These figures, however, only partially capture what is changing, because the current phase is defined less by visible output gains than by a shift in how decisions themselves are produced. In the United States, AI-related investment is already contributing to growth at the margin, around 1.1% of GDP in early 2025, even as broader productivity effects remain uneven and difficult to isolate. This suggests that the structure of decision-making is evolving faster than the data used to measure it.
At the same time, intelligence is no longer concentrated within individual institutions but distributed across models, platforms and external providers. Decisions increasingly emerge from the interaction of multiple systems rather than a single internal process, with direct implications for where control sits and how outcomes are understood.
AI is already embedded in portfolios
Artificial intelligence is already present in portfolios, even where it is not explicitly recognised. Over the past two years, a small group of technology companies closely linked to AI development has accounted for a disproportionate share of equity market performance, with the “Magnificent Seven” contributing more than 60% of S&P 500 returns in 2024 alone.
For globally diversified investors, this exposure is not incidental but structural. Market-cap weighting reinforces it, and passive allocation embeds it further, allowing portfolios to appear diversified across sectors and geographies while remaining dependent on a narrow set of technological drivers tied to AI adoption, cloud infrastructure and semiconductor supply chains.
Elevate Your Wealth Game: Empowering UHNWIs for Simplified Asset Management. Altoo Platform Preview
Inside financial institutions, adoption has moved beyond experimentation. Around 65% of U.S. financial institutions already deploy AI in production, and a significant share plan further increases in investment, indicating that what was recently optional is becoming part of the operating baseline.
The speed of adoption outside finance reinforces this shift. ChatGPT reached 100 million users within months and has since approached 700 million weekly active users, signalling a durable change in behaviour rather than a temporary surge in interest.
Language as the new interface of finance
The underlying technology is not new in principle. Systems capable of classification, prediction and optimisation have existed for years, but what has changed is their ability to operate across language, the medium through which most financial work is conducted.
Investment memos, due diligence reports, risk commentary and client communication all rely on language. Once systems can operate fluently within that medium, they begin to participate directly in processes that were previously difficult to automate.
The OECD definition of artificial intelligence highlights this overlap. AI systems infer from data how to produce outputs such as predictions, recommendations or decisions, and finance is built on exactly these outputs. This alignment is structural, which explains why its implications extend beyond any single use case.
The economic shift is measurable, but uneven
The economic impact of artificial intelligence is becoming visible, though not evenly distributed. While there is still debate about how much of recent productivity growth can be attributed to AI, its effects are increasingly evident in cost structures and operational efficiency.
Estimates suggest that AI applications could generate up to USD 4.4 trillion in annual productivity gains, particularly in knowledge-intensive functions. At the same time, projections indicate that North America could see a GDP uplift of around 14.5% by 2030, reflecting early adoption and capital concentration.
At the firm level, the effects are more incremental but still meaningful. Asset managers report time savings of 30 to 60 minutes per employee per day in research, reporting and internal workflows, which, when aggregated, begin to alter cost trajectories and operating leverage.
As these tools become widely available, efficiency gains are absorbed by the system, compressing margins and making excess returns harder to sustain. Access to AI alone does not create advantage; outcomes depend on how it is applied.
Portfolios reflect the shift more than they reveal it
Artificial intelligence is not a separate investment theme. Its effects are embedded in how portfolios behave, rather than isolated within specific allocations.
The concentration of returns in a small number of technology companies is one visible expression. A less visible layer sits beneath it in the form of infrastructure providers, including firms supplying cloud capacity, specialised semiconductors and data centre infrastructure, which occupy positions in the AI value chain comparable to earlier control points in energy or telecommunications.
Private markets add a further dimension. Between 30% and 40% of global venture capital funding is now directed towards AI-related companies, often at valuations that assume sustained growth. Some of these expectations will be realised, while others will not.
As a result, portfolios can appear diversified while remaining dependent on a narrow set of technological assumptions. This dependency is rarely visible in standard reporting, but it shapes outcomes.
