Artificial intelligence (AI)-driven investment strategies could help pension schemes improve risk-adjusted returns without increasing portfolio risk, according to SmartWealth Asset Management AG CEO, Dr Miro Mitev.

Speaking to Pensions Age, Mitev argued that systematic, AI-based portfolio construction could reduce human bias in investment decision-making while improving the efficiency of portfolio optimisation.

“At the end, what success means is actually the performance for a client,” he stated.

“If you implement these kinds of tools correctly and properly, and you are following the rules without any kind of human interventions, then the results are obvious - very good.”

Mitev explained that his work with AI in financial markets dated back to the late 1990s, when he researched the use of neural networks to forecast stock returns during his postgraduate studies.

“I started this in 2000… my master's thesis was about the application of artificial neural networks for forecasting of stock returns,” he said.

“At that stage, neural networks were showing quite better results compared to traditional regression models.”

He then established SmartWealth in 2016, with the asset manager launching two Irish-domiciled alternative investment funds last year that used a fully automated AI investment process.

According to Mitev, the strategy combines machine learning models, neural networks and genetic algorithms to analyse financial data, forecast returns and construct portfolios.

The system first selects an investment universe using algorithms, then generates forecasts of asset returns and optimises portfolio allocations.

Trades are then executed electronically, with human involvement largely limited to oversight and monitoring.

“The funds are based 100 per cent on a fully AI-driven investing process,” Mitev said.

“The human element is just overseeing the entire process… There is no human intervention in the sense of overruling or changing results.”

He argued that removing human decision-making from portfolio construction can help reduce behavioural biases and improve the speed and accuracy of portfolio optimisation.

“AI can analyse thousands of potential portfolio combinations and optimise allocations for the highest expected return for a given level of risk. This is something that a human cannot calculate themselves.”

Mitev also suggested that automated execution can reduce implementation delays in portfolio management.

“With AI, the optimisation and execution can be done in seconds,” he stated.

“Normally, if you try to optimise such a complex portfolio manually, you might need one, two or three hours.”

Interest in AI across financial services has grown rapidly in recent years, with asset managers increasingly exploring the use of machine learning and data analytics in investment processes.

However, while AI tools are widely used in areas such as quantitative trading and data analysis, fully automated AI-driven investment strategies remain relatively uncommon in institutional portfolios.

Within the pensions industry, adoption of AI has so far been more prominent in operational areas such as administration, fraud detection and member communications.

On the investment side, schemes have generally taken a more cautious approach due to governance requirements and concerns around transparency and model risk.

Meanwhile, industry observers have also warned that AI-driven investment systems can raise questions around explainability, particularly when trustees must demonstrate that investment decisions are well understood and consistent with fiduciary duties.

Nonetheless, Mitev believed systematic AI-driven strategies could play a growing role in institutional portfolios as investors seek to improve risk-adjusted returns.

“Institutional investors are increasingly looking for ways to optimise the risk-return balance of their portfolios,” he stressed.

“If you can increase the return for the same level of risk - or reduce volatility and drawdowns while maintaining returns - that is something many pension funds are interested in.”

With this in mind, Mitev suggested that smaller and more flexible pension schemes may be among the earliest adopters of AI-driven investment approaches.

He noted that large pension funds often relied heavily on passive investment strategies due to regulatory frameworks and liquidity considerations, which can limit the use of more active or systematic investment models.

However, Mitev argued that AI-based portfolio optimisation could help schemes improve returns while continuing to invest in liquid assets.

“You can increase the Sharpe ratio for institutional portfolios… reducing drawdowns, reducing volatility and also increasing the return simultaneously," he said.

“That is something we already see - smaller and more flexible pension funds are starting to go in this direction.”

---