AI promises to make operational risk faster, cheaper and more dynamic, but technology alone will not deliver transformation. Risk leaders must redesign governance, develop judgment-led skills and prepare for continuous decision-making while ensuring automation does not remove the experience needed to create tomorrow’s senior risk professionals
Financial institutions must stop
treating artificial intelligence simply as another efficiency tool and start
preparing their governance, operating models and people for a fundamentally
different form of risk management.
The technology could enable
continuous, data-driven risk analysis, but existing structures are unlikely to
cope with the speed at which that information is generated.
Speaking at CeFPro’s recent Risk
Evolve conference, a senior board consultant at an operational risk industry
organization said attitudes had changed markedly in six months.
Experiments are moving closer to
production as institutions recognize two fundamental effects of AI - cheaper
access to data and cheaper software development.
For risk teams, that could finally
unlock information that has historically been too expensive or difficult to use
because it was fragmented across structured and unstructured sources.
The senior consultant said a broadly
consistent vision was emerging among risk leaders.
“If you were to articulate a North
Star for what risk management looked like in five years’ time,” he said, it
would be “data-driven,” continuous, dynamic and automated.
Institutions are initially targeting
efficiency, including automating labor-intensive regulatory processes, before
progressing toward enhanced capabilities such as richer scenario analysis using
more timely and comprehensive data.
The next stage could involve entirely
new risk applications rather than simply performing existing activities faster
or better.
However, continuous analysis creates
a governance problem.
“If you’re generating risk analysis
24-7 on a continuous basis, how does that fit with your current governance
structure?” the consultant asked. “I will answer for every institution in this
room, it just doesn’t.”
Technology transformation must
consequently be accompanied by changes to culture, skills and operating models.
A senior compliance risk and
governance executive at a systemic European bank agreed that institutions need
practical governance built around clearly defined problems and boundaries.
AI combines multiple existing risks
rather than representing a single new category, she argued, meaning controls
should be embedded within existing structures wherever possible.
“Risk should be built in every step
of the way,” the compliance executive said. Risk and compliance teams should
also educate employees deploying AI so they understand those risks without
repeatedly creating additional approval stages.
A senior cybersecurity industry
leader presented the challenge more starkly, arguing that “the biggest risk in
AI is not adopting it.”
Simply using AI to reduce costs will
not be sufficient, he warned. Competitors able to use the same resources to
innovate dramatically faster could transform banking products, internal
workflows and customer interactions.
Smaller organizations are already
demonstrating an ability to turn ideas into functioning prototypes at
unprecedented speed.
That acceleration places new demands
on operational risk.
The senior board consultant argued
that financial institutions are deliberately assuming greater non-financial
risk because doing so enables them to move faster, integrate technology and
data, launch projects and pursue growth.
Risk functions therefore need to
recognize that they are helping organizations make strategic trade-offs rather
than merely responding to a more dangerous environment.
He identified “speed and scale” as
the two defining challenges.
Financial institutions are relatively
well prepared for scale after years spent standardizing, simplifying and
automating processes. Speed is more problematic.
Traditional risk approaches work
effectively in stable environments with established data, expertise and
repeatable controls. Emerging risks operate differently, with limited
information and potentially unreliable historical evidence.
“I’m pretty confident we’ll get there
on scale and I’m pretty worried that speed’s a much harder problem,” he said.
AI could also transform the skills
risk teams require. Routine activities are increasingly susceptible to
automation, increasing the value of creativity, judgment, context and
decision-making.
The consultant compared future risk
professionals with “hunters” and “farmers,” arguing that many repetitive
farming activities will ultimately be performed by machines.