All videos
0:00 / 0:00
ai

6 in 10 Enterprises Don't Know Why Their AI Is Failing

Eye on AI17 July 2026Watch on YouTube

Description

6 in 10 enterprises can't tell you why their AI workloads are failing. Paul Appleby, CEO of Virtana, explains why finding root cause in an AI factory is fundamentally harder than in traditional enterprise IT, and why that gap matters enormously as companies bet more of their critical operations on AI infrastructure. When an AI workload slows down or stops, the question isn't just "what broke." It's "which layer of a deeply complex, heterogeneous system is responsible", and most enterprises today don't have a way to answer that automatically. That's the real risk hiding inside the AI investment boom. If 6 in 10 enterprises can't diagnose their own AI failures, what does that say about how ready most companies actually are to run AI at production scale? Full conversation with Paul Appleby of Virtana on Eye on A.I.

What you'll learn

  • 6 in 10 enterprises cannot explain why their AI systems fail, creating significant risk for companies deploying AI at production scale.
  • Root cause diagnosis in AI environments is fundamentally harder than traditional IT because AI systems comprise many more complex, heterogeneous layers.
  • Most companies lack automatic diagnostic tools to quickly identify which layer of an AI system is responsible for failures.
  • AI failures require answering not just 'what broke' but 'which layer of the system is responsible', a much more complex challenge than traditional IT troubleshooting.

Frequently asked questions

Why is diagnosing AI failures harder than troubleshooting traditional IT problems?
AI systems are heterogeneous and consist of many more complex layers than traditional IT. When an AI system fails, it is not enough to know what broke, you must determine which specific layer of the system is responsible, which is much harder to automate.
What risk does the lack of diagnostic tools pose for enterprises?
Without automatic diagnostic capabilities, enterprises cannot quickly identify and resolve AI failures. This becomes critical as companies increasingly rely on AI infrastructure for their operations at production scale.
What does the fact that 6 in 10 enterprises cannot diagnose their AI failures reveal about their readiness?
It suggests most companies are not truly ready to run AI at production scale. They are investing in AI without the necessary tools and knowledge to resolve issues when they arise.
What is the underlying problem hidden behind the AI investment boom?
Many companies are investing in AI infrastructure without having sufficient solutions for diagnostics and monitoring. This lack of visibility into complex AI systems is the real risk underlying growing AI investments.

Topics

ai operationsenterprise aiai infrastructurediagnosticsproduction scale