Artificial intelligence has moved from novelty to infrastructure across the advisory profession. The firms that thrive in the coming era may be those that use it to deepen, rather than replace, the human relationship at the center of wealth management, and the difference between those two paths is a matter of deliberate design.
From Disruption to Infrastructure
The early conversation about AI in wealth management fixated on disruption: would algorithms replace advisors? Several years into real adoption, that framing looks increasingly beside the point. Robo-advice captured a segment of the market and then plateaued, while the technology itself migrated into the operations of traditional firms. The more interesting question now is not whether AI replaces the advisor, but how thoroughly it transforms what an advisor can do for each family.
The profession has absorbed general-purpose technology before, and the pattern tends to repeat. Portfolio accounting systems did not eliminate the investment professional; they raised the standard for what a client could reasonably expect to see. Financial planning software did not eliminate the planner; it turned planning from an artisanal exercise into a repeatable discipline and moved the practitioner from calculation toward interpretation. Client relationship systems did not eliminate the relationship; they made the relationship legible to a team rather than resident in one person’s memory. In each cycle, the technology commoditized a task and revalued a judgment. There is little reason to expect the present cycle to behave differently, and considerable reason to expect it to move faster, because this generation of models arrives already competent at language, synthesis, and inference rather than requiring years of workflow-specific engineering.
What separates this cycle from its predecessors is the nature of the input. Earlier systems required structured data: fields, tables, and codes that somebody had to populate by hand. Language models operate on the unstructured material that constitutes the actual substance of a wealth management relationship, which is to say trust instruments, partnership agreements, operating documents, K-1s, meeting notes, correspondence, term sheets, and the accumulated context of decades of family decisions. That material has historically been expensive to read, difficult to search, and effectively invisible to a firm’s systems. Making it legible is the substantive change, and it is a change that favors complexity. The families whose circumstances resisted systematization are precisely the families for whom the new capabilities may matter more.
The Arithmetic of Attention
Start with the mundane and consequential: time. Studies of advisor workflows have consistently found that client-facing professionals spend a minority of their hours actually facing clients, with the balance consumed by meeting preparation, documentation, data gathering, and administrative coordination. AI compresses each of those categories. Meeting preparation that once took hours can take minutes. Notes transcribe and file themselves. Account data, planning scenarios, and document summaries assemble on demand. The arithmetic of a service calendar changes when the overhead per relationship falls by half.
The consequence deserves to be stated plainly, because it is easy to mistake for a cost story. Reclaimed hours can be spent in three ways, and the choice among them may determine what kind of firm emerges from this decade. A firm can serve more families with the same staff, which converts the technology into margin. It can serve the same families with fewer people, which converts it into headcount reduction. Or it can serve the same families with the same team at materially greater depth, which converts it into service. The third path is the harder one to measure and, in our view, the one with the longer half-life. Capacity has the potential to strengthen the client relationship.
Depth is not an abstraction. It looks like a second read on an estate document before it is signed rather than after. It looks like a call placed the week a lockup schedule shifts rather than the quarter after. It looks like a planning team that has already modeled the charitable alternative, the entity alternative, and the do-nothing alternative before the family asks which merits consideration. None of that is new in kind. What changes is how many families a team can hold to that standard at the same time, and how consistently the standard survives a busy quarter.
Capacity returned to the client relationship tends to compound. Capacity returned to the income statement tends to be competed away.
Four Capabilities, One Operating Model
It helps to be specific about what the term refers to inside an advisory business, because a single word flattens four distinct capabilities that carry different returns and different risks.
Generative systems produce drafts: memoranda, summaries, correspondence, presentation material, and code. The return is speed. The risk is fluency without accuracy, which makes human review non-negotiable before anything reaches a family.
Predictive systems estimate outcomes from historical and current data: cash flow ranges, liquidity needs, tax results under alternative structures, funding probabilities for a stated objective. We offer a variety of choices with different potential returns.. The risk is false precision, because a model’s confidence interval reflects its inputs rather than the world.
Assistive systems recommend and automate inside a workflow: flagging an unharvested loss, surfacing an account that has drifted from policy, queuing an outreach when a triggering event appears. The return is coverage. The risk is complacency, because a system that has quietly stopped running looks identical to a system with nothing to report.
