Essay · Agentic Wealth
What agentic rails mean for the advisor P&L
Overhead in an advisory firm rises with size. Every technology this profession has bought so far was priced beside the labor, which is why. Agents land on a different line of the income statement.
The line that has only gone one direction
An advisory firm's income statement is short. Revenue is a fee, usually on assets. Underneath it sits overhead — rent, software, compliance, operations staff, management, everything that is not the client-facing work. Whatever survives that pays the people who do the client work, and the owner who runs the place.
Michael Kitces's benchmarking work puts the overhead ratio near 34% of revenue for a firm doing $1 million. At $2 million it is 39%. At $10 million it is 40%. At $50 million and above it is still 40%.
Scale helps once, at the very bottom, and barely. Doubling revenue from roughly $500,000 to $1 million takes the overhead ratio from 35% to 34% — one point, for twice the work. Past that it climbs, then holds at 40% for as far as the data runs. Kitces's summary: the larger the advisory firm, the higher the overhead expenses tend to get, which he calls the opposite of achieving scale.
Most of what is strange about this industry's economics follows from that. It is why the second hundred clients are less profitable than the first, and why so many good practices stop at whatever size the owner can hold personally. It is also a large part of why owners sell, though the sale usually gets explained as succession — a buyer at a higher multiple can pay today for a margin the seller was never going to produce on their own.
Why it has stayed put
Every dollar of new revenue in this business has arrived carrying its own labor.
More households mean more meetings, more paperwork, more reconciling, more documentation. More people to do that work. Then more people to coordinate the people, and that last part is what compounds. A firm at $10 million has an operations manager, a chief compliance officer, someone running client service, and a layer of management whose job is making sure the other layers work. Every one of those roles is necessary. Every one of them is also overhead.
Technology has been available the whole time and has moved this very little. Wealth managers spend somewhere between 5% and 10% of revenue on technology, against a 15% benchmark for financial services generally. Each system bought capability and brought its own operating cost: someone to administer it, someone to reconcile its output, someone to keep its assumptions current.
A second cost arrives once a firm has several of them. The systems ask for the same information and do not talk to each other, so the same household gets entered, checked and corrected in four places. Keeping those four copies consistent becomes its own job. In most industries that is what enterprise software is for, and enterprise software has not been available to a firm this size at a price it could pay. So it has been done by people, and people are overhead.
The measured result is public and it is modest. Schwab's benchmarking has clients per professional going from 53 in 2019 to 61 in 2023. Operations and administration hours per client went from 17 to 15 across the same four years. That is roughly 15% of productivity, over four years, in a period when this profession bought a great deal of software.
Where the cost lands
The difference with agentic infrastructure is in what is being purchased.
Prior systems were tools, and a tool needs an operator. The operator is a person, and the person is on the overhead line. That is why the technology budget and the technology's cost were never the same number, and why a firm could double its software spend without the income statement improving.
An agent does the work itself, against a written methodology, and shows what it did. Once it is running, nobody on the team has to manage it or become an expert in it. Work bought as work is priced against labor rather than beside it. That is the first claim any technology in this profession has had on the labor line.
It reaches the data problem underneath as well. Agents read and write across systems that were never built to talk to each other, and software has gotten cheap enough that a small firm can afford the result. One consistent set of household data, with intelligence that can read it, used to take an enterprise budget. A firm of three can have it now, which is most of what agency means for an owner.
The condition
That holds only where an agent can reach the work.
An agent is bounded by what it can read from and write to. Notetaking has no dependency on a firm's systems at all, which is why it is where most of this profession's adoption has stopped. Everything that would move the overhead line is the other kind of work — mapping a household's data, ingesting statements and trust documents and K-1s, moving an account, reconciling a position. All of it requires reaching into systems that were never built to be reached into by anything other than a person at a screen.
Where reads are limited to what a screen renders and a write means someone clicks, the agent proposes and a person still executes. The handoff survives. So does the role built around the handoff, and so does the overhead.
So the argument is conditional on the substrate. Work bought as work lands on the labor line only where the work can be done end to end. Pointed at systems it cannot reach, the same agent buys a faster version of what a person could already do, and the overhead ratio does what it has always done.
What a point of overhead is worth
Each point of overhead ratio removed is a point of operating margin.
At $2 million of revenue, one point is $20,000 a year. Move overhead from 39% to 33% and the firm keeps $120,000 more annually, on the same revenue and the same clients. The 39% comes from the benchmarking data. The six points is our estimate. There is no benchmarking finding behind it yet, and it is the number in this piece most worth arguing with.
Then it compounds in a way most owners underprice. RIA transactions in 2025 cleared at a median 11.6 times EBITDA, up roughly 40% since 2020, across 60 deals tracked by Advisor Growth Strategies. That figure gets repeated everywhere, and for most owners it is the wrong one. The average seller in that set was around $500 million. Practices between $100 million and $250 million have been transacting closer to 6.5 to 8.5 times adjusted EBITDA.
Use the multiple that applies to your own firm. At seven times, that same $120,000 of recurring margin is about $840,000 of enterprise value. Which means the decision about what to do with recovered capacity and the price of the firm at sale are the same decision, made a decade apart.
The case against
Three arguments cut against this.
This profession has never once banked an efficiency gain. Kitces's research on financial planning software found that advisors used it to go deeper rather than faster. When account aggregation took the manual work out of data gathering, they spent the recovered time on more qualitative discovery with the same households. The software with the highest time cost also carries the highest user satisfaction, which should trouble anyone selling efficiency. Given the choice, advisors picked comprehensiveness every time. If agents follow that pattern, overhead does not move at all. Advisors use the recovered hours on the households they already have, the advice gets better, and the bottom of the income statement looks exactly as it did.
Kitces argues the same point forward, about AI specifically. His position is that it should raise the quality of advice rather than the number of clients, and that the 200-client version of this ends in burnout. He is the most careful researcher this profession has.
The gain may not be the owner's to keep. A capacity gain available to every practice at once tends to get competed away, and price is where it usually goes. An advisor who passes it through as a lower fee has given the entire thing to the client, which is a fine outcome for the client and produces nothing on the income statement.
What I believe
I don't know whether overhead falls. I don't know how much of any gain gets competed away, or how quickly (but I do know that those that don't embrace agentic AI won't be able to compete). I don't know the timing, and neither does anyone selling it.
I will stand behind something narrower. The advisory P&L is a labor P&L — it always has been, and the overhead data says so in a way that is hard to argue with. For as long as this profession has measured itself, the only technology available to it was priced beside that labor. That is a different relationship to the cost structure than any prior system had.
Whether it arrives as margin is a choice among three: margin, depth, or price. Those three are not worth the same. A buyer pays a multiple for margin. Depth pays back through retention and referrals, more slowly and less directly, and buyers do pay for it eventually. A fee cut shows up on neither statement but increases efficiency and the probability of a successful client outcome.
An advisory practice is a math problem wearing a lifestyle choice. Prior to agentic AI, software just made the math more complicated.
That choice now has a price attached, and the price is calculable. It never was before, because there was nothing to decide.
The overhead ratio is the number that will come down as the work no one wants in these services gets done by agents.
Further reading: Kitces — why advisory firms don't actually scale with growth · Kitces — why financial planning software doesn't make advisors faster · Kitces — why the second 100 clients are less profitable than the first · Schwab 2024 RIA Benchmarking Study · RIA valuations at a median 11.6x EBITDA in 2025 — Advisor Growth Strategies via WealthManagement.com · CT Acquisitions — RIA M&A multiples by firm size · Kitces on AI raising quality rather than scale