Generative AI is making investment information easier to obtain. It is not making the consequences of an investment decision easier to carry.

Ask a capable generative-AI tool to compare two ETFs and it can produce a useful first pass in seconds: index exposure, stated fees, concentration, currency, distribution policy, historical volatility and tracking characteristics. It may also summarise a prospectus, identify questions and explain unfamiliar terminology.

That is genuine progress. It also removes some of the scarcity value once attached to finding and organising information.

But a faster comparison is not the same as a better decision. The output may be incomplete, outdated or wrong. More importantly, it does not know why the capital exists, what liabilities it must meet, which losses the investor can absorb, which losses the family will actually tolerate, or what other assets and commitments already sit outside the screen.

The question for financial advice is therefore not whether a machine can retrieve investment information. It can. The harder question is what professional value remains once information is abundant.

Adoption is real. Displacement is not yet proven.

The evidence points to a hybrid transition rather than a clean substitution.

In HSBC’s 2026 survey of 9,993 affluent and high-net-worth investors aged 21–69 across ten markets, 73% said they used AI for finance and investment. Yet only 12% said AI was the most influential factor in their last investment decision, compared with 37% who cited financial professionals and institutions. Respondents valued advisers for judgement and validation, identifying errors in AI-generated data and interpreting complex information in personal context.

CFA Institute research similarly found that roughly one-third of surveyed Gen Z and millennial investors had used generative AI for financial learning. The technology is becoming part of how investors educate themselves before a conversation begins.

EY reports that around 29% of client assets are already self-directed and that wealthy clients use 2.3 wealth managers on average. Those figures describe a broader change in behaviour and competition: clients have more tools, more relationships and more willingness to test whether an adviser adds value.

They do not establish that GPT, Claude or generative AI alone has already compressed advisory margins. Natixis offers a useful counterweight. Its 2026 research polled 300 US advisers within a global sample of 2,950 professionals across 23 countries. Only 7% of those US advisers named improved self-directed tools, including generative AI, as their biggest current competitive threat, although 35% expected that to become the case within five years. The pressure is credible. Its timing and commercial impact remain uneven.

The information layer is becoming a commodity

Basic product information should become easier and cheaper to access. That is good for investors and uncomfortable only for advice models that depend on information asymmetry.

An ETF does not become less useful because an AI tool can explain it. Low-cost, transparent and liquid instruments can remain efficient building blocks. The shift is that describing those building blocks is no longer a strong source of differentiation.

The same applies to generic market commentary and model allocations. A client can now arrive with a plausible summary, several product comparisons and a draft portfolio. The quality will vary, but the starting point has changed. Repeating what the client has already read does not demonstrate professional judgement.

This should not push advisers towards needless complexity. Replacing an easily explained liquid instrument with an opaque alternative merely to appear sophisticated would invert the purpose of advice. Complexity is not value. A product earns a place only if its role, trade-offs and fit can be explained in the context of the investor’s objectives and constraints.

A product answer is not a portfolio decision

The distance between comparison and advice becomes visible when the questions move from product features to consequences.

What is the capital for? Which cash flows must be funded over the next two, five and ten years? What happens if a business sale is delayed, a property requires further capital or a family distribution occurs earlier than expected? Which exposures already exist through an operating company, pension, property portfolio or concentrated holding? Who has authority to act when markets fall? Which decision will become hardest to maintain under stress?

These are portfolio-architecture questions. They require assets to be read against liabilities, timing, concentration, currency, tax and legal structures, governance and human behaviour. Tax and legal conclusions belong with appropriately qualified advisers, but they cannot be ignored when a financial decision is designed.

Regulation recognises this distinction. ESMA’s May 2024 statement on AI in investment services makes clear that firms using AI remain subject to MiFID II organisational and conduct requirements and the obligation to act in the client’s best interests. MiFID II suitability requirements depend on client-specific knowledge, experience, objectives and characteristics. Product data alone cannot supply that context.

Liquidity is not a footnote

Portfolio construction is often presented as a choice between expected return and volatility. For many private investors and families, the binding constraint is time.

A portfolio may look diversified on an asset-class chart while remaining vulnerable to a single liquidity event. Commitments can overlap. Distributions can arrive later than planned. Listed assets may need to be sold at an inconvenient moment to fund taxes, family needs, capital calls or an operating business. A long-term asset can be economically attractive and still be wrong for a balance sheet that needs flexibility.

