AI can generate explanations, portfolio ideas and personalised financial prompts at almost no marginal cost. That does not make regulated advice worthless. It changes what clients are actually paying for.
The Abundance Problem
For most of financial history, advice was scarce because expertise was scarce. A client paid for access to someone who could interpret markets, compare products, build a portfolio, translate regulation, explain trade-offs and turn a vague objective into an actionable plan. Artificial intelligence changes that scarcity. A consumer can now ask for a retirement drawdown framework, a portfolio stress test, a comparison of bond ladders and annuities, or a plain-English explanation of a fund prospectus—and receive an answer in seconds.
The consequence is easy to misunderstand. If an AI system can generate unlimited financial recommendations, it is tempting to conclude that advice itself is becoming worthless. The more useful conclusion is narrower: information, explanation and first-pass analysis are becoming radically cheaper. The economic value of advice is migrating away from producing an answer and toward establishing whether the answer is appropriate, implementing it correctly, managing conflicts, absorbing accountability and helping a person act when the decision is emotionally or financially consequential.
The Financial Conduct Authority’s 2026 Mills Review makes the disruption explicit. It says AI agents may increasingly construct, rebalance and execute investment portfolios on behalf of consumers and that this could compress margins on traditional advice. The same review also warns about suitability, opacity, hallucinated advice, loss of consumer agency and the possibility that value migrates to AI providers outside the traditional regulatory perimeter.
Core thesis: AI makes recommendations abundant. It does not make fiduciary responsibility, suitability, trust, behavioural discipline, tax context, family complexity or accountability abundant. Those are where the durable price of advice is likely to migrate.
“Infinite Advice” Is Not the Same as Regulated Advice
The first distinction is legal as much as economic. A chatbot can produce something that sounds like advice without necessarily being a regulated adviser, and regulators are already wrestling with the boundary between information, guidance, personalised suggestions and formal recommendations.
In the United Kingdom, the FCA’s Advice Guidance Boundary Review was created because conventional advice is not reaching enough people. The regulator estimates that around 23 million consumers are underserved by the markets for advice and guidance. Its new targeted-support regime, which took effect on 6 April 2026, allows authorised firms to give ready-made suggestions to consumer segments using limited information, while requiring firms to make clear that targeted support is not the same as comprehensive individualised investment advice.
That policy architecture is revealing. It does not assume every financial decision deserves a full adviser relationship. Instead, it creates a continuum: general information, targeted support, simplified advice and comprehensive personalised advice. The FCA’s 2026 consultation on simplifying pensions and investment advice similarly seeks to make lower-cost advice easier to provide without abandoning consumer protection.
AI fits naturally into the lower-cost end of that continuum. It can triage needs, explain choices, personalise prompts and automate routine calculations. But when the system moves from “here are the options” to “this is what you should do, given your full circumstances,” the economic and regulatory burden rises. Someone must know enough about the client to establish suitability; someone must control conflicts; someone must maintain records; and someone must answer when the recommendation causes harm.
The Price of Information Is Falling Faster Than the Price of Responsibility
The strongest case for AI is not that it can replace every adviser. It is that it can industrialise the expensive preparation work around advice. Research summaries, meeting notes, portfolio diagnostics, product comparisons, client communications and first drafts can all be generated or assembled at machine speed.
Major wealth managers are already using AI in exactly this way. UBS says its AI-powered wealth-management tools provide thousands of advisers with personalised client intelligence and meeting briefings, allowing advisers to spend less time searching for information and more time delivering advice. The bank reports that its STAAT Insights platform is used by nearly 90% of eligible adviser teams and estimates that it saves about 1,200 hours of meeting preparation each week. Those are UBS-reported operational figures, not independently audited measures of client outcomes, but they show where firms expect AI to create value: beneath the adviser, not instead of the adviser.
Vanguard has taken a similar approach. Its generative-AI client-summary capability for advisers is designed to turn investment research into customisable client communications, including required disclosures. Morgan Stanley’s AI @ Morgan Stanley Debrief automates meeting notes, action items and draft follow-up emails while leaving the adviser to edit, decide and communicate.
This pattern matters because it changes the cost structure of advice. If the machine absorbs the production of analysis, then charging a premium merely for possessing information becomes harder. The adviser has to justify the fee through something the machine has not commoditised: judgement, context, implementation, access, coordination and accountability.
Investors Are Already Showing Where the Trust Boundary Sits
The emerging consumer evidence points toward a hybrid market rather than a clean handover from humans to machines. In HSBC’s June 2026 global survey of roughly 10,000 affluent and high-net-worth investors across 10 markets, 73% said they use AI for finance and investment. Yet only 12% said AI was the most influential factor in their last investment decision. Financial professionals and institutions were the main source of investment ideas for 62% of respondents, versus 32% for AI, and the gap widened at the point of commitment.
