Skip to content

Blog

AI modernization works as a delivery problem, not a code translation

Translating COBOL to Java one to one reproduces the old debt. Treated as delivery, with architecture decisions, encoded context and a migration factory, AI-driven modernization works.

Written by

Software Factory

Published: 26 March 2026 Updated: 2 min read

Anthropic’s COBOL blog post shook IBM’s stock price. Thoughtworks just published the missing playbook.

AI-driven modernization works. Just not as a 1:1 code translation.

Direct translation keeps the mess

Direct code translation, COBOL to Java, faithfully reproduces existing architectural constraints, accumulated technical debt and outdated design decisions. You get the same mess, in a different language. Faster.

That’s vibe coding applied to legacy systems.

Understand first, plan second, execute third

The Thoughtworks team lays out a 5-phase approach. Understand first, plan second, execute third. Sound familiar? That’s agentic engineering applied to modernization.

  • Start with architecture decisions. What stays, what goes, what gets redesigned. AI can’t make these calls. Engineers can.
  • Encode those decisions as context. Architecture guidelines, bounded contexts, quality gates, target patterns. Feed them to the agent.
  • Let AI accelerate the repetitive parts. Parsing, pattern detection, boilerplate conversion, test generation. This is where it shines.
  • Build a migration factory. Extract features, generate target code, validate, ship. Same three steps on repeat. Scales with compute.
  • Keep SMEs in the loop. Every mainframe modernization program requires strong domain knowledge. The agent handles volume, humans handle judgment.

Context makes the output reliable

In our experience, the pattern holds everywhere. Teams that prompt without context get plausible garbage. Teams that engineer context (architecture guidelines, quality standards, bounded contexts) get reliable output at scale.

One migration team we worked with went from a 2-year manual timeline to a factory that runs on repeat. The difference wasn’t a better model. It was a better factory.

Why this is good news

Modernization isn’t impossible with AI. It’s achievable. But it requires what good software engineering has always required: understanding the system before you change it, encoding that understanding, and building repeatable processes around it.

AI makes the coding part much cheaper. The engineering part (understanding what to keep, what to discard and what to redesign) is where the value lives. And that’s exactly what experienced teams are built for.

The teams that combine deep domain knowledge with agentic engineering modernize faster, and they modernize better.

Are you translating code 1:1? Or engineering the context that makes modernization actually work?

Further reading

  1. Strong automated gates separate a dark factory from a bullshit factory Above a certain complexity, human code review lowers quality at agent speed. Quality inversion, the automation ratchet, and what a software dark factory needs instead of approvals. 1 min readShort post
  2. Every company will build its own software factory The question shifts from which editor a team uses to what its factory looks like. Four stages lead from autocomplete to a factory that improves itself, and no team can buy its factory off the shelf. 2 min readShort post
  3. The agent shipped a story while the developer slept In a pilot, an agent implemented a complete user story overnight. When work no longer needs a person awake, sprint planning, team sizing and the developer's job have to be rethought. 2 min readShort post

All 23 posts on Software Factory

ClientsClient work

Test the constraint in your system

The Factory Readiness Assessment connects this research to your value stream, your data, and the next decision.

Training on this topic: Agentic Software Factory: Advanced Team Training
See the assessment

Want to check one team first? The self-check: twelve yes or no questions, three minutes, instant result, no account. Take the self-check