Useful advice therefore turns liquidity into an explicit architecture: near-term obligations, resilience reserves, investable liquid capital, contingent commitments and genuinely long-duration capital. It tests sequences rather than relying on one central forecast. It asks what can be changed, what cannot and who decides when assumptions fail.

No prompt can answer those questions well without accurate, complete and current information about the investor. Even with that information, the responsibility for defining the assumptions, challenging inconsistencies and documenting the decision does not disappear.

Family context changes the investment problem

Private wealth is rarely a single-person optimisation exercise. One family may contain a founder seeking strategic flexibility, a next generation prioritising liquidity, trustees focused on mandate discipline and beneficiaries with different time horizons.

A technically coherent allocation can fail if the decision rights are unclear or the family has never agreed what the capital is meant to do. Behaviour matters as much as stated risk tolerance. People who accept volatility in a questionnaire may react differently when a concentrated holding falls, distributions stop or a commitment becomes illiquid.

The professional task is not to manufacture agreement. It is to expose trade-offs early, create a decision process and make the portfolio intelligible to the people who must live with it. That may involve coordination with legal, tax, corporate-finance and other relevant advisers. The value lies partly in connecting decisions that product screens naturally separate.

Private markets increase the need for judgement, not the certainty of reward

As public-market information becomes easier to compare, private-market and real-asset opportunities may appear to offer greater differentiation. They can widen the opportunity set and provide exposure to assets, companies or financing situations unavailable through a standard listed instrument.

They are not automatically better than ETFs, and they do not make risk disappear.

Private opportunities may involve limited information, negotiated governance, uncertain valuation, concentrated exposures, capital calls, long holding periods and restricted liquidity. Access is not the same as selection; selection is not the same as suitability; and neither is a guarantee of performance.

Their assessment therefore has to begin with the whole portfolio. What role is the opportunity expected to play? Which risk is being added? How is value created? What can impair it? Are incentives aligned? What governance and information rights exist? How reliable are cash-flow assumptions? What happens if exit timing extends? Can the investor fund commitments without weakening the liquid core?

Those questions are precisely where a generic comparison becomes least sufficient. They require evidence, challenge and a clear boundary between an interesting opportunity and an appropriate decision.

The emerging value stack of advice

The adviser of the next decade should not compete with AI on the speed of producing a summary. The more durable proposition has five layers.

First, verification: identify the source, date and limitations of information, and challenge outputs that are incomplete or fabricated. In January 2024, FINRA, the SEC Office of Investor Education and Advocacy, and NASAA jointly warned investors that AI-generated information may be inaccurate, misleading, outdated or fabricated and should not be the sole basis for investment decisions.

Second, framing: define the real decision before comparing products. An apparently simple investment question may actually be about liquidity, control, succession, concentration or timing.

Third, architecture: connect liquid assets, private commitments, real assets, liabilities and contingencies into one portfolio rather than treating each opportunity in isolation.

Fourth, behaviour and governance: establish decision rights, escalation rules and a process that can survive stress, disagreement and changing circumstances.

Fifth, accountability: document why a course of action was considered, which evidence supported it, which risks were accepted and which conditions would require review.

AI can strengthen every one of these layers when governed well. It can accelerate research, surface inconsistencies, explore scenarios and improve preparation. The professional advantage will not come from refusing the tool. It will come from using it without outsourcing judgement to it.

Advice has to move beyond the answer

The information advantage is narrowing. That does not make advice valueless; it makes weak sources of value easier to see.

An adviser who primarily retrieves facts and describes products will face growing pressure from self-directed platforms and generative AI. An adviser who understands the investor’s full context, designs liquidity, connects decisions, tests complex opportunities and remains accountable for the process is solving a different problem.

The future is not ETFs versus private markets, or humans versus machines. It is an efficient information layer combined with a more demanding standard of judgement.

At CGPH Banque, work in Investment Advisory & Private Markets can include opportunity assessment and selection support within a defined mandate. It does not replace the client’s relevant advisers for suitability, allocation, liquidity or implementation decisions, and it does not imply discretionary management, distribution or guaranteed access.

A prompt can help explain the choice. Professional value begins with understanding what the choice will do to the person, family or balance sheet behind it.

References and further reading

This article is provided for general information only. It does not constitute investment, legal, tax, accounting or regulatory advice, a recommendation, an offer or a suitability assessment. Private-market and real-asset investments may involve illiquidity, valuation uncertainty, concentration, loss of capital and other material risks. Outcomes depend on the investor’s circumstances, the terms of each opportunity, market conditions and decisions taken with the relevant advisers.