Wealth Management, Deconstructed

We took a top-tier family office practice, broke it into its smallest units of work, and rebuilt it from the bottom up. The result was clean: 46 distinct jobs, which cluster naturally into three functions - maintaining the household model, managing the client relationship, and keeping the record.

The aim is to explore wealth management from first principles - reshape it, shake it, and see what falls out. The industry describes itself in familiar categories - planning, investing, reporting, etc - we'll attempt to set those aside and rebuild the job from the ground up.

That framework makes it possible to see where the hours actually go, where the value actually sits, and how the work reshapes in a systematic era.

What a financial advisor actually is

Strip away the tasks and what remains is this: a fiduciary agent managing risk, capital, and tax to maximize a family's after-fee purchasing power - under uncertainty, human emotion, and law. Everything an advisor does serves one of three functions:

The model. The advisor's primary job is keeping the household model accurate - assets, obligations, goals, relationships, preferences. Markets move. Tax law changes. Children are born. Jobs change. Documents expire. Every change updates the model, and recommendations become the deterministic output of that model. The better the model, the better the decisions.

The relationship (the “load”). What the client is often paying for is to transfer a burden - the cognitive and emotional load of watching everything, all the time - from their household to an expert. It's a relationship function: the confidence that someone competent is watching, that you're on the right path, nothing gets missed, that risks won't surprise you and opportunities won't be lost.

The spine. Beneath both runs a record-keeping function: every recommendation needs evidence, every decision needs context, every action needs a record. In a fiduciary business, the model must be a model of record.

Model jobs fail in dollars. Load jobs fail in client churn. Spine jobs fail in court.

Deconstructing the work

Of the 46 jobs, 27 sit in the model (59%), 15 in the load (33%), and 4 in the spine (9%). Hour-weighted, the model consumes nearly 70% of advisor time - it contains the chronic time sinks that keep practices busy. The load jobs, the retention-critical and client-facing pieces, take the least time but carry the highest perceived value. The spine takes the fewest hours and carries the highest stakes.

Once viewed through this lens - three functions, and 46 underlying jobs - three core findings emerge:

• Labor and value are inversely distributed. Model work is invisible when done well. Load work is what the client sees and feels. Call it the practice's 80/20: roughly 80% of advisor hours go to maintaining the model - ordering, reconciling, preparing to make a decision - while roughly 80% of the value clients perceive sits in the load: the proactive outreach, a call during the drawdown, the flagged stale beneficiary form. Clients cannot observe the quality of the model; they can feel whether someone is watching. That feeling is what earns the referral.

The industry, in other words, spends most of its time where clients see the least value, and creates most of its perceived value where it spends the least time. A practice can be analytically brilliant and still churn clients, or analytically mediocre and retain forever through perceived attentiveness. It is why response speed on small requests matters more than it should on paper: clients use it to calibrate trust for the large ones.

• Workflows are event-driven. The industry organizes itself around lifecycle stages - prospecting, onboarding, planning, investing, growth - but the work does not happen that way. It happens because something triggered it. A liquidity event occurred. An RMD deadline. A threshold was crossed. A market moved. A portfolio drifted. Events are the impetus for work; event-driven is the correct architecture for a modern practice.

• Advisors assemble more than they decide. Decisions carry all of the consequences, yet receive the fewest minutes. Most advisor time goes to aggregating information, reconciling data, and coordinating people and systems - not to judgment. And wealth management is full of one-way doors: when to realize gains, whether to convert to a Roth, how to structure an estate, when to unwind a concentrated position. These decisions cannot easily be undone, which makes the scarcity of decision-time the industry's strangest resource allocation. The advisor's scarce resource isn't administration. It's judgment.

The structural cracks

The same lens exposes why practices can't scale: the industry is solving continuous problems with periodic human attention, across fragmented systems, on partial data. The cost is visible in the data - output per employee flat at every size band, across 2,700 firms. We measure it in Why aren't we talking about operating leverage in wealth?

• Continuous oversight cannot be a human job. An advisor cannot manually run 50 variables across 200 households every day. In the practices we speak with, the worst-case scenario is a client catching an advisor on their heels - a missed wire, an overdrawn account, unfamiliarity with the client's own portfolio. Reactive is bad; proactive is good. The product is vigilance, and vigilance is precisely what periodic human attention cannot deliver.
• Personalization is an interoperability problem. Advice is impossible to scale if the client notes, profile, portfolio management system, and the held-away assets aren't talking to each other. Humans become the integration layer, re-assembling context by hand for every job - a pattern that does not survive 200 households.
• Advice quality is bounded by state quality. The real household state is fragmented, adversarially formatted, and decaying - held-away assets, unread trust documents, K-1s, the client's actual risk behavior versus their stated one. Under partial observability, the best possible policy is worse than under full observability: blind spots don't just risk a specific miss; they degrade every decision, because every decision is made against a wrong belief state. To build the best model, solve data acquisition first - otherwise the system runs perfect loops over an incomplete world.

Human bandwidth is the constraint. There must be a fiduciary - or less eloquently, a throat to choke - but leverage has to come from somewhere other than headcount. It can only come from systems.

The primitive unit of wealth management

The 46 jobs are not the fundamental unit of wealth management work. A ‘job' is a human-led bundle of tasks. Decompose the job, and there exists a more primitive unit: the quantum of work (QoW).

