Model risk is no longer confined to specialist teams, it is now embedded across the entire organization. Briony Pentecost explains how AI and advanced analytics have blurred traditional definitions of models, requiring a shift toward enterprise-wide gover
Model
risk management is undergoing a fundamental shift as financial institutions
confront the realities of AI adoption and increasingly complex decision-making
tools.
What
was once a clearly defined function governed by technical definitions and
regulatory frameworks is now evolving into a broader, enterprise-wide
discipline that cuts across every business line and risk category.
According
to Briony Pentecost, the traditional approach to model risk was rooted in
classification. Tools were either designated as models and subjected to formal
governance, or they were not.
“It
seemed like it should be easy,” she explains, describing early frameworks based
on regulatory definitions. “Does this item meet the definition? Sort it this
way. If not, sort it that way.”
In
practice, however, that binary approach has proven increasingly inadequate.
Pentecost argues that many tools exist on a spectrum rather than fitting neatly
into predefined categories.
“It’s
not so black and white. It’s really more of a spectrum,” she says, noting that
even experienced practitioners can reach different conclusions about the same
tool used in the same way.
This
ambiguity reflects a deeper issue within the industry. By focusing on whether
something meets a technical definition of a model, institutions risk missing
the broader objective of risk management.
“It
really misses the spirit of the guidance,” Pentecost said. “What kind of risk
does this pose? Does this warrant model risk or any other sort of oversight?”
That
shift in perspective is being accelerated by the rise of AI and advanced
analytics, which are introducing new forms of risk that extend beyond
traditional model error.
Pentecost
highlights the growing importance of information security, bias, and fair
banking considerations, all of which now converge with model risk in ways that
were not previously as prominent.
“It’s
really not fair to put it on one group,” she says. “It’s more interdisciplinary
than it’s ever been.”
As
a result, model risk management is no longer confined to specialized teams.
Instead, it requires collaboration across multiple functions, including
technology, compliance, and business units.
This
creates both opportunities and challenges. On one hand, it enables a more
holistic understanding of risk. On the other, it introduces complexity in
coordination and governance.
Pentecost
emphasized that many organizations are still grappling with how to embed this
more distributed approach.
