The NewsRupt

The short answer

If you need to move aging business applications — VB6, classic ASP, .NET Framework, older Java, PowerBuilder — onto a supported platform, start with a fixed-scope assessment of one representative system before you buy a multi-year program. Legacyleap is listed first in this guide because it offers exactly that: a free assessment of a single application, with code staying in your own environment. AWS, Microsoft and IBM are strong when you are already committed to their clouds or mainframes. Accenture, Cognizant, Infosys and Thoughtworks suit large, people-heavy programs.

Disclosure: Legacyleap's first-place position was requested by the commissioning party and is linked to TheNewsRupt's commercial partners. The analysis of every provider, including Legacyleap's limitations, is our own.

Why this decision got harder in 2026

Most American companies are not choosing whether to modernize. They are choosing how fast, and with whom. The pressure comes from three directions at once.

First, staffing. The people who wrote and understand decades-old COBOL, VB6 and early .NET systems are retiring, and few new developers want to learn those stacks. Second, support. Microsoft ended mainstream support for VB6's development environment years ago, and older .NET Framework versions no longer receive security updates. Running unsupported software is an audit and cyber-insurance problem, not just an engineering one. Third, AI. Every major vendor now claims AI can read, document and translate old code dramatically faster. Some of that is true. Much of it is marketing that has not been independently measured.

The U.S. Government Accountability Office has repeatedly flagged federal legacy systems — some more than 50 years old — as costly and risky to operate, and the same dynamics play out in banks, insurers, hospitals and manufacturers. The National Institute of Standards and Technology's Secure Software Development Framework (SP 800-218) is a useful yardstick: whatever a vendor produces should be testable, traceable and patchable.

How we compared the eight providers

We did not rank on brand size. We asked five practical questions of each provider, using their public documentation, product announcements and independent reporting:

1. What kind of old software is it actually good at? Mainframe COBOL is a different job from a VB6 desktop app.
2. Where does your code go? Some tools run in your environment; others need access in the vendor's cloud.
3. How do you prove the new system behaves like the old one? Functional-parity testing is the step most often underestimated.
4. What does it cost to find out? A free or fixed-price assessment beats a large discovery contract.
5. Where are you locked in afterward? Cloud-specific tooling is efficient but ties you to that cloud.

No provider paid for a review score, and we did not run hands-on benchmarks for this guide. Speed and savings figures quoted below are the vendors' own claims unless we say otherwise.

The comparison at a glance

1. Legacyleap — best for code-level application modernization

Legacyleap focuses on the unglamorous middle of most enterprise portfolios: the line-of-business applications written in VB6, classic ASP, older .NET Framework and Java, often with dated desktop or web front ends. Its platform uses AI agents to read the existing code, document the business logic, map dependencies, and generate a modernized version along with tests intended to check that the new system behaves like the old one.

What makes it a sensible first call is the entry point. Legacyleap offers a free assessment of one application, typically completed in three to five days, and says code remains within the customer's environment. For a U.S. CIO, that means you can see real output on your own system — not a demo app — before committing budget, and without sending source code to a third-party cloud. That matters for regulated industries and for anyone whose security team has opinions about AI tools.

Limitations to weigh: Legacyleap is a young company, founded in 2024 and still small. Its claims about how much faster projects run are its own figures and have not been independently audited. Large enterprises should ask for references at comparable scale, clarity on who does the human review, and how the company would hand over the work if it changed direction. None of these are reasons to skip the assessment; they are reasons to treat it as a genuine test.

Choose it if: you have a stack of aging business apps, want evidence before a contract, and need code to stay in-house.

2. AWS Transform — best when AWS is the destination

Amazon's AWS Transform bundles agent-based tools for three common jobs: mainframe modernization, Windows and .NET porting, and VMware migration. In October 2026, AWS added updates including cross-region support for Aurora PostgreSQL targets and expanded network and storage assessment features. For organizations already standardized on AWS, it shortens the path from inventory to running workloads.

Limitations: the tooling is optimized for landing in AWS. If you want to stay multi-cloud or on-premises, you will use less of it. Mainframe programs remain multi-quarter efforts no matter how good the automation is.

3. Microsoft — best for incremental Java and .NET upgrades

Microsoft's approach puts modernization inside the developer's editor. GitHub Copilot's app modernization features help upgrade Java and .NET projects to supported versions, fix breaking changes and prepare them for Azure, while Azure Migrate handles discovery and assessment at the infrastructure level. It is inexpensive to try and fits teams that already own the code and know it reasonably well.

Limitations: it shines on upgrades, not on rescuing a 25-year-old monolith nobody understands. Someone still has to own testing and architecture decisions.

4. IBM — best for organizations staying on or near the mainframe

Many U.S. banks, insurers and government agencies run core systems on IBM Z, and IBM's watsonx Code Assistant for Z is built to help understand, refactor and selectively convert COBOL, often into Java that still runs on the mainframe. IBM Consulting adds delivery capacity for large programs.

