AI simulation gives wealth managers a crystal ball

The wealth management industry must negotiate a turning point, with artificial intelligence-driven simulation now offering firms a powerful new way to anticipate client behaviour and respond to change.

For many years, these firms have relied on surveys, sentiment tracking, focus groups and A/B testing to predict how clients might react to new products, pricing or market shifts.

But these traditional methods are slow and biased, as people do not always act as they say they will, says Ugur Hamaloglu, EY Americas Wealth & Asset Management consulting leader.

“The age of guessing what clients want is over,” says Mr Hamaloglu, who previously headed up EY’s tech consulting division. “Firms can now know what their clients will do, before they do it.”

Agentic AI changes the game. Instead of measuring what investors feel today, simulations use synthetic data to model how real clients behave. Whether adjusting fees, launching new products, or targeting younger investors, firms can now forecast likely outcomes. “This isn’t just about understanding what clients are thinking now,” says Mr Hamaloglu. “It’s about anticipating what they may think and do tomorrow.”

Exposing blind spots

EY recently partnered with AI simulation start-up Aaru to test the technology on its annual wealth and asset management survey, normally conducted across more than 30 markets over several months. Using simulation, the same survey was replicated in a single day.

A “digital replica” of the target group was created, mirroring the demographics of traditional respondents. AI-generated responses to more than 50 questions showed a 90 per cent correlation with the original survey. But it is the remaining 10 per cent that proved most revealing.

For instance, while 82 per cent of survey respondents said they would stick with their parents’ adviser after inheriting wealth, the simulation predicted just 43 per cent. Industry studies suggest only 20-30 per cent do.

Likewise, while 69 per cent claimed to prefer a single financial provider, simulation found only 37 per cent did; real-world data shows just 33 per cent of high net worth individuals use only one adviser.

“That 10 per cent is important, as it gives us insight into behavioural bias,” says Mr Hamaloglu. “When you’re in a peer group, you answer in a certain way. But when you’re by yourself, you act differently.”

Simulation, he argues, helps reveal these blind spots, especially under stress or during life transitions.

Simulation equation

Markets move quickly, driven by geopolitics, digital disruption and generational shifts. Traditional research cycles struggle to keep pace, he explains. Simulation, by contrast, allows firms to test hundreds of scenarios at once, from pricing changes to shifts in client behaviour, and do so with unprecedented speed.

“You can pivot, and change your direction, how you’re investing in your future experiences in a matter of days or weeks, instead of months. That’s the big change,” says Mr Hamaloglu. Digital journeys, messaging strategies and service models can now be adapted in real time. “The whole ecosystem will be impacted by that.”

At the heart of simulation are synthetic populations, virtual individuals or ‘agents’, created using real — and where needed — synthetic data.

Machine learning draws on public sources like census data, IMF reports, sales figures and social media sentiment, to identify the traits that matter in each scenario such as age, income or risk appetite. These agents behave in ways that reflect real-world decision-making.

“Wealth managers could work with providers to create synthetic replicas of their client base, allowing them to test scenarios and generate insights,” says Mr Hamaloglu. Context, such as location and culture, is critical in simulation models, to adapt “persona group data” to a bank’s specific client base. A private bank in Hong Kong, for instance, will require a very different data model from a retail bank in Italy.

Still, some remain cautious. “My challenge is making sure the models genuinely capture the experiences of my customers, not just replay outcomes seen elsewhere,” says Matt Handley, London-based CEO Wealth at Swedish bank Handelsbanken.

“I want to test products and services directly with our clients and prospects. I’d need convincing before fully trusting an AI model in isolation.”

He also highlights the importance of trust and transparency in AI, especially in regulated sectors such as wealth management.

“As we extend the use of large language models and generative AI, the biggest challenge we face is ensuring the solutions we give customers are repeatable, fair and in the customers’ best interest,” states Mr Handley.

Sharper insights

For greater accuracy, explains Mr Hamaloglu at EY, firms can integrate internal data, like transaction records, investment preferences and spending patterns with synthetic datasets.

“The type of information you have depends on the industry you’re in. Firms can layer this data on top of synthetic datasets, if regulations allow them to use it, to help fine-tune the results for their specific business.”

Simulation also supports portfolio management by revealing behavioural drivers such as risk tolerance. “Investors often claim they can handle steep losses, only to panic-sell after a modest dip,” says Mr Hamaloglu.

Simulation can test reactions to volatility, interest in private markets, new pricing models, or major life events, such as inheritance, career shifts, as well as tax and regulatory changes.

“The applications are limited only by imagination,” he says.

Although simulation is not new in other industries, its application using synthetic and proprietary data in wealth management has only gained traction “in the past six to eight months,” says Mr Hamaloglu.

Data crater

Still, integrating proprietary data introduces new challenges. While public synthetic data carries “minimal risk”, internal client data demands strong privacy safeguards, robust governance, and stringent oversight.

Change management also matters. Advisers may distrust or resist model outputs that contradict their intuition or experience, whether in how they engage clients or adapt software and digital tools, he warns. Changing mindsets and building trust in the outputs is key.

Another major hurdle is around data consolidation and management, says Mr Hamaloglu. “It might sound simple - they have all the transaction data - but most firms struggle to combine that with synthetic data and run effective simulations. It takes significant effort to consolidate, clean and aggregate the data properly.”

Interestingly, cost is less of a barrier. Off-the-shelf models are affordable, he says; integration with internal data is what drives higher expenditure.

Simulation is fast becoming a differentiator. Mr Hamaloglu predicts the industry is moving towards “continuous intelligence,” where decisions are based on real-time behavioural data rather than backward-looking research. “Within three years, no major strategic decision will be made without it. Organisations that haven’t started by then will be left behind.”

Yet, statistics tell a more measured story today. While 95 per cent of wealth and asset management firms are already deploying at least three GenAI use cases, just 7 per cent of firms currently report using agentic AI, according to EY in early 2025. Adoption is expected to accelerate, however, with nearly 85 per cent anticipating full integration of agentic AI within two years, particularly in areas like risk analysis and customer service.

Across Europe, another EY study conducted last year found that only 9 per cent of financial services firms saw themselves ahead of the curve in AI integration, with almost a third admitting to falling behind.

But technology, on its own, is not enough. “It’s more than a technology question, it’s about human behaviour,” notes Mr Hamaloglu. The well-known gap between what people say and what they do means AI is only useful if firms are asking the right questions.

“It’s the historical knowledge and experience within these firms that will guide them in using these tools effectively, to uncover behavioural biases and apply insights that directly improve client strategies and decision-making.”