The same survey found that investors primarily use AI for research and analysis, strategy support and second opinions. When a decision becomes consequential, they place more value on reassurance, strategic expertise, judgement, validation and the ability to catch mistakes in machine-generated data. HSBC commissioned the survey from Ipsos, so it should be read as company-sponsored research rather than neutral industry census data. Even so, its scale and cross-market design make the direction useful: AI is becoming the research companion before it becomes the final decision-maker.
The FCA is also seeing willingness to delegate. Its July 2026 Mills Review publication says research with more than 5,000 UK retail-finance consumers found that 20%—equivalent to roughly 11 million UK adults—would be likely to use AI capable of acting autonomously within pre-set goals. But the regulator also reports consumer concerns about trust and control. The tension is the market: consumers want the convenience of automation without surrendering confidence that someone is accountable.
What Humans Can Still Charge For
The most defensible human premium is not the ability to predict markets. Few advisers can credibly sell that. It is the ability to connect multiple decisions that an isolated model may treat separately. Retirement income depends on tax, longevity, spending flexibility, sequence-of-returns risk and family obligations. A business owner’s portfolio cannot be assessed cleanly without understanding the private company, borrowing needs, insurance, estate planning and the timing of a possible sale. A family office may care as much about governance, succession and liquidity as it does about asset allocation.
Those problems are not impossible for AI. In principle, a sufficiently capable system with permissioned access to financial accounts, legal documents, tax records and household preferences could integrate them. But the technical ability to model a household is not the same as the institutional authority to take responsibility for the model’s recommendation.
The second premium is behavioural. Vanguard’s 2025–26 advice research argues that clients value advice through financial planning, portfolio construction, emotional reassurance and time savings. In its July 2025 survey of more than 13,000 Vanguard investors, “peace of mind” was the top reason respondents sought advice; Vanguard reports that 86% of advised clients felt more at ease and 76% spent less time worrying about finances. Because the research comes from an advice provider, those findings should be treated as evidence of client-reported value rather than proof of causal investment outperformance.
That caveat is important. The human adviser’s moat should not be romanticised. Empathy can be simulated, trust can be misplaced, and humans carry their own biases and conflicts. The argument is not that humans are inherently safer. It is that investors may continue to pay for a person or institution that can be challenged, held responsible and asked to interpret a decision in the context of a life rather than a prompt.
When “Free Advice” Has a Business Model
The abundance of AI guidance creates another problem: if advice appears free, the investor needs to know what is funding it. Recommendation engines can monetise product placement, cash balances, order flow, data, cross-selling or distribution economics. The relevant question is not simply what the AI recommends, but what the provider is rewarded for recommending.
A 2026 enforcement case provides a useful non-AI warning. The U.S. Securities and Exchange Commission’s order involving Ally Invest Advisors concerned “no advisory fee” robo-adviser accounts that allocated 30% of client assets to cash. The SEC found that the adviser failed to fully and fairly disclose a conflict because the cash allocation was selected in part to compensate for revenue lost by not charging an advisory fee. Ally agreed to a $500,000 civil penalty without admitting or denying the findings.
The lesson extends directly to AI. Zero-price recommendations can still be economically valuable to the provider. As AI becomes the interface through which consumers compare funds, pensions, insurance and savings products, conflicts embedded in ranking, model objectives or commercial partnerships may matter as much as the explicit fee. “Free” advice can become expensive if the recommendation engine optimises for the wrong principal.
Regulators Will Price Accountability Into the Market
The regulatory direction across major markets is not to exempt machine advice from existing duties. It is to insist that established investor-protection obligations continue to apply when AI is part of the process.
In Europe, ESMA’s statement on AI in investment services says firms using AI remain subject to MiFID II requirements, including organisational standards, conduct rules and the obligation to act in the client’s best interest. ESMA specifically highlights algorithmic bias, data quality, opaque decision-making, overreliance on AI, privacy and security as material risks. The message is straightforward: a firm cannot outsource its regulatory duties to a model.
In the United States, the policy picture is still developing. In a February 2026 speech, the SEC’s Director of the Division of Investment Management invited the industry to imagine fund- or adviser-provided AI agents that could answer investor questions from official disclosures in plain English. He also flagged unresolved issues over liability, marketing, registration and supervision. Those remarks are not an SEC rule and should not be read as Commission policy. They are still notable because they frame AI as a potential bridge to information while acknowledging that the regulatory status of an AI “adviser” depends on what it actually does.
As the boundary hardens, compliance itself becomes part of advice economics. A regulated bank or adviser is selling not only a recommendation but a controlled process: identity, suitability, records, conflict management, supervision, data protection, complaints handling and recourse. AI can reduce the cost of that process. It does not make the process optional.
The Counterargument: Human Advice Could Still Be Overpriced
There is a credible opposing view. Much of what the traditional advice industry charges for can be automated. Basic diversification, periodic rebalancing, tax-loss prompts, savings nudges, risk profiling, retirement projections and product comparisons are rules-based or data-intensive tasks. If AI systems become more reliable and are connected to open-finance data, they could deliver more personalised guidance to millions of people who cannot justify a conventional adviser fee.