A QoW is the atomic unit of wealth operations. It is a portable, auditable object that carries its own logic and decision framework, data inputs, outputs, and evidence wherever it travels. Every QoW follows a common pattern: something triggers the work; the work requires data of known provenance and freshness; it executes against the firm's methodology; a decision is made; and the result is captured back into the operating spine - creating an increasingly intelligent, self-improving system.

Once work is represented at this level, it becomes observable, measurable, and systematizable.

Triggers, not lifecycle stages, are the natural axis of day-to-day work. Every quantum triggers one of four ways:

• Scheduled - the trigger time is known at creation (RMDs, K-1 season, quarterly reporting, review cycles). The threshold is a T-minus window; the easiest tier to systematize.
• Monitored - a computed condition on state the system already maintains (drift bands, idle cash, harvestable losses, concentration limits, engagement decay). Pure watching. A machine evaluating continuously always beats a human checking periodically.
• Event - a discrete fact arrives from outside and can't be recomputed if missed (drawdowns, law changes, life transitions, maturing CDs, capital-call notices). The signal is generic; the obligation is per-household. One fact fans out into many quanta.
• Request - the payload is an ask, not a fact (service calls, complex asks, wires). The clock starts at receipt; latency is the trust metric.

That is the operating loop of a modern practice: a signal - scheduled, monitored, event, or request - triggers a quantum; the quantum executes; the result lands in an auditable spine.

Systematic wealth management moves one way: Signal → QoW → decision → record. Weight the signals, and the advisor's queue prioritizes itself.

We elaborate on how we transition to system-led wealth management here.

A new operating model

1. The model automates. Monitoring, vigilance, and loops are machines' home turf. Systems not only monitor 24/7 without fail, but they proactively surface and rank the decisions that actually deserve an advisor's attention. The advisor's job shifts from searching to deciding.

This is how advisor capacity is created. Human judgment relocates into system design. The chronic time sinks of the model function are exactly the work where error is measurable and the feedback is fast - introduce determinism, reconciliation gives you ground truth, a drift breach is binary, a wash-sale window is arithmetic, an RMD is proactively addressed.

The marginal cost of watching one more household collapses.

2. The load gets the hours; the 80/20 inverts. The model function frees the load function. The open hours flow to functions machines cannot perform, where the human premium lives - presence, judgment at transitions, a well-timed call. Make the easy things easy so the human can do the hard things. Advisor judgment gets its due, enabled by firm-encoded workflows.

The advisor's day becomes a queue of staged workflows, each arriving with context assembled, firm logic encoded, rationale and impact simulation attached. Hours of assembly give way to minutes at the gate. The fiduciary boundary becomes a per-quantum gate. Reversible, bounded, deterministic decisions (tax-lot selection, wash-sale checks) sit near full autonomy. One-way doors - Roth conversions, estate elections - the attention they are due, enabled by the hours freed up by automating the two-way door tasks. And autonomy is earned, not declared: a quantum type is promoted only against the track record the spine has already captured.

3. The spine creates institutional memory. In a practice built on quanta, the record writes itself. Every quantum carries its own evidence - what it saw, what logic it applied, what the human decided - captured in the moment of the work rather than reconstructed later for an examiner. The work also becomes countable: quanta run per household per month is a number a firm can audit, contract on, and compare. The cost of a workflow collapses to pennies of compute plus a moment of human judgment.

And the memory compounds. Every time an advisor changes a recommendation instead of approving it, the system records the difference - what it proposed, what the human did instead. Across thousands of decisions, those differences add up to something the industry has never had: the firm's judgment, written down.

The implications

The first-order value of automation is not a cheaper practice - it is a larger serviceable book.

Serving a household stops costing an advisor's hours and starts costing whatever it takes to run that household's work - a number that falls as the system earns more autonomy. Family-office-grade service has historically required $50–100K a year of human labor per relationship, which is why it started at $10M in assets. Remove the labor and, by our modeling, it breaks even at $500K–$1M households. The path of least resistance for firms chasing AUM - next-gen inheritors, lower minimums, operating leverage - is the same path that brings family-office service to the mass affluent.

We see this in the field: the PE groups rolling up RIAs largely don't yet track operating leverage, and are caught off guard when asked. We take that as a sign of how early this is.

The spine serves as the prerequisite for everything above it: you cannot put semi-autonomous execution into a fiduciary practice until the record is automatic. How does a CCO approve this? Every quantum type starts at recommend; promotion to approve-to-act, then act-with-review, then act is backed by a verifiable track record the spine has already captured - approval rates, delta magnitudes, outcome verification. The audit log is simultaneously the compliance record and the promotion file.

The advisor evolves. Enduring value migrates toward complex decisions, tax strategy, family dynamics, and the transfer of cognitive load. Personalization, meanwhile, is a data problem - unified household state, clean security data, document understanding, firm methodology - and data problems get solved, encoded, and distributed.

What comes next

The practice of today runs 46 jobs on periodic human attention: inputs scattered across systems, advisors carrying context by hand, monitoring with gaps, output capped by hiring. Firms respond by rationing - higher minimums, narrower coverage, teams stretched past capacity - routinely forgoing revenue rather than breaking the model.

The practice of tomorrow runs the same 46 jobs - deconstructed into quanta, triggered by signals, confirmed at the edges - continuously, for 100% of households. Human judgment relocates into system design and concentrates at the decision boundary. Value accrues to complex decisions, expert strategy, and the transfer of the load. The model becomes systematic.

The relationship remains deeply human.