Conversational systems handle dialogue: internal retrieval across a firm’s own knowledge base, service logistics, onboarding coordination. The return is responsiveness. The risk is placing a machine where a person belongs.
A serious operating model treats these separately. Generative work sits behind named human review. Predictive work sits behind a stated methodology and disclosed assumptions. Assistive work sits behind testing and monitoring. Conversational work sits behind a clear boundary around what a family experiences directly. Firms that collapse all four into a single enthusiasm tend to inherit the risks of each without capturing the return of any.
Complexity as a Data Problem
For ultra-high-net-worth families, the implications run deeper than efficiency. Complex balance sheets spanning operating businesses, real estate partnerships, alternatives, insurance structures, and multiple trusts generate enormous information surfaces, and the traditional service model reviews that surface episodically: quarterly meetings, annual planning cycles, tax season. AI makes continuous monitoring feasible. Concentration drift, tax-loss opportunities, maturing lockups, document inconsistencies, and provisions affected by new legislation can surface as they emerge rather than when the calendar happens to catch them.
Consider the shape of a representative balance sheet at this level. A founding family may hold a concentrated position in a private company, a set of grantor trusts funded at different valuations under different exemption regimes, several real estate partnerships with distinct debt maturities and depreciation profiles, a private fund portfolio with capital calls arriving on independent schedules, a donor-advised fund, a private foundation carrying its own distribution requirement, and a residence pattern spanning two states with divergent domicile rules. Each element is intelligible in isolation. The interactions are where value and risk accumulate, and the interactions are what an episodic review process handles poorly, because a calendar organizes attention by account rather than by consequence.
A useful way to frame the opportunity is that a family balance sheet is a graph rather than a list. Entities own entities. Trusts hold interests in partnerships that hold interests in operating companies. A change at one node propagates to others. When a firm can hold that graph in a machine-readable form, a question that once required a week of coordination among counsel, accountant, and advisor can become a query. How does estate tax exposure shift if the operating company is valued at a different multiple? Which entities hold assets with embedded gains that could interact with a contemplated charitable strategy? Which trust provisions reference a statute that has since been amended? These are ordinary questions. They have simply been expensive enough to ask sparingly.
A family balance sheet is a graph rather than a list. Entities own entities, a change at one node propagates, and the interactions are where value and risk accumulate.
From Episodic Review to Continuous Awareness
The service model that follows from this looks different in cadence rather than in kind. Instead of a quarterly review that reconstructs a family’s position from scratch, a firm maintains a standing picture and examines the deltas. A concentrated position crosses a threshold set in the investment policy statement. A partnership approaches a debt maturity in a repricing environment. A trustee appointment references an individual who is no longer serving. A state of residence accumulates day counts that approach a statutory line. A private fund issues a capital call that interacts with a planned distribution. Each of these is discoverable today by a diligent professional with sufficient time. The change is that discovery becomes systematic rather than dependent on which item happens to surface during the review that quarter.
Continuous awareness carries its own failure mode, and it should be designed against from the outset. A system that surfaces everything surfaces nothing, because alert volume is the fastest route to alert fatigue. Materiality thresholds, suppression rules, and human triage are not administrative afterthoughts; they are the substance of whether the capability produces attention or noise. The discipline that matters is deciding in advance what rises to a client conversation, what rises to an internal review, and what is logged and left alone. That decision is a judgment about the family, which means a person makes it.
Compressing the Research Cycle
Research and analysis capabilities follow the same pattern. A planning team can now stress-test a family’s plan against dozens of scenarios, synthesize the practical effects of a new tax bill within days of passage, or analyze the interaction between an estate structure and a proposed transaction at a depth that previously required weeks of specialist time. The quality ceiling of advice rises when the cost of rigorous analysis falls.
The legislative cycle offers a concrete illustration. When a reconciliation bill runs to hundreds of pages and reaches provisions touching qualified small business stock, the estate and gift exemption, opportunity zones, depreciation, and the treatment of pass-through income, the traditional sequence is familiar: wait for the summaries, wait for the technical commentary, then revisit affected families as the annual review cycle permits. Under the One Big Beautiful Bill Act enacted in July 2025, for example, qualified small business stock moved to a tiered exclusion of 50 percent at three years, 75 percent at four years, and 100 percent at five years, with the per-issuer cap raised to $15 million and the gross asset threshold raised to $75 million. The federal estate and gift exemption moved to $15 million per individual and $30 million per married couple on a permanent basis, indexed for inflation beginning in 2027. The opportunity zone program was made permanent, introducing a distinction between the original program and its successor framework that matters for anyone holding or contemplating a deferred gain.