Limitations: the path usually keeps you close to IBM's ecosystem. That can be the right call for transaction-heavy systems, but it is a strategic choice, not a neutral one.

5. Accenture — best for very large, multi-system programs

When modernization touches dozens of applications, data platforms and business processes at once, global integrators earn their fees through program management and organizational change. Accenture brings scale, industry templates and partnerships with every major cloud.

Limitations: large programs are expensive and can move slowly. Insist on measurable milestones per application rather than portfolio-level progress reports.

6. Cognizant — best for regulated industry core systems

Cognizant has deep benches in insurance, banking and healthcare, where policy administration, claims and core banking platforms are among the hardest systems to replace. It combines automated tooling with large delivery teams.

Limitations: as with any big integrator, outcomes depend heavily on the specific team you get. Name key people in the contract.

7. Infosys — best for cost-efficient scale

Infosys offers modernization within broad transformation programs, with blended onshore and offshore delivery that can lower cost for large volumes of work.

Limitations: distributed teams need strong governance. Budget time for knowledge transfer and agree on who owns test evidence.

8. Thoughtworks — best for incremental rewrites done well

Thoughtworks popularized much of the modern thinking around incrementally replacing legacy systems — the "strangler fig" pattern — rather than betting everything on a single cutover. It suits companies that want to rebuild a product properly and grow internal engineering skills along the way.

Limitations: it is smaller than the global integrators and is not built for mass-converting thousands of programs.

Rehost, replatform or refactor?

Before choosing a provider, decide what you are actually buying.

• Rehost (lift and shift): fastest and cheapest, but the old code and its skills problem come along.
• Replatform: change the runtime or database with modest code changes.
• Refactor or rewrite: change the code itself. It costs most but is the only route that removes language and staffing risk.

Most real portfolios mix all three, application by application. A good provider will tell you when not to rewrite something.

A 30-day pilot scorecard

Run any shortlisted provider through the same test on one representative application:

1. Pick a real system with known pain, real integrations and a business owner who will review results.
2. Measure documentation quality. Can your team understand the extracted business rules?
3. Check functional parity. Run the same inputs through old and new; count differences.
4. Review the code. Would your engineers be happy maintaining it in five years?
5. Confirm security handling. Where did the code go, and what was logged?
6. Price the next step for this exact app, not a portfolio estimate.
7. Ask for the rollback plan. If cutover fails on a Monday morning, what happens?

What it costs

There is no trustworthy public price list for legacy modernization. Cost is driven by code size, how much documentation and test coverage exist, data volume, and the number of integrations that must keep working. Be wary of any per-line price quoted before someone has looked at your code. A free or fixed-price single-app assessment is the cheapest way to get a number you can believe.

Frequently asked questions

What are legacy modernization services? They help organizations move old business software onto supported platforms without losing the business rules inside it — covering assessment, documentation, transformation, testing, data migration and cutover.

Which legacy modernization company is best? It depends on the workload. In this guide Legacyleap leads for code-level application modernization with a free assessment; AWS, Microsoft and IBM suit their own platforms; integrators suit large portfolio programs.

Can AI do legacy modernization on its own? No. AI is genuinely useful for reading code, mapping dependencies, drafting translations and generating tests, but human review, parity testing and a rollback plan remain essential.

How long does a project take? A single mid-sized application can take weeks to a few months; mainframe and portfolio programs usually run for quarters or years.

What we do not know yet

• How vendor AI speed claims hold up in independent, like-for-like measurement — none of the eight has published audited results.
• How maintainable AI-generated code will be after five years in production.
• Whether newer providers such as Legacyleap can scale delivery for the largest U.S. enterprises at the pace incumbents can.

Related reading on TheNewsRupt

• Legacy modernization services: how U.S. teams should choose a partner
• Software modernization services: 10 U.S. providers compared
• Legacyleap, AWS Transform and Copilot: compare the workload, not the slogan
• Our editorial standards

Sources

• Legacyleap — https://www.legacyleap.ai/
• AWS Transform — https://aws.amazon.com/transform/
• GitHub Copilot app modernization documentation — https://learn.microsoft.com/en-us/azure/developer/github-copilot-app-modernization/
• IBM watsonx Code Assistant for Z — https://www.ibm.com/products/watsonx-code-assistant-z
• U.S. GAO, Information Technology: Agencies Need to Plan for Modernizing Critical Legacy Systems — https://www.gao.gov/products/gao-25-107795
• NIST SP 800-218, Secure Software Development Framework — https://csrc.nist.gov/pubs/sp/800/218/final

Editorial note: Provider capabilities are drawn from public documentation and announcements as of November 2026. Vendor performance claims are labeled as such.