Leveraging controlled innovation — the second core element of robust Model Risk Management

Model risk has ceased to be attributed solely to the underlying mathematics. Rapid developments in regulation have driven a broader model landscape which ranges from data viability through to use and implementation systems. As such, Model Risk Management innovation is required to keep pace with competitors and robustly identify, assess and control model risk.

In this new environment, innovation is no longer just a nice to have. It’s integral to long-term, robust Model Risk Management (MRM).

Innovation can come in many forms. It can be expanding model risk frameworks and validation standards to include new model types such as fraud and compliance. It can be learning to use the full scope of data available in a structured format. And it can even be the heightened use of technology through process automation, generative technologies or cloud-based technology platforms.

What is clear is that innovation can be a large-scale programme or a series of small changes. But, in all cases, you need to ensure that innovation is controlled.

We’ve already spoken about the three core elements every bank needs to get right if they want to robustly manage model risk, including establishing a positive model risk culture. In this blog, we’ll discuss the second of those elements: keeping pace with the changing risk landscape through controlled innovation.

Or, if you’re ready for the next step, you can read our in-depth blog on improving MRM efficiency.

And remember, for more in-depth insights, you can always download our specialist Model Risk Management Guide. Learn more about the guide from this link, or read it straight away by clicking here.

Why Model Risk Management Innovation matters

Model risk is a risk in its own right, reported to board and with dedicated Senior Management Responsibility. But this hasn’t always been the case, and regulatory requirements have increased significantly across all areas of the model risk lifecycle.

For example, validations require increased scope; model risk reporting requires greater depth, and deterministic quantitative and model risk mitigants now fall clearly in scope. The increasing demand for resources is surpassing the effectiveness and relevance of traditional products, processes and controls. Model Risk, as with any risk, requires a holistic view and a traceable, single source of truth.

That leaves organisations with a choice: manage the risks of every model with no process or infrastructure changes, or innovate.

Following the current path might appear consistent and have the potential benefit of being embedded and understood, but over time it’s likely to exponentially increase risk through manual workarounds (for example, aggregating model data required for board reporting from disconnected solutions). It’s easy to see where hidden costs and large regulatory fines can transpire if organisations fail to innovate and keep pace with the changing risk landscape.

Why innovation must be controlled innovation

It’s important to create an environment that supports and embraces controlled innovation across risk and the wider organisation. This will require simple policies which are easily understood and relatable, with frameworks defining a greater breadth of dynamic controls. Remember, the link between innovation and MRM must be transparent and bought into by everyone in the organisation — a point we highlighted in our previous article on driving a positive model risk culture. 

As the Bank of England recently acknowledged, it’s imperative that innovation is controlled, and not just a reactive response. Decisions need to be suitably justified, incorporating alternatives and scenario planning. Models and use cases need to have defined success criteria, be logged within inventories, and have controls established and proportionate validations undertaken.

Most importantly, the model risk needs to be explained and understood by the board. Control is important to maintain trust (both from senior management and the regulator) and correct prioritisation — which ultimately helps manage costs and improve business efficiency.

In the spotlight: generative AI

First of all, it’s important to note that generative AI isn’t the only form of innovation. Innovation can be as simple as the expansion of frameworks and controls, improving embedded model inventories, or creating a common data key between isolated systems. There is also the potential for technology-based change such as machine learning or digitalisation and business automation. 

All of this means use case selection is paramount. So, before investing in a new generative AI solution, ask yourself: do we actually need this? Do we want this added complexity, and what’s the benefit of doing so? Do we have the capabilities to create or adapt risk management frameworks, implement controls and validate the solution?

The benefits of leveraging generative AI within Model Risk Management

The benefits from generative AI (again acknowledged by the Bank of England) include process optimisation and productivity. To put these into perspective, here are three model risk specific examples to consider:

• Model risk understanding. The core output of a validation is a concise report, which leverages a detailed methodology document and implementation specification. Large Language Models (LLM) within an AI system can ingest documents and interrogate the resulting data to identify trends or inconsistencies across different model types and explore unknown model linkages and frequent limitations. This would allow a deeper understanding of model risk.
• Assisted technical commentary. To help draft diverse technical descriptions of the modelling process and results of validation or monitoring figures, historic commentary can be used alongside model outputs within a generative AI system to provide an empirically based narrative that’s ready for further interpretation. This can be a huge time saver when preparing documentation or reporting.
• Efficient board reporting. Improving the communication of reporting is also a useful benefit of generative AI. Rather than sharing reports as a simple document, you can leverage generative AI systems to provide deeper user interactions – readers can ask questions within the document and get an immediate answer, even to complex queries.

Our advice?

Innovation is only effective when it’s safe; risks are hard to retrospectively control. For instance, you may want to implement an MRM efficiency through the adaptation of a third-party generative technology within the validation process. But, if you aren’t fully aware of all the potential impacts it could have, it can result in unintended and severe model risk consequences.

AI within MRM can provide efficiencies, but it does not abdicate decision maker responsibility for interpretation and communication. Use cases are not exempt from model risk management principles and validation. These will still need to be interpreted, explained and controlled despite their dynamic nature and complexity.

How can 4most support you with Model Risk Management innovation?

We have many years of experience working within banks and regulated businesses across numerous jurisdictions. In that time, we’ve supported clients to implement a broad range of innovations within their businesses — from gap assessments to expanded model risk frameworks and validation standards, and from market-leading machine learning model monitoring through to automation. We understand how to integrate these changes into operations in an effective and compliant way, so that you can innovate in a controlled manner.

One of the key benefits of working with a technical consultancy like 4most is that our experts are constantly exploring new innovations like generative AI and developing their skills within them. We are model risk management experts, industry-acknowledged data scientists and MLOps engineers with persistent hands-on generative AI experience. That’s why we’ve been able to quickly embed different generative AI capabilities across various areas of our business, with a particular focus on MRM.

What’s next?

For more guidance on achieving robust MRM, you can read our in-depth blogs on…

• The three core elements of robust MRM
• Establishing a positive model risk culture
• Improving MRM efficiency

Or you can read our specialist guide…

… and get the latest best practice guidance for MRM. You can learn more about the guide here, or you can read it straight away by clicking this link.

If you’re looking to keep pace with the changing risk landscape through innovation, but want to do so in a controlled, safe, and expert manner, let us know how we can help by completing our short contact form.