
When resilience means planning for the failure of others
Jon Eilbeck,
Chief Operating Officer
1
AI is changing the resilience landscape by making vulnerabilities across third-party services and digital supply chains easier to detect and exploit.
2
Traditional backup arrangements are no longer enough when multiple organisations depend on the same platforms, software and infrastructure.
3
Technology resilience now depends on how well a business can operate when critical dependencies fail beyond its direct control.
For many years, technological resilience has focused on physical assets. If one site failed, another could take over. In the event of a fire, flood or power outage, the business could continue to function.
Today, the challenge is far more complex. Organisations depend on an intricate web of external providers for cloud infrastructure, software, data and operational services. At the same time, AI is making weaknesses across that network easier to identify and, potentially, easier to exploit. Resilience means understanding what happens when a critical dependency outside your control breaks down.
Why resilience has become a harder question

One of the defining features of the modern technology landscape is the concentration of essential services in the hands of a relatively small number of providers. Businesses may believe they have diversified their risks, yet still find themselves dependent on the same underlying infrastructure, software vendors or data suppliers as many others.
Cybersecurity is part of the picture, but a major provider does not need to be hacked to trigger serious disruption. In a highly connected environment, incidents can rapidly leave affected organisations with few immediate options. AI may increase the speed at which weaknesses are uncovered, but the broader challenge is structural, as businesses are increasingly reliant on systems and services they do not fully control.
This is why resilience planning has become more layered. It is no longer enough to ask, “What if one data centre fails?” Leaders now have to ask, “What if a critical supplier is unavailable for days or weeks?”, “What if our primary and backup arrangements depend on the same provider?” and “What can we still do if a core external service becomes unavailable?”
These are harder questions, and they now sit at the heart of operational resilience.
The hidden risks of shared dependency
As reliance on third parties increases, resilience depends on understanding which dependencies are critical. That means looking beyond direct suppliers to the wider chain of services beneath them.
A firm may know which company provides its software or cloud hosting.
It may be less clear which other vendors support that provider, where data is stored, how recovery would work in practice or how easily the service could be replaced. In some cases, what appears to be diversification on the surface turns out to be concentration underneath. AI may help organisations map these interconnections more effectively, but it also reveals just how complex and opaque some of these dependencies have become.
This means dependency mapping has become important, as organisations need to know not only who their suppliers are, but which business services rely on them, what would happen if they failed and how long disruption could be tolerated.
Mapping what matters
Resilience means understanding what happends when a critical dependency outside your control breaks.

One way to improve resilience is by preparing a “digital lifeboat” – a separate resource of the minimum tools, data and processes needed to keep the organisation functioning if there is a significant disruption.
For a bank, asset manager or market infrastructure provider, that might mean maintaining access to recent positions; the ability to monitor exposures, communicate with clients and counterparties; or the systems to carry out a narrow set of essential transactions. For another kind of business, it might mean being able to pay employees, serve customers or protect sensitive records.
The aim is to identify what absolutely must continue, preserve the capabilities needed to support it and plan for a period of degraded but controlled operation.
Preparing for the worst: a digital lifeboat
AI is already improving the ability to scan code, analyse systems and identify vulnerabilities that might previously have gone unnoticed. In the wrong hands, the same capability can help uncover weak points across supply chains more quickly and at greater scale.
At the same time, many AI tools and services are themselves provided by third parties, adding another layer of dependency to already complex technology estates. Businesses may benefit from the speed and insight these tools offer, while also becoming even more reliant on platforms and providers they do not fully control.
AI therefore cuts both ways. It can strengthen visibility and accelerate response, but it can also increase the speed at which weaknesses are identified and deepen reliance on external providers. In that sense, AI is intensifying the resilience challenge.
Where AI fits into the picture
As reliance on third parties increases, resilience depends on understanding which dependencies are critical.
The old model of resilience was built around recovery from rare events, while the new model is much closer to continuous adaptation. Organisations need to test for supplier outages, corrupted updates, dependency failures and disruption that affects multiple parties at once.
They need to challenge assumptions about substitutability, question whether backup arrangements are truly independent and practise operating with less than full functionality. They need to understand which services can be restored quickly, which cannot, and what workarounds exist if recovery takes longer than expected.
The hard truth for leaders is that resilience now begins with accepting that some important failures will happen beyond the organisation’s direct control. AI is making the technology landscape more transparent, but also more exposed, by reducing the time and effort required to identify vulnerabilities across interconnected systems.
The question is no longer only whether systems can be recovered. It is whether the business can continue to function when a critical provider, platform or shared service fails.
The next phase of resilience will be continuous
Jon Eilbeck is Chief Operating Officer at BNY Investments.