Each of those changes may affect a different subset of families, and none of the analysis is conceptually difficult. It is a matter of identifying which families hold which facts, then reaching them while the decision is still live. That identification step is what a well-structured system does well, and it is where the calendar has historically failed. The judgment about what a family should do with the answer remains where it has been. What compresses is the interval between a change in the world and a considered conversation about it. In planning, that interval is frequently the difference between a strategy that is available and a strategy that has closed.
Where the Risks Live
The risks deserve equal attention, and treating them seriously is itself a competitive differentiator. Models can be confidently wrong, and in a profession built on accuracy, an unverified fabrication in a client deliverable is a meaningful failure. Data privacy demands rigorous governance, particularly for families whose financial information is sensitive by nature and valuable to bad actors. Regulators have made clear that the obligations of supervision and care extend fully to AI-assisted work, which means firms need documented oversight rather than enthusiasm alone.
That expectation has a concrete regulatory footprint. In March 2024, the Securities and Exchange Commission settled charges against two registered investment advisers over statements about their use of artificial intelligence that the orders found to be false or misleading, citing the Marketing Rule under Advisers Act Rule 206(4)-1 and the Compliance Rule under Rule 206(4)-7. The conduct at issue was promotional rather than technical: describing capabilities the firms did not possess. In June 2025, the Commission withdrew its 2023 proposal addressing conflicts of interest associated with the use of predictive data analytics, indicating that any future rulemaking in that area would begin with a new proposal. The withdrawal removed a prospective rule. It did not alter the existing framework, under which an adviser’s fiduciary duty, advertising obligations, compliance program requirements, books and records rules, and privacy obligations apply to AI-assisted activity in the same way they apply to every other activity.
Beneath the regulatory layer sit several operational exposures worth naming directly:
- Accuracy and verification. Fluent output is not verified output. A citation, a statutory reference, a basis calculation, or a document summary that reaches a family without human confirmation is an unmanaged risk regardless of how well it reads.
- Confidentiality and data boundaries. Which systems receive client information, whether inputs are retained or used for training, where data resides, and which vendors sit in the chain are questions that belong in a written policy and in vendor diligence rather than in individual discretion.
- Model and vendor concentration. Capability sourced from a single provider inherits that provider’s outages, policy changes, pricing decisions, and model deprecations. Architecture that assumes substitution tends to age better than architecture that assumes permanence.
- Silent degradation. A model updated upstream can behave differently on the same task without any visible signal. Periodic testing against known cases is how a firm learns this before a client does.
- Reconstruction and records. If a recommendation was informed by an AI-assisted analysis, a firm should be able to reconstruct what was run, on what inputs, reviewed by whom, and when. Reconstruction is a recordkeeping discipline, and it is easier to build at the outset than to retrofit.
- Description and marketing. How a firm characterizes its technology to clients and prospects is itself regulated conduct. Precision about what a system does, and restraint about what it does not, is both a compliance obligation and a credibility asset.
The Subtler Risk: Automating the Relationship
There is also a subtler strategic risk: firms that automate the relationship itself may discover they have automated away the reason clients chose them. Wealth management at the upper end has remained an intensely personal profession because the decisions are intensely personal. Families do not hire a firm to generate documents; they hire judgment, discretion, and a steward who understands what the wealth is for. Technology that interposes itself between the family and that judgment subtracts value even while adding efficiency.
The distinction worth holding is between technology that is felt and technology that is seen. A family should feel the effects of a capable system: the call that arrives before the question does, the analysis that is already complete, the detail that nobody had to be reminded of. A family should rarely be routed into one. There is a version of efficiency that reads to a client as neglect, and at this level of relationship the difference registers immediately. A principal who receives an automated summary in place of a conversation has been handed a document and denied the counsel.
This is also a competitive observation rather than a sentimental one. As analysis becomes abundant, analysis becomes less differentiating. What remains scarce is the judgment that decides which analysis matters, the discretion that a family extends over time, and the institutional memory that understands why a structure was built the way it was. Those are the assets that a firm should be spending its reclaimed hours to build.
