We took a close look at Ionate, the company turning mainframe COBOL and Oracle Forms estates into cloud-native microservices with a guarantee most vendors won't touch: 100% business parity, delivered in months instead of years. Inside: why verified equivalence is the wedge, which cornered CIOs actually buy this, and what the modernization category keeps measuring wrong.
Every few years, the mainframe dies again. It hasn't yet. Somewhere between 60% and 80% of IT budgets at large enterprises still goes to keeping legacy estates alive, and the COBOL workforce keeping them breathing is retiring faster than anyone can replace it. Oracle Forms is end-of-life. IBM licensing bills keep climbing. And the standard answer from the industry has been some version of: brace yourself for a two-to-five-year rewrite, pray the cutover works, and hope the business rules nobody documented survive the trip.
That fear is the actual product category. Not COBOL. Fear.
The wedge: parity, not translation
Ionate's core platform, APPDATE, converts mainframe COBOL, Oracle Forms, AS/400 and similar estates into cloud-native microservices, and the claim they hang everything on is 100% business parity: every rule, every edge case, every process preserved exactly as it was. Not "mostly equivalent." Not "functionally comparable." Guaranteed identical behavior, validated through parallel runs where the old and new systems execute side by side until the outputs match.
That guarantee is the wedge because it's the one thing buyers can't get anywhere else. A lift-and-shift or replatform vendor can promise speed; a manual rewrite can promise a clean architecture; a generic LLM tool can promise cheap code generation. None of them can promise that the migrated payroll system will process the exact same 30-year-old edge case the same way on day one. Ionate backs it with a decade of proprietary AI/ML R&D on what they describe as an institutional dataset of legacy transformation patterns, and claims 99.999% code accuracy across 500M+ lines transformed and 50+ enterprise deployments. The 10 years matter. This is not a wrapper bolted onto a foundation model in 2023; the pattern library predates the LLM boom, and their own positioning takes direct shots at "we'll try some AI tools" as a strategy.
It's hard to copy for the same reason it's hard to believe: the dataset only exists if you've done the migrations. A competitor starting today is a decade behind on training data, and no amount of model cleverness substitutes for having seen how ten thousand undocumented COBOL quirks actually behave.
The ICP they actually win: the cornered CIO
This lands hardest with a specific profile: a Fortune 500-scale CIO or application-modernization lead sitting on an end-of-life platform with a hard deadline and no appetite for a big-bang rewrite. The named customers all fit it. UPS converted JCL batch to Spring Batch. SERPRO, running tax systems for Brazil's Receita Federal, moved ADABAS Natural to Angular and Java microservices at national scale. UNFI took COBOL/DB2/CICS to Angular and Java. These are not experimental workloads. They are the systems where an outage makes the news.
The second buyer is quieter and arguably more strategic: global SIs. Deloitte, Accenture, NTT Data, Kyndryl, EY, Capgemini, Infosys, Atos — 18 partners — use Ionate as a delivery platform, which turns the incumbents' own delivery arms into a channel. Meanwhile MIRA Codex Studio targets smaller projects under 500K lines and developer-led teams, with a SOC 2, air-gapped option for buyers who won't let code leave the building. That matters for banks and government agencies, which is most of their book.
The Hapvida and Davivienda cases show the pattern: Hapvida/GNDI delivered in half the projected timeline with 60% annual cost savings; Davivienda cut time-to-market by 80% and eliminated its Oracle Forms EOL risk. Averages across their case studies run to 82% cost reduction and 3–6 month timelines against an industry norm of 2–5 years.
What the category still gets wrong: optimizing for the wrong unit of work
Here's the contrarian read. The modernization industry overwhelmingly optimizes for lines of code migrated per dollar. That's the wrong number. What actually kills these projects is semantic drift — the thousand tiny behavioral differences between old and new systems that only surface in production, when a discount calculation rounds differently or a compliance rule fires at the wrong threshold. Rule-based tools miss the edge cases. Manual rewrites document them wrong. LLM assistants generate plausible code that quietly changes behavior.
Ionate optimizes for verified behavioral equivalence instead, and prices the engagement on that basis. That's also why their timelines compress so aggressively: the expensive part of migration was never writing the new code. It was the years of parallel testing and defect archaeology needed to prove the new system matches the old one. Automate the proof and the calendar collapses.
The honest caveat: 100% parity guarantees are only as good as the validation behind them, and buyers should pressure-test the parallel-run methodology hard. But that's a diligence question, not a category objection.
The takeaway for operators watching this space: the winner in legacy modernization will not be whoever generates the best code. It will be whoever can prove, mechanically and repeatedly, that nothing changed except the plumbing. If a vendor can't show you how parity gets verified — not claimed, verified — walk.