Industry
Capital Markets
Unlocking Exponential Future Growth in Capital Market Firms
Highlights
- Volatility underpins pricing, risk, and regulatory decisions across financial institutions, yet many existing models struggle when markets shift.
- The challenge is no longer ensuring accuracy in calm markets, but stability, transparency, and trust during stress.
- Adopting a hybrid approach that combines proven financial models with controlled AI to build resilient, enterprisegrade volatility systems is the way forward.
On this page
On this pageinpage
The volatility challenge
Implied volatility plays a critical role in today’s financial markets.
The challenge is no longer limited to creating models that work for a single market situation. It’s about ensuring that volatility models remain stable, transparent, and reliable as market conditions change. Traditional models such as Black–Scholes and Stochastic Alpha Beta Rho (SABR) provide clarity but can falter under stress or when data is sparse. Machine learning (ML) models adapt quickly but raise concerns about oversight and trust.
European-style options, a cornerstone of institutional investment strategies and a key source of market insight, offer a clean view of price and volatility relationships, making them critical for pricing, risk, and capital decisions as well as regulatory compliance. For European-style options, volatility surfaces are updated daily and serve as a foundation for decisions across multiple business areas, even when markets are unpredictable.
The challenge is, even in these structured markets, volatility models can become fragile when conditions shift. A model that looks accurate today may behave unpredictably tomorrow, especially for short-dated options and out-of-the-money strikes, where risk is most sensitive. This isn’t just a technical issue. Volatility surfaces feed core processes such as transfer pricing, xVA (value adjustment), risk capital, and client pricing. When calibration becomes unstable, it can trigger unexpected P&L swings, reconciliation problems, and manual overrides. In short, volatility doesn’t just reflect uncertainty, it can amplify it.
Industry experience and research now point toward hybrid solutions—combining the structure of traditional models with the flexibility of ML. The focus is shifting from small accuracy gains to building systems that are resilient, consistent over time, and capable of supporting enterprise-level decisions.
The gaps in traditional approaches
Traditional volatility models such as Black–Scholes and SABR have long been trusted because they provide structure and are easy to explain.
They help traders and risk managers understand market patterns and meet regulatory expectations.
But here’s the challenge: these models assume markets behave in predictable ways. In reality, conditions can change suddenly—liquidity shifts, pricing skews, and volatility spikes materialize without warning. When that occurs, models need frequent recalibration, and each adjustment makes them more fragile over time. This continual adjustment consumes expert time, complicates governance, and erodes trust in the outputs used by business and control functions. Traditional approaches are not wrong; however, they are neither designed nor equipped to handle the combination of volatility, scrutiny, and interconnectedness that capital markets firms now face.
AI for better risk models
ML is increasingly recognized as a valuable tool for financial markets, offering the ability to uncover patterns and adapt to evolving data landscapes.
Its strengths such as rapid calibration, scalability to accommodate new datasets, and the potential for short-term performance gains make it an attractive addition to the risk management toolkit. However, integrating AI into capital markets firms’ volatility frameworks brings its own set of challenges:
Governance and trust: These models are hard to explain to regulators and auditors. They often produce results that don’t align with basic pricing principles, creating compliance risk.
Fragility under stress: ML assumes the future will look like the past. When markets behave unexpectedly—the very moments risk managers care about—these models can fail.
Operational complexity: Traditional models require heavy data preparation, strong controls, and constant monitoring. In fragmented markets, this adds cost and risk.
The above can perhaps be better explained through a real-world example: a leading asset management firm based in North America incorporated AI into its quantitative risk management (QRM) framework. With this shift, the firm realized the following benefits:
- Rapid scaling to accommodate new datasets.
- 7% improvement in alpha due to improved calibration leading to better short-term performance.
- 15% improvement in go-to-market timelines.
While we successfully delivered the project to the client, the engagement highlighted the need for robust enterprise-wide integration and stress testing for sustained alpha growth. It also underscored the importance of legacy infrastructure modernization to overcome latency and throughput challenges. The key learning was that achieving this would demand a hybrid model that combined AI’s strengths with established controls and infrastructure, which can in turn deliver long-term benefits.
Clearly, AI should be viewed as complementary to traditional approaches rather than a replacement. When thoughtfully integrated, it can enhance decision-making and operational efficiency, provided its limitations are acknowledged and managed.
The hybrid paradigm: A practical approach
The most effective path to better volatility management is to combine the strengths of traditional models and ML.
