Fresche X-Modernize AI review: is it worth the sales call?

Fresche Solutions launched X-Modernize AI on November 4, 2025, calling it the third generation of its RPG and Synon-to-Java conversion technology. Nine months later, there is still no independent review of it anywhere. This one is built entirely from what Fresche publishes, what the tool actually claims to do, and what’s conspicuously missing, starting with a price.

What X-Modernize AI actually is

X-Modernize AI is not a standalone product you install. It’s built on top of X-Analysis, Fresche’s existing forensic analysis engine, which scans an IBM i application and maps its dependencies, business rules, and cross-references before any code gets touched. That distinction matters more than the marketing copy lets on. X-Analysis does the reading. X-Modernize AI does the writing.

Fresche describes the conversion engine as a “custom orchestrated agentic AI model” that is LLM-agnostic, meaning it routes different parts of a conversion job to whichever underlying model handles that task best, rather than running everything through a single model. The pitch is that this produces “handwritten-like” Java rather than a literal, line-by-line transliteration of RPG logic, and that it converts an application and its interdependent programs together instead of one isolated program at a time.

That last point is the real differentiator versus throwing RPG source at a general-purpose coding assistant. A subfile program rarely stands alone. It calls service programs, reads physical and logical files, and shares copybooks with a dozen other objects. Convert it in isolation and you’ve moved the bug, not fixed it. Fresche’s argument is that grounding the conversion in X-Analysis’s dependency map first is what keeps a multi-program conversion from drifting out of sync with itself. There’s no independent test of that claim, but the architecture at least addresses a real, specific failure mode instead of a generic one.

The 50% faster claim: what’s behind it

Fresche’s headline number is that X-Modernize AI converts legacy applications to Java “50% faster than manual rewrite.” Read that phrase carefully. It’s a comparison against a human team rewriting the application by hand, not against a competing AI-assisted tool, and not against Fresche’s own previous-generation conversion product.

No published case study, benchmark, or third-party audit backs the number up as of this review. That doesn’t make it false. Vendor-reported productivity numbers in this space are common and directionally plausible; IBM has made similar claims for Bob’s productivity gains, and those numbers also came with the same caveat of being self-reported. The honest read is that “50% faster than a human rewrite” is a reasonable starting assumption for a scoping conversation, not a number you should put in a budget proposal without your own proof of concept behind it.

Pricing: there isn’t one, and here’s what that means

Search Fresche’s site, the press release, and every syndicated version of the announcement, and you will not find a number. X-Modernize AI is sold exclusively through Modernization-as-a-Service, a subscription structure Fresche describes as “pay-as-you-go” with no runtime license and full access to the converted source code once you have it, but the actual rate depends entirely on your application portfolio’s size and complexity.

This isn’t unusual for enterprise modernization tooling. ARCAD MCP Server and IBM’s watsonx Code Assistant for i share the same pattern of custom, sales-gated pricing. But it does mean you cannot compare X-Modernize AI to a self-serve tool like a Bob subscription on cost alone. You’re not buying software. You’re entering a scoped engagement, and the price only exists once Fresche has looked at your code.

Don’t confuse the free trial with the free proof of concept

Fresche offers two different “try it free” paths, and mixing them up will waste a phone call. X-Analysis has a genuine self-serve free trial, no credit card required, that you can start today. It documents your applications and maps dependencies, and it’s a legitimately useful tool on its own even if you never touch the conversion product.

X-Modernize AI does not have that kind of trial. What it has is a “free, low-risk Proof of Concept,” which is a scoped engagement where Fresche runs the conversion process against a sample of your actual code so you can evaluate the output before committing to anything. That’s a reasonable way to de-risk an enterprise purchase, but it’s a sales process with a Fresche team involved, not a sandbox you log into on a Saturday. If you want to poke at the underlying analysis technology this weekend, start with X-Analysis. If you want to see what your RPG looks like converted to Java, book the POC and expect a conversation first.

Who this is actually for

X-Modernize AI is built for a specific situation: a shop with a real, funded, multi-year mandate to move off RPG or Synon and onto Java, usually because the developers who understand the RPG are getting close to retirement and the business has decided modernization is cheaper than continuing to compete for scarce RPG talent. If that’s your shop, the free POC is worth the meeting. You’ll get a concrete sizing and a real Java sample against your own code, which beats guessing from a vendor deck.

If you’re an IBM i shop that just wants to try AI on your codebase this month, this is the wrong tool to start with. There’s no self-serve entry point, no pricing to evaluate against a budget line, and no reason to start a sales process before you know whether AI-assisted work even fits your team. Point that curiosity at something with a real trial first.

How it stacks up against the DIY and Bob path

Most IBM i shops in 2026 aren’t choosing between X-Modernize AI and nothing. They’re choosing between a full conversion engagement and incrementally using AI tools they already have access to, like Bob or a general coding assistant, to document and gradually refactor code in place.

That second path is slower and doesn’t get you to Java, but it’s cheap to start, has no sales process, and lets a single developer test the water this week instead of scoping a multi-month engagement. It’s also the right first move even if you end up going the Fresche route eventually, because a documented, well-understood codebase is exactly what makes a conversion engagement go smoothly instead of turning into a discovery project. If your team hasn’t captured what your senior RPG developers know yet, that work matters more right now than picking a conversion vendor. Our IBM i AI Field Guide walks through the documentation and knowledge-capture prompts that make that first move concrete.

Bottom line

X-Modernize AI is a credible, well-architected answer to a real problem, built by a company with two decades of modernization experience behind it, but it is not a tool you can evaluate the way you’d evaluate a SaaS subscription. There’s no price, no self-serve trial, and no independent proof yet that the 50% number holds up outside Fresche’s own materials. If your shop has board-level buy-in for a Java conversion and the budget to match, take the free proof of concept and judge the output against your own code, not the marketing page. If you don’t have that mandate yet, spend this quarter documenting what your RPG programs actually do while the people who wrote them are still around to ask, and revisit X-Modernize AI once that groundwork is done.

Frequently Asked Questions

How much does Fresche X-Modernize AI cost?

There is no public price list. Fresche sells it as Modernization-as-a-Service, custom quoted per engagement based on the size and complexity of your application portfolio. The entry point is a free, low-risk proof of concept, not a published rate card.

Is there a free trial for X-Modernize AI?

Not a self-serve one. The free trial you’ll find on Fresche’s site is for X-Analysis, the separate documentation and dependency-mapping tool that X-Modernize AI is built on. X-Modernize AI itself is only available through a sales conversation and a scoped proof of concept.

What languages does X-Modernize AI convert RPG to?

Java is the primary target, with Fresche also supporting Synon-to-Java conversion. The tool is marketed as LLM-agnostic, meaning it selects different underlying language models for different parts of the conversion task rather than running everything through one model.

Is the 50% faster conversion claim verified?

No independent benchmark or published case study confirms it as of this review. It’s Fresche’s own comparison against manual, human-only rewrites of the same code, not against other AI-assisted modernization tools. Treat it as a starting point for your own proof of concept, not a guarantee.