AI-powered trading strategies: how Capital Markets are balancing efficiency and disruption

Artificial intelligence (AI) is reshaping industries worldwide—and capital markets are no exception. For investment banks and asset managers, the focus is twofold: harnessing AI for near-term efficiency gains while assessing its potential for deeper, structural disruption.

A recent Risk.net survey conducted in partnership with Murex reveals two distinct adoption dynamics:

1. Rapid integration of Generative AI (GenAI)—leveraging large language models (LLMs) and process automation tools for immediate productivity improvements.
2. Slower emergence of transformative AI applications—particularly in trading, portfolio optimisation, and risk management.

Generative AI: Delivering Immediate Operational Impact

Machine learning (ML) is already well established in finance, particularly in anomaly detection and fraud prevention. GenAI takes operational efficiency further by accelerating complex, human-intensive tasks.

Practical applications include:

• Automated document summarisation—processing financial filings, legal contracts, and regulatory updates in seconds.
• Complex derivatives term-sheet extraction—speeding up reconciliation and reducing operational risk.
• Personalised client communications—enabling tailored, data-driven marketing content.

While these uses are incremental rather than revolutionary, they offer measurable productivity gains—freeing up skilled resources for higher-value activities.

Machine learning for derivatives pricing: speed meets accuracy

One of the most promising AI applications in the front office is replicating complex derivatives pricing models using ML. Over the past decade, financial institutions have faced exponential growth in computational demands—driven by more complex quantitative models, regulatory stress testing, and centralised valuation adjustment management.

Murex’s research shows that well-designed neural network architectures can replicate pricing models with high accuracy at unprecedented speed. This enables:

• Use of sophisticated models in high-volume evaluation contexts (e.g., potential future exposure in credit risk).
• Significant cost savings in risk system infrastructure.

Survey insight: 65% of respondents are “very positive” or “cautiously optimistic” about ML for derivatives pricing.

Murex’s white paper on derivatives pricing with neural networks provides a practical framework for building architectures that capture product specifics and model dynamics—ensuring robust, regulator-ready adoption.

Deep hedging: potential game-changer, but not without hurdles

Deep hedging—first described in 2018—uses deep learning and reinforcement learning to optimise portfolio hedging strategies. Unlike traditional methods based on Greeks, deep hedging incorporates market frictions (liquidity, transaction costs) and blends model-driven with real-world assumptions.

Survey insight:

• ~33% of respondents see it as “significantly” or “moderately” disruptive.
• 31% believe practical and regulatory barriers could limit adoption.

Key challenges:

• Industry inertia—Decades of reliance on established models and validation frameworks make cultural and structural change difficult.
• Technical constraints—Real-time hedging requires frequent recalibration; deep neural networks demand significant computational resources.

Murex’s proof-of-concept projects suggest selective adoption is the most pragmatic path forward. High-value candidates include exotic products like autocallables that require complex stochastic modelling. Here, deep hedging can complement traditional methods—assisting traders without dismantling existing frameworks.

A Pragmatic roadmap for AI in Capital Markets

The integration of AI into capital markets is a balancing act—between immediate operational wins and the slower, riskier path of structural transformation.

Murex recommends:

1. Iterative adoption—Start with contained, high-value use cases to minimise disruption.
2. Close collaboration between business and technology teams—Ensure models address real trading and risk needs.
3. Robust validation and governance—Meet the stringent regulatory and operational demands of the industry.

As AI capabilities mature, its success in capital markets will depend on its measurable business value, its ability to integrate with existing financial frameworks, and the trust it builds among traders, risk managers, and regulators