Mark Carroll, Founder – Graduate ERM Program at Boston University, warns that hidden cloud dependencies and inadequate supplier testing are exposing weaknesses in third-party resilience. He explains why financial institutions must strengthen dependency mapping, demand evidence of recovery capabilities and prepare for cloud outages rather than assuming infrastructure diversification guarantees operational continuity.
Financial institutions risk
overestimating their operational resilience by overlooking hidden cloud
dependencies and relying on supplier assurance measures that provide little
evidence of how critical services would perform during a major disruption.
Mark Carroll, Founder of the Graduate
ERM Program at Boston University, warns that growing dependence on cloud
infrastructure and artificial intelligence requires organizations to rethink
how they identify critical suppliers, assess recovery capabilities and prepare
for outages.
His central argument is that
established third-party risk management principles remain relevant, but
organizations must extend their application beyond immediate contractual
relationships to encompass the underlying technology infrastructure supporting
essential services.
One of the most significant
challenges is identifying fourth-party dependencies - suppliers and technology
providers operating behind an institution's direct vendors.
Carroll recommends the "Use of
Service Level Agreements (SLAs) as well as the identification of single and
sole source vendors being used by a direct vendor."
He also advocates examining the
results of supplier exercises involving data dependencies, helping
organizations identify vulnerabilities that conventional vendor assessments may
overlook.
However, mapping dependencies
represents only part of the challenge. Carroll argues that financial
institutions need stronger evidence that suppliers can actually withstand and
recover from disruption.
Certifications and audit reports may
establish that appropriate processes exist, but they cannot necessarily
demonstrate whether those arrangements will work during a genuine operational
crisis.
Instead, Carroll recommends
"Access to testing results from recovery tabletops, full scale assessments
and actual crisis events as well as vendor performance data."
He identifies delivery performance,
completion rates, inventory management and production practices as additional
indicators that can help organizations evaluate the operational resilience of
critical suppliers.
The problem becomes more complicated
when institutions depend heavily on a small number of major cloud providers.
Diversifying infrastructure across
multiple suppliers may appear an obvious solution, but Carroll questions
whether such arrangements are commercially or operationally realistic for many
organizations.
Where diversification is impractical,
he argues that firms should prioritize maintaining access to essential
operational information when cloud services become unavailable.
"Protect yourself with good
solid data management," Carroll advises.
He recommends generating regular
reports covering inventory balances, production status and financial
conditions, providing a fallback source of information when institutions lose
access to their normal systems.
Such arrangements could enable
organizations to retain visibility over critical activities even when a
technology outage leaves them effectively operating blind.
The increasing integration of AI into
cloud-hosted services adds another layer of complexity, particularly where
organizations cannot readily identify the infrastructure supporting automated
processes.
Nevertheless, Carroll rejects the
suggestion that these developments necessarily require fundamentally new
operational risk management frameworks.
"Operational Risk frameworks
will not change but the scope of these frameworks will evolve," he says.
He argues that the complexity of
cloud-hosted technologies has allowed some dependencies to escape the scrutiny
traditionally applied to single-source or sole-source suppliers.