The evolving economics of model risk management in financial services
As model inventories expand and artificial intelligence becomes more embedded, the cost of model risk management (MRM) is coming under increased scrutiny. While many organisations continue to view MRM primarily through the lens of headcount, this perspective is becoming increasingly limited as the nature of model risk evolves.
In practice, the economics of MRM are shaped by a broader set of factors, including policy scope, operating models, governance maturity, and operational efficiency. At the same time, AI and Gen AI is beginning to reshape both the volume and complexity of models, while also introducing new opportunities for efficiency. Together, these developments are prompting firms to rethink how MRM functions are structured and funded.
Rethinking cost beyond headcount
Headcount remains the most visible component of MRM cost, which explains why it often dominates internal discussions. However, organisations that focus solely on full-time equivalents risk overlooking the structural drivers that determine workload.
Policy design is often the most significant of these factors. Decisions around what constitutes a model, how models are tiered and how frequently validations are required can materially influence the size of the model inventory and the level of effort required to maintain it. This is closely linked to the risk appetite setting and plays a key role in scoping and should include an assessment of the cost of failure.
As firms broaden model definitions to include novel risks such as climate risk and AI, inventories are expanding and validation demand is increasing.
Governance frameworks also shape cost outcomes. Complex approval processes, fragmented ownership and inconsistent documentation standards can introduce delays and rework across the validation lifecycle. In contrast, clear accountability and streamlined governance can improve throughput without increasing resources.
Operational efficiency plays a similar role with submission quality and stakeholder engagement influencing the level of effort required from validation teams. Where documentation is incomplete or engagement is limited, validations often become iterative and resource-intensive. These inefficiencies can quickly translate into rising costs.
Is the pyramid resourcing model still fit for purpose?
The traditional pyramid-shaped resourcing model has long been a feature of validation functions, with a small number of senior reviewers supported by a layer of mediors and a foundation of junior analysts. This structure has historically aligned with validation activities that were more standardised and documentation-driven.
However, validation work is evolving. Increasing model complexity, the growth of machine learning and heightened regulatory expectations are shifting the balance towards judgement-based assessments. These developments often require deeper expertise, closer stakeholder engagement, and more iterative validation processes.
As a result, some organisations are reconsidering whether the traditional pyramid remains optimal. A greater reliance on experienced practitioners may improve validation quality and reduce rework, even if it alters the apparent cost structure. In certain cases, flatter team structures or hybrid models are emerging, particularly where specialised skills are required.
This shift does not necessarily imply that validation functions will become uniformly senior-heavy. Instead, it reflects the need for more flexible resourcing models that align with the complexity and diversity of modern model portfolios. Juniors may also be up to date with the latest developments and can bring a different and fresh perspective to the table.
At 4most, we’re increasingly supporting banks on validation engagements and are seeing a shift in how validation functions are resourced. Teams are hiring fewer graduates and increasingly outsourcing validation activities to us.
With IFRS 9 and IRB validation practices now well established, and with greater levels of automation in place, banks are starting to look beyond credit risk. There is growing demand for expertise in newer areas such as generative AI. As these domains are less mature, they rely more heavily on judgement and senior experience.
As a result, the operating model of validation functions should not be considered in isolation. It needs to reflect broader changes, including the expanding scope of models, as well as developments in AI and automation.
Outsourcing offers a way to address these shifts. It allows teams to remain lean, keeps base costs low, and provides the flexibility to scale up during periods of peak demand or access specialist expertise when needed.
Why governance and culture are becoming cost drivers
While policy and resourcing decisions remain central, governance maturity and organisational culture are becoming increasingly important in shaping MRM economics. The way organisations approach model development, documentation and engagement can significantly influence validation effort.
A mature model risk culture often results in better submission quality and clearer documentation, reducing validation cycles and rework. Where model owners engage early with validation teams, issues are typically identified and resolved more efficiently. This dynamic can lower operational friction and improve overall cost efficiency.
Clear escalation paths, well-defined roles, and consistent standards help streamline validation processes. Conversely, fragmented governance structures and unclear ownership can lead to delays and duplicated effort.
As model inventories continue to grow, organisations are increasingly recognising that governance and culture are not only risk management considerations, but also economic ones.
How AI could reshape MRM economics
The growing use of AI introduces both opportunities and challenges for MRM cost structures. In the short term, expanding the model inventory to include AI models is likely to increase workload, particularly as organisations update policies and develop new validation approaches.
For AI, banks need to determine how and what to include in the model inventory and the depth of procedures required. AI models and their use cases often require specialised expertise and more nuanced assessments, placing additional demands on validation teams.
At the same time, AI is beginning to improve efficiency across the validation lifecycle. Tools that support documentation drafting, code review and testing automation have the potential to reduce manual effort and improve consistency. Over time, these capabilities may help offset the increased workload associated with more complex models.
The net impact of AI is therefore likely to evolve. In the near term, organisations may experience rising costs as they adapt policies and build new capabilities. Over the longer term, automation and efficiency gains could reshape cost structures and reduce reliance on manual validation processes. Banks are already taking the first steps in using AI to process and review documents.
What this means for organisations navigating MRM evolution
As MRM continues to evolve, organisations are increasingly moving beyond headcount as the primary measure of cost. Policy design, governance maturity, and operating model choices are becoming central to managing MRM economics effectively.
At the same time, the growing influence of AI is prompting firms to rethink how validation functions operate. While the long-term potential for efficiency is significant, achieving these gains will depend on thoughtful policy design and investment in operating models.
Organisations that take a broader view of MRM economics will be better positioned to manage growing model complexity and regulatory expectations in a cost-efficient way. Those that continue to focus primarily on reducing headcount may actually have higher overall MRM costs as a result. In the end, MRM economics are always a balancing act between the direct cost involved with MRM and the potential cost of model failure.
