A two-word sentence can trigger a nine-zero decision.
“Translate COBOL.”
When Anthropic explained how Claude Code could read, analyze, and accelerate COBOL modernization work, part of the market “heard” something else: “the mainframe becomes useless.” IBM then saw its sharpest one-day drop in more than 25 years, with roughly $40 billion in market value wiped out. (Reuters)
But that interpretation is often an illusion. And illusions have a superpower in both companies and markets: they spread faster than reality.
Translation isn’t modernization: the difference between the skin and the nervous system
Yes—AI can accelerate work that has historically been “cost-prohibitive”: codebase exploration, dependency mapping, test generation, and conversion to languages like Java or Python. Anthropic says this directly in its COBOL modernization post and related “code modernization” resources. (Anthropic)
But modernizing a living COBOL estate is not just changing syntax:
- it means re-architecting data (often shaped by decades of constraints that code alone can’t reveal),
- dealing with the runtime (CICS/IMS, batch, schedulers, messaging, etc.),
- preserving transactional integrity and correctness guarantees that don’t tolerate approximation,
- maintaining performance, latency, throughput, availability,
- and protecting a deeply hardware/software-coupled stack built over decades.
IBM’s response is explicit: modernization isn’t a language problem, and translation captures almost none of the real complexity of the integrated environment. (IBM Newsroom)
In short: translation touches the surface; modernization touches the living system.
The useful (and dangerous) fantasy: “a tool” that replaces “a system”
Why was the market reaction so violent?
Because our brains love narrative shortcuts:
- an announcement → a simple cause → a total conclusion
- a tool → a rupture → a disappearance
Financial coverage also emphasized this “AI disruption fear” dynamic, while some analysts pushed back by noting COBOL modernization isn’t new—and that mainframe stickiness is structural, not emotional. (Barron’s)
This pattern is common: a real innovation triggers outsized storytelling, then fast decisions that can become hard to undo.
“I think I know”: the real tipping point (my book, chapter 6)
In my book, chapter 6, I explain why these shocks happen: the most dangerous territory isn’t ignorance—it’s implicit assumptions.
The ones we don’t even state, so we never test them.
Here, the implicit assumption sounds like:
“If COBOL can be translated quickly, then the mainframe and its entire ecosystem lose their value.”
That’s not a conclusion. It’s a hypothesis. And hypotheses should be tested—not “priced in” blindly.
What AI truly accelerates… and what it doesn’t replace
What AI can accelerate (and it’s significant):
- application inventory,
- dependency understanding,
- test and documentation generation,
- mechanical conversion of certain modules,
- parts of incremental migration when governance is strong. (Anthropic) (CIO.com)
What it doesn’t replace:
- governance (prioritization, risk, sequencing, accountability),
- target architecture choices,
- compliance, security, auditability,
- operational strategy and resilience trajectory,
- human responsibility when critical systems fail. (IBM Newsroom)
A simple summary: AI compresses effort—but it doesn’t eliminate consequences.
The real takeaway: watch for “trigger-words” inside your organization
Markets moved on a short phrase. Organizations do too.
Single words can trigger disproportionate decisions:
- “automatable”
- “obsolete”
- “replaceable”
- “legacy”
- “cloud-first”
- “AI-first”
- “quick win”
Sometimes they help. But they become dangerous when they substitute for thinking.
The question that matters
👉 In your organization, what was the last “translation” (a word, a phrase, a slogan) that triggered a disproportionate decision—before the implicit assumption was stated and tested?
References
(Reuters) = https://www.reuters.com/business/ibm-posts-steepest-daily-drop-since-2000-after-anthropic-says-ai-can-modernize-2026-02-24/
(Anthropic) = https://claude.com/blog/how-ai-helps-break-cost-barrier-cobol-modernization
(IBM Newsroom) = https://newsroom.ibm.com/blog-lost-in-translation-what-the-ai-code-debate-keeps-getting-wrong
(Barron’s) = https://www.barrons.com/articles/ibm-stock-had-worst-day-in-25-years-ai-disruption-fears-5f632d6c
(CIO.com) = https://www.cio.com/article/4137185/anthropics-claim-that-ai-can-quickly-refactor-cobol-rattles-ibm-investors.html