A Design Standard: Absorb the Mechanical, Supply the Meaningful
The design principle that resolves this tension is straightforward to state and demanding to execute: AI should absorb the mechanical so that professionals can supply the meaningful. Every hour reclaimed from preparation is an hour available for the conversations that matter, the thinking that produces insight, and the proactive outreach that families remember. The technology is designed to be unobtrusive, with the aim of providing a firm that is attentive and responsive to client needs.
Translating that principle into practice tends to require a small number of standing commitments:
- Purpose before deployment. A defined problem, a defined owner, and a defined measure of whether the capability improved a client outcome, established before the tool is adopted rather than after.
- Named human review. A professional accountable by name reviews any AI-assisted material before it reaches a family, and that review is recorded.
- Stated methodology. Where analysis informs advice, the assumptions, data sources, and limitations are documented and available, so that a conclusion can be examined rather than accepted.
- Written data boundaries. Clear rules governing which information enters which systems, retained in policy and reflected in vendor agreements and diligence.
- Testing and periodic review. Scheduled evaluation against known cases, with results documented, because a capability that is not tested is a capability that is assumed.
- Truthful description. Public and client-facing characterizations of the technology match what the systems actually do, in substance and in emphasis.
- Judgment reserved. Decisions involving family dynamics, legacy, trade-offs among competing goals, and irreversible commitments remain with people, by design and not by default.
None of these commitments is exotic. Taken together, they are the difference between a firm that has adopted tools and a firm that has built an operating model, and that difference tends to become visible over a full market cycle rather than in a product demonstration.
What Does Not Change
A fiduciary standard is indifferent to the instrument used to meet it. The duty of care and the duty of loyalty apply to an analysis produced in thirty seconds exactly as they apply to one produced over three weeks.Faster production may potentially enhance the standard, as it could lead to more efficient use of resources. A firm that can examine a question thoroughly is answerable for having examined it.
The questions that resist modeling also remain where they have been. Which child is prepared to receive what, and when. Whether a family shares the size of the balance sheet with the generation coming behind it, and how. What the wealth is intended to accomplish once the arithmetic of sufficiency has been settled. How to hold a family together through a liquidity event that changes the relative position of siblings. Whether a founder is ready to sell the company that has been an identity as much as an asset. These conversations are the substance of the work at this level, and they are not analytical problems with better tooling waiting to solve them. They are human problems, and they are handled by people who have earned the standing to raise them.
Our View
Our view is that AI is leverage for judgment rather than a substitute for it. The capabilities arriving now can extend what a thoughtful team sees, how quickly it responds, and how much complexity it can hold per family. The decisions that matter, the ones involving family dynamics, legacy, competing goals, and the weight of irreversible choices, remain irreducibly human. Firms that strive to balance multiple perspectives may contribute to the evolving nature of the profession.
We think of this as amplified intelligence rather than artificial intelligence: advanced capability combined with human judgment, applied to the specific circumstances of a specific family, governed by people who are accountable for the outcome. The technology is a means of extending attention. Attention, with care over decades, is a product that families often consider.
That has been the substance of this profession since long before the current cycle, and we expect it to remain the substance well after the vocabulary changes again.
About Certuity. Certuity, LLC is a fee-only registered investment adviser serving high-net-worth and ultra-high-net-worth families, principals, and institutions, with offices across the United States. Our approach combines certainty and ingenuity: disciplined planning for a wide range of possible outcomes, paired with forward-thinking, tailored solutions.
Important disclosures. This material is provided for educational and informational purposes and reflects the views of Certuity, LLC as of the date of publication. It is not intended as investment, tax, legal, or accounting advice, and it does not constitute a recommendation or an offer to buy or sell any security or engage any service. Views and statements regarding technology, regulation, and market conditions are subject to change without notice. Tax and estate planning provisions referenced are summarized at a high level, may be subject to further guidance, technical correction, or amendment, and may apply differently based on individual circumstances; readers should consult their own tax and legal advisers regarding their situation. Third-party information is believed to be from reliable sources, though accuracy and completeness are not represented or warranted. Investing involves risk, including possible loss of principal. Past performance is not indicative of future results. Certuity, LLC is registered as an investment adviser with the U.S. Securities and Exchange Commission; registration does not imply a certain level of skill or training. Additional information about Certuity is available in our Form ADV Part 2A, available upon request or at adviserinfo.sec.gov.