When AI Stops: Operational Resilience in an AI world

When AI stops

At its most basic level, AI is no different to many other technologies. It depends on hardware, software, infrastructure, data and third parties to function, and shares many of the same vulnerabilities to disruptive risks. Many of the foundational concepts of operational resilience apply directly to AI-enabled services. The core principles: building capabilities to anticipate disruption, absorb shock, and adapt to mitigate impacts remain relevant.

But there are significant differences from traditional technology.  AI capabilities are more exposed to rapidly changing risks, including sovereignty, regulation and capacity of infrastructure.

Within organisations, AI presents unique failure modes such as model drift and hallucinations, which can silently degrade the service without visible disruption.

For high-profile use cases, the cost of AI-related disruption may extend beyond financial impacts. Public sentiment is still evolving, and a small amount of disruption has the potential to impact trust and adoption.

Leaders must recognise that disruption is going to occur, and plan accordingly.

This article examines the unique risks which AI presents for operational resilience, and steps which should be taken to minimise the impacts of disruption.