a larger effort.
4. Build the business case. Use the pilot results to build a phased modernization proposal. Our CIO playbook for AI on legacy systems provides a framework for structuring the financial case, aligning stakeholders, and sequencing the rollout.
5. Do not wait for a crisis. Proactive, planned modernization costs 2-4 times less than emergency migration triggered by a system failure or the sudden departure of a key developer. The economics of AI-assisted modernization have just shifted dramatically in your favor — act while you have the luxury of planning.

Key Takeaways