Analyst(s): Mitch Ashley
Publication Date: February 26, 2026
Anthropic’s Claude Code blog on COBOL modernization sparked a sensitive topic. Should we rely upon Claude or IBM to modernize COBOL? Modernizing applications is more than converting codebases to a new programming language.
What is Covered in This Article:
- Anthropic published a technical blog on February 23, 2026, positioning Claude Code for COBOL modernization, targeting the discovery, analysis, and documentation phases that make legacy migration prohibitively expensive.
- IBM’s Project Bob, an AI-first IDE built on VS Code with a multi-model architecture that includes Claude, targets the same COBOL modernization buyers. IBM i CTO Steve Will confirmed Project Bob will replace watsonx Code Assistant for i, covering RPG, CL, SQL, COBOL, Java, and Python on IBM i at launch.
- Successful COBOL modernization programs require business scoping, technical assessment, data migration planning, behavioral equivalence validation, observability, and organizational change management, in addition to code translation.
- AI tools accelerate the code analysis and transformation phases of modernization, but do not yet address the full program scope that determines whether migrations succeed or fail.
- Enterprises in regulated industries face additional requirements around audit trails, behavioral equivalence, and compliance that neither Claude Code nor Project Bob fully resolves today.
The News: Anthropic’s February 23, 2026, blog post positions Claude Code directly against that problem, arguing that AI changes the economics of COBOL modernization by automating the phases that make it expensive.
The post describes a structured methodology. Claude Code reads the full COBOL codebase and maps dependencies, identifies program entry points, traces execution paths, and surfaces implicit couplings through shared data structures and file operations that static analysis tools miss. Workflow documentation follows automatically, generating diagrams of processing pipelines that exist only in the code itself. Risk assessment then identifies high-coupling modules, isolated components ready for early migration, and areas of accumulated technical debt.
The market reaction was immediate. IBM shares fell 13% on February 23, their worst single-day drop since October 2000, with the stock down 27% in February. Accenture and Cognizant also declined, reflecting investor concern that both companies’ legacy modernization practices face the same disruption risk.
IBM responded the same day, publishing a counter-argument that the modernization challenge is not a COBOL language problem but an IBM Z platform problem. The company’s position: translating code captures almost none of the actual complexity, and the platform’s value comes from decades of hardware-software integration that code translation does not move.
IBM vs. Anthropic: A Tale of the COBOL Modernization Tape
Analyst Take: COBOL modernization has been a persistent enterprise problem for decades: the code runs critical systems, the engineers who wrote it retired, and the institutional knowledge left with them.
Anthropic’s blog and IBM’s Project Bob both address the code analysis and transformation layer of modernization. It is also one phase of a program that touches business strategy, technical architecture, data migration, behavioral validation, testing, observability, governance, organizational skills, and change management simultaneously. Choosing the right AI tool for code discovery while skipping the other dimensions does not produce a successful migration. It produces faster discovery of a program that still fails.
What Modernization Actually Requires
The road to application modernization is littered with project failures, and COBOL programs have more than their share. They fail on predictable fronts. Business goals get defined too loosely: “move off the mainframe” is not a success criterion. Teams skip deep technical assessment, assuming one strategy fits all components when the right answer is nearly always keep some, refactor some, replatform some, and replace some. Data migration and integration get underestimated consistently; data loss, corruption, and broken interfaces are among the highest-impact risks in any COBOL migration.
Behavioral equivalence is where regulated industries hit the wall. Regression testing, golden datasets, and parallel runs are required to confirm the new system matches legacy behavior where it matters, not just on known test cases. Edge cases accumulated over decades encode regulatory logic and business rules that exist nowhere else.When migrated code diverges from legacy behavior in production, passing tests proves nothing. The real questions are whether the legacy behavior was itself correct and whether anyone can prove it.
Observability has to be in place before changes go in, not after they break something. Feature flags, metrics, and rollback capability are not nice-to-haves in a COBOL migration. They can be the difference between a contained problem and an incident that reaches the board. And none of this works without the people who understand the legacy system staying involved through the entire program. The institutional knowledge that is left with retiring developers lives partly in the code. It also lives in the heads of the domain experts still on staff.
Where AI Tools Actually Fit
Against that backdrop, Anthropic’s Claude Code and IBM’s Project Bob address a real and valuable slice of the problem. Code discovery, dependency mapping, implicit coupling detection, workflow documentation, and initial risk stratification are genuinely difficult and genuinely expensive at scale. AI that compresses those phases from months to weeks to days and hours changes the program economics meaningfully.
IBM’s approach is structurally stronger for enterprise programs. Project Bob runs a multi-model architecture that routes Claude, Mistral, Meta’s Llama, and IBM’s own models to tasks based on accuracy, latency, and cost. IBM i CTO Steve Will confirmed at TechXchange 2025 that Project Bob replaces watsonx Code Assistant for i, covering COBOL for i, RPG, CL, SQL, Java, and Python at launch, delivering capability “far faster than we were going to be able to do it with our old approach.” Project Bob also embeds FedRAMP, HIPAA, and PCI compliance context and inline security scanning directly in the IDE, which matters for regulated industries where the audit trail starts at the first line of changed code.
Tale of the Tape
IBM or Anthropic: Who has the farther reach, greater speed, and the best team? The right sequence is to define business success criteria, assess the full technical estate, identify which components to keep, refactor, replatform, or replace, and establish the behavioral equivalence and data migration strategy before the tool selection decision becomes load-bearing. Applying AI to that program architecture work may deliver higher ROI than applying it to code rewriting alone.
Claude Code gives development teams direct access to a capable model with no intermediary, which compresses timelines for teams ready to assemble their own governance, testing, and validation layers, backed by one of the fastest innovation cadences in the industry.
IBM brings something different: institutional knowledge of enterprise-scale modernization programs, operational depth with the largest regulated organizations globally, an ecosystem of partners equipped to manage the business, organizational, and process dimensions that code analysis never touches, decades of data management experience, and its own AI models embedded in a platform purpose-built for this work.
The comparison is not Claude Code versus Project Bob. It is who can put together the best team for modernization, inclusive of technology choices.
The best answer may well be a combined IBM plus Claude team.
What to Watch:
- Project Bob’s general availability timeline is load-bearing for IBM’s position. The product is in tech preview with a waitlist. If enterprise COBOL buyers move toward Claude Code while Project Bob remains in preview, Anthropic’s blog will have done its job. Watch for IBM’s GA announcement and the customer evidence that comes with it.
- Behavioral equivalence tooling is the gap neither vendor has closed. The first platform to demonstrate provable behavioral equivalence between legacy COBOL and migrated code, with a referenceable production deployment in banking or government, changes the compliance conversation for the entire market.
- SI response shapes enterprise adoption. Accenture, Capgemini, and other large integrators built COBOL modernization practices over decades. They can build on Claude Code, Project Bob, or both. Watch whether they position AI tools as accelerators inside their existing methodology or let tool vendors define the program scope directly with enterprise buyers.
- Amazon Q Developer and GitHub Copilot are investing in legacy modernization with enterprise distribution advantages that neither Anthropic nor IBM matches in developer toolchain penetration. Watch whether legacy modernization capability becomes a feature of the tools developers already use rather than a specialized product category.
See Anthropic’s How AI helps break the cost barrier to COBOL modernization and IBM’s Lost in Translation: What the AI code debate keeps getting wrong response posts for more information.
Disclosure: Futurum is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.
Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of Futurum as a whole.
Other Insights from Futurum:
Futurum’s 2026 Key Issues & Predictions Report
AWS’s Deploy-to-AWS Plugin: Frictionless Deployment or Developer Honeypot?
Claude Found 500 Zero-Days. Who Patches Them Before Attackers Arrive?
Transparency Through Observability is Essential to AI Success
Author Information
Mitch Ashley is VP and Practice Lead of Software Lifecycle Engineering for The Futurum Group. Mitch has over 30+ years of experience as an entrepreneur, industry analyst, product development, and IT leader, with expertise in software engineering, cybersecurity, DevOps, DevSecOps, cloud, and AI. As an entrepreneur, CTO, CIO, and head of engineering, Mitch led the creation of award-winning cybersecurity products utilized in the private and public sectors, including the U.S. Department of Defense and all military branches. Mitch also led managed PKI services for broadband, Wi-Fi, IoT, energy management and 5G industries, product certification test labs, an online SaaS (93m transactions annually), and the development of video-on-demand and Internet cable services, and a national broadband network.
Mitch shares his experiences as an analyst, keynote and conference speaker, panelist, host, moderator, and expert interviewer discussing CIO/CTO leadership, product and software development, DevOps, DevSecOps, containerization, container orchestration, AI/ML/GenAI, platform engineering, SRE, and cybersecurity. He publishes his research on futurumgroup.com and TechstrongResearch.com/resources. He hosts multiple award-winning video and podcast series, including DevOps Unbound, CISO Talk, and Techstrong Gang.