Operational change inside wealth management
The shift becomes clearer at the level of daily processes. Portfolio monitoring is no longer tied to periodic reporting cycles, as systems now track positions continuously, identify deviations and generate alerts in real time. This changes both the timing and the nature of decision-making.
Client expectations are evolving alongside these capabilities. Around 46% expect personalised reporting, 44% demand digital investment solutions and roughly one third anticipate simplified execution. Together, these expectations are reshaping service models.
Large institutions are already adapting. AI tools are increasingly used to summarise client interactions and support follow-up actions, reallocating time towards interpretation and client engagement. This shift also affects internal structures. Entry-level analytical roles become less central, while initial research is increasingly automated and redistributed across systems.
Switzerland: structural strength, technological dependency
Switzerland remains one of the most important centres for global wealth management, accounting for roughly 25% of cross-border private wealth. This position is built on stability, trust and institutional quality.
Artificial intelligence introduces a different dynamic. The most advanced models are largely developed in the United States and China, while European capabilities remain stronger in research than in commercial deployment. This creates a dependency that is technological rather than financial.
Data from FINMA shows that across approximately 400 institutions, a substantial share already uses AI or is developing applications. For every system in production, roughly two remain in development, many of them relying on external providers.
This reliance limits transparency into how decisions are generated and concentrates risk in a small number of providers. Switzerland controls a significant share of global wealth, but increasingly depends on infrastructure it does not control.
More intelligence, less alignment
Firms rarely adopt a single system. Instead, they integrate multiple tools over time, each addressing a specific function such as market analysis, reporting or compliance, improving efficiency in isolation.
The challenge arises when these outputs need to be combined. Probabilistic systems can produce different answers to the same question without either being clearly incorrect.
If these signals are not aligned, the result is inconsistency. Decisions may appear coherent in isolation while diverging at the portfolio level. The issue is less about access to information and more about the difficulty of reconciling it. The constraint has shifted from access to interpretation.
Risk changes form
The risks associated with artificial intelligence rarely appear as visible failures. Instead, they emerge as small distortions, such as misinterpretations of data, outdated pattern recognition or subtle biases in recommendations.
Because these outputs remain plausible, errors become harder to detect, creating a form of risk that accumulates gradually rather than appearing as a clear break.
Dependency adds another layer. Reliance on external providers reduces transparency into how conclusions are reached while increasing concentration risk.
Regulators are aware of this. FINMA highlights model risk, third-party dependency and cyber threats as key concerns, while accountability ultimately remains unchanged.
Adoption outpaces trust
Behaviour and perception remain misaligned. In the United States, around 19% of individuals trust AI in financial services and roughly 10% are comfortable with automated decision-making. Switzerland shows a more balanced picture, with around 46% indicating a willingness to trust AI, despite continued emphasis on transparency. Despite this, adoption continues to expand. Systems are integrated because they improve efficiency rather than because they are fully trusted.
Where advantage shifts
The structure of competition is changing. Information is no longer scarce, and analytical capability is increasingly accessible. Differences between firms now emerge from how these capabilities are organised and integrated. Those that align data, models and processes produce more consistent outcomes than those that accumulate tools without coordination. Advantage is shifting from access to structure.
What Matters to Remember
Artificial intelligence is becoming part of how wealth is managed on a day-to-day basis, whether it is explicitly recognised or not.
In practical terms, this shows up in continuous portfolio monitoring, more granular reporting and the ability to respond to changes without relying on fixed reporting cycles.
At the same time, providers are building increasingly complex systems behind the scenes. Understanding how these systems interact, how signals are prioritised and how external dependencies are managed becomes part of evaluating the service itself.
As these capabilities become more widely available, differences between providers emerge less from the tools they use and more from how they are structured and overseen.
For clients, the focus moves towards transparency, including how decisions are formed, which systems are involved and where responsibility ultimately sits.
Transparency is becoming central to wealth management, shaping how decisions are understood and where responsibility sits. Artificial intelligence is reshaping how portfolios are managed and risks are assessed; this upcoming article series examines what follows.
