Share with your CIO
KPMG’s 2026 Global Tech Report puts a number on a problem most CIOs already feel: roughly 60% of organizations say AI investment is outpacing their governance capabilities. The result is what KPMG calls “AI dark zombies,” deployed models that drift from their original purpose, go unsupported, accumulate cost, and never get properly retired. One insurance client had to write enterprise architecture rules on the fly during a production rollout, only to watch those rules get overwritten at scale. The full piece lays out the IT maturity capabilities required to prevent it.
What this means for your business
The organizations most exposed here aren’t the ones moving slowly on AI. They’re the ones moving fast on pilots while IT infrastructure stays flat. If your enterprise architecture, asset management, and data governance capabilities haven’t materially improved in the last 18 months, every successful proof-of-concept is quietly building a liability. KPMG’s data that AI leaders hold a 44-point maturity advantage in enterprise architecture over laggards isn’t a benchmark to aspire to; it’s the structural gap explaining why some organizations scale and others accumulate wreckage.
The “perfect prototype problem” is a recognizable failure mode, and the governance deficits it exposes are real. That said, this piece is sponsored content written by KPMG to promote its IT Maturity Assessment product, which shapes the argument toward a formal assessment engagement as the natural first move. The underlying logic still holds, but the framing conveniently skips the possibility that organizations with strong internal EA and product ownership disciplines may already have adequate scaffolding without an external diagnostic. The question worth asking isn’t “do we need an assessment” but “do we actually know what AI is running in our environment right now and who owns each deployment.”
The leading indicator to watch is your help desk’s unknown-system ticket volume. When AI deployments start generating support requests that no team claims ownership of, the dark zombie population has already formed. That’s the moment lifecycle debt becomes visible, and by then the cost of cleanup typically exceeds what governed deployment would have required. The budget conversation shifts from “what do we invest in AI governance” to “what does it cost to audit and retire everything we can’t account for,” and the second number is almost always larger.
Concept deep-dive: AI lifecycle management
AI lifecycle management is the practice of treating a deployed model the way you’d treat any enterprise software asset: tracking ownership, monitoring performance drift, and retiring it when it no longer serves its purpose. Most organizations do this well for traditional software and poorly for AI, because models can degrade silently as data and business conditions change without throwing an obvious error. The business risk isn’t a crashed system; it’s a system still running, still making decisions, and no longer making the right ones.
Based on reporting from Is your AI strategy creating “dark zombies”?, originally published 2026-09-25 10:04:00.