The FCA’s targeted-support regime effectively concedes this point: comprehensive advice is not economically necessary for every decision. A large part of the market can be served by cheaper, more standardised interventions. The regulator is explicitly trying to widen access while maintaining a distinction between scaled support and holistic personal advice.
Younger consumers may accelerate that pressure. HSBC’s survey shows much higher AI use among Gen Z and Millennials, and the FCA’s willingness-to-delegate findings suggest that a meaningful minority of consumers could eventually allow AI to act within preset limits. If trust improves, the human adviser may move from default interface to escalation layer—called in for divorce, inheritance, business sale, retirement, litigation, concentrated wealth or other high-stakes transitions.
That would still be a large structural change. It would mean the human adviser is paid less for routine attention and more for exceptional judgement.
A Three-Tier Advice Market Is Emerging
The most plausible end-state is not “AI advice versus human advice.” It is a tiered market.
At the first tier, financial information becomes nearly infinite. Consumers can ask unlimited questions, request explanations and generate scenarios at negligible incremental cost. This layer is useful but vulnerable to hallucination, hidden commercial incentives and regulatory ambiguity.
At the second tier, regulated support becomes scalable. Targeted support, robo-portfolios and AI-assisted guidance provide bounded recommendations for defined needs. The product is not a human relationship; it is a governed decision process delivered cheaply and repeatedly.
At the third tier, advice becomes more human precisely because machines do more of the routine work. Clients pay for an accountable professional who integrates complexity, coordinates specialists, challenges assumptions, manages behaviour and stays responsible when the plan must change.
The middle of the old advice model is therefore the most exposed: advisers charging premium prices for generic market commentary, standard model portfolios and periodic reviews that technology can reproduce. If the service can be reduced to “here is what happened in markets and here is a diversified portfolio,” AI will compress its price.
What Banks, Wealth Managers and Fintechs Need to Reprice
For banks and wealth managers, the strategic mistake would be to defend information scarcity that no longer exists. Research, commentary and product comparison will increasingly be table stakes. The differentiator should be an advice architecture that combines clean client data, permissioned AI, specialist expertise, transparent incentives and a human decision-maker for consequential cases.
This also changes adviser productivity economics. If AI can prepare the meeting, summarise the portfolio, draft the follow-up and identify a life event, then adviser capacity should rise. Firms will face a choice: use that capacity to reduce fees, serve more clients, deepen service for existing clients, or expand margins. Competitive pressure will determine how much of the productivity gain reaches the customer.
For fintechs, the opportunity is not merely to produce smarter recommendations. It is to build trusted infrastructure around them: traceable data sources, explainable assumptions, suitability controls, audit trails, escalation paths and clear disclosures about where the service ends. The winning product may be the one that knows when not to give an answer.
For regulators, the priority is to preserve accountability as the interface changes. The hardest questions will concern systems that provide something functionally equivalent to personalised advice while sitting outside traditional licensing boundaries. The FCA has explicitly identified that risk. A regulatory perimeter designed around the identity of the provider will come under pressure when the consumer’s main intermediary is an AI agent.
For investors, the practical test is simple: before paying for advice—or following advice that appears free—ask who is responsible for the outcome, how the provider is paid, what information the recommendation used, which information it could not see, what conflicts exist and what happens when the model is wrong.
Evidence vs inference: Evidence shows rapid investor use of AI, regulatory moves to scale lower-cost support, and major wealth firms deploying AI to augment advisers. The conclusion that fees will migrate toward accountability, complex judgement and behavioural support is an inference from those trends—not a proven universal outcome. Fully autonomous regulated advice could become more capable and trusted than today’s systems, particularly for standardised financial decisions.
Conclusion: Advice Is Worth What Cannot Be Generated on Demand
AI is turning financial knowledge from a scarce asset into a utility. That is good news for consumers. Explanations become easier to obtain, product comparisons faster, portfolio analysis cheaper and support more personalised. The advice gap can narrow because the cost of serving an additional customer falls.
But abundance changes the meaning of value. When anyone can generate a plausible financial answer, the premium shifts to knowing whether the answer fits the person, whether the incentives behind it are clean, whether it will survive contact with taxes and family realities, whether the client will actually follow it, and who will take responsibility if it fails.
Financial advice therefore does not become worthless when it becomes infinite. The commodity part of advice becomes cheaper. The accountable part becomes easier to see.
References
3. Financial Conduct Authority — Advice Guidance Boundary Review
5. Financial Conduct Authority — CP26/10: Simplifying the pensions and investment advice rules
10. Vanguard — Four ways financial advice delivers true value
11. Vanguard — Vanguard unveils generative AI client summaries for financial advisors (5 May 2025)
12. UBS — Innovation and AI at UBS: AI-powered insights, advisor-led relationships