Hybrid volatility systems (see Figure 1) use traditional models to provide structure, transparency, and alignment with pricing logic. ML components are then layered on top to capture residual behaviors that parametric models do not fully explain, reduce noise in sparse or stressed markets, and stabilize surfaces from one day to the next. The result is a volatility surface that retains financial meaning but behaves more coherently through time.
Figure 1: A hybrid volatility model
The foundational layer consists of traditional parametric models, providing structure, transparency, and pricing logic. On top, ML modules are integrated to absorb anomalies, model residual behaviors, and stabilize volatility surfaces. Data flows bi-directionally, with continuous monitoring and feedback ensuring resilience, explainability, and adaptability in stressed market conditions. By adopting such a layered approach, capital markets firms can augment established methods with ML techniques, driving both short-term performance and long-term robustness.
This hybrid approach is particularly powerful during market volatility. The parametric layer anchors implied volatility, and the ML layer reacts to nonlinear shifts, producing a surface that is both stable and responsive.
Treating volatility modelling as an enterprise capability brings clarity from traditional methods and flexibility from adaptive techniques. Prioritizing transparency, consistency, and stress-readiness ensures resilience far beyond marginal accuracy gains. Hybrid architectures deliver this balance without sacrificing rigor.
We recommend a practical starting point:
- Recognize the hidden costs of daily recalibration.
- Pilot hybrid approaches in lower-risk areas like long-dated index options.
- Include stability metrics in model risk frameworks.
By starting with these focused steps, firms can build momentum toward a more resilient and adaptive volatility management framework.
Case-in-point
A global bank’s markets division leveraged advanced semantic models and open-source formats to scale its AI capabilities across the enterprise.
By integrating new AI tools with existing infrastructure, the bank established a robust foundation for AI and GenAI adoption. This integration enabled:
- Enhanced risk analytics: Improved volatility surfaces for more stable pricing and risk metrics.
- Operational efficiency: Seamless legacy integration reduced calibration failures as well as manual intervention by ~20%.
- Greater resilience: Scalable architecture supported rapid response to market shifts and strengthened governance.
A leading Canadian bank’s fintech subsidiary, focused on retail, personal, and business banking, adopted a similar approach to scale AI capabilities across the organization. By building on foundational AI capabilities, robust and scalable data, and integrating ML pipelines with its existing tech stack, the organization achieved:
- Reliable volatility management: AI-driven data cleansing delivered consistent risk and pricing models, even in volatile market conditions.
- Accelerated innovation: Scalable AI solutions sped up the launch of new risk analytics and client offerings.
These successes demonstrate that enterprise AI adoption in capital markets is not just about technology, it’s about building resilient, integrated systems that enhance volatility modeling, risk management, and business agility.
Business impact
Resilient volatility models deliver measurable business benefits across the organization.
The effects are felt in functions such as trading, risk, and finance:
- Fewer disruptions: More stable models mean fewer calibration failures and less manual intervention, improving trading efficiency.
- Clearer P&L drivers: Volatility moves for market reasons rather than model instability make performance easier to explain and attribute.
- Smoother pricing and risk metrics: Reduced volatility noise leads to more consistent valuations and exposure calculations.
- Improved governance and regulatory confidence: Traditional models provide interpretability, while ML is applied in controlled, well-defined roles.
- Better monitoring and early warning: Historical archives and stability metrics help detect drift before it becomes a problem.
These benefits are especially critical in European-style options, where volatility and stability directly impact pricing, risk management, and client confidence.
Toward future resilience
Volatility systems now sit at the heart of decisions on pricing, capital, and client confidence.
The question is no longer limited to identifying the best performing model but about keeping systems stable and explainable when markets are most uncertain given future technology evolutions, changing monetary policies, and geo-political headwinds.
Rapid, non-linear shifts are making financial markets increasingly complex and unpredictable; in this environment, traditional, static, and rule-based models will fail to deliver accurate, real-time predictions. The need of the hour is to build models equipped with the ability to adapt to perpetually changing financial markets and the resultant volatility. In our view, the hybrid model comes with the resilience to manage volatility as it combines the structure of proven traditional models with the adaptability of modern techniques, ensuring decisions remain credible even in volatile conditions.
The time for action is now. And capital markets firms that make this shift at the earliest will gain not just operational efficiency, but strategic advantage as they navigate a complex environment, marked by global geo-political tensions, shifting policies, and AI-driven transformations, that demands robust decisions.