Somebody on the buying committee will ask for the ROI, and the answer usually arrives as hours saved per advisor per week. That number is easy to compute and hard to defend a year later, because the hours rarely show up anywhere a partner can see them. A 2026 industry survey found only a very loose link between what wealth firms spend on AI and the value they can measure. This post sets out the unit that does show up, assets per ask, works it through for a 12-advisor, $1.5B firm, and describes a 90-day proof that makes the renewal decision for you.
Measured in hours saved, AI's return at an advisory firm is real but capped at the cost of the time it frees. Measured in introductions, it compounds: in our worked example, one extra introduction per advisor per quarter would add 0.8 to 4.0 points of organic growth. The evidence that matters is whether asks happen and trace to assets.
- In F2 Strategy's 2026 survey of 40 firms with $8.6 trillion in assets, co-founder Doug Fritz described only a very loose correlation between AI spend and measurable value.
- Hours saved is a real return with a ceiling: it can never exceed the cost of the time it frees, and it does not compound unless the time is redeployed.
- Assets per ask is the unit that compounds. In the worked example, one extra introduction per advisor per quarter would add 0.8 to 4.0 points of organic growth.
- The evidence that managing referrals pays is consistent but correlational, and causal evidence for any AI product is young, including ours. Ask vendors for asks, introductions and traced assets, not logins.
- A 90-day proof on one segment of the book, with a baseline ask rate, answers the ROI question with the firm's own numbers.
Why is AI ROI so hard to show at wealth firms?
Spending has moved faster than measurement. In F2 Strategy's 2026 survey of 40 RIAs, wealth management firms and broker-dealers representing $8.6 trillion in assets, F2 co-founder Doug Fritz described "a very loose correlation in 2026 between firms' spend on both AI technology and its tokens and a meaningful measurable value." None of the bank and trust respondents had formally measured their AI projects. Among the firms that