How to leverage AI in investments: Generali Asset Management's view

By Enrico Scarin, CFA , Head of Portfolio Solutions, ESG & Investment Data Science

What are the main challenges you face in integrating AI into asset management processes?

The fundamental challenge lies in the correct balance between consolidated processes and disruptive innovations.

AI improves operational efficiency, but supporting investment decisions in active management is complex. The industry has not yet fully defined the governance of AI-related investment decisions.

Transparency, ethics, and accountability are crucial themes in this context. For example, transparency requires that algorithms be understandable and verifiable. Ethics implies avoiding biases in AI models that could negatively influence decisions. Accountability means that AI decisions must be traceable and attributable, ensuring that there is always human responsibility behind every choice.

Presenting our results to the financial business, we have placed great emphasis on understanding the initial assumptions, such as calibrating a TEV (tracking error volatility) in line with that desired by our portfolios, or deciding on a realistic portfolio turnover, or, particularly relevant, including transaction costs in the strategy calibrated with the help of portfolio managers themselves, according to the specific experiences and contingencies of their respective reference markets.

How do you measure the success of AI-based initiatives and projects?

The science of business processes requires specific KPIs, and AI is no exception. It is essential to build cost indicators, time management efficiency, customer satisfaction, and market benchmarking to evaluate the cost/benefit ratio.

For example, the internal case of LLM algorithms to produce due diligence reports on investment funds clearly shows the time savings and monetary advantage projected over a year.

When the contribution must instead support a qualitative decision, such as investment decisions, reliable back-tests and a sufficiently long time horizon become crucial. These are the main challenges we face in measuring the results obtained; the opportunities are truly numerous.

How can AI improve risk management in investment portfolios?

Risk management processes, being quantitative and data-driven, can greatly benefit from AI, especially with incomplete time series or to massively process alternative data sources in risk models.

For portfolio construction, we have recently developed a tactical asset allocation strategy that uses machine learning to systematically increase expected returns within a robust risk framework.

This strategy is yielding promising results, with AI supporting short-term tactical allocations in combination with long-term strategies, improving all major risk and risk-adjusted performance indicators.

How do you see your work and the investment area changing in the next 3 years?

Integrating new technologies into the business has always been part of our mission.

The challenge is to manage project timelines consciously. AI advances so rapidly that a 3-year horizon seems too long.

We know that applications in the investment business require project management of a maximum of 6 months, with developments and applications analyzed several times a year.

We believe that the investment value chain, from data science to asset allocation, from stock selection to building reliable back-tests, will be pervaded by AI in every element well before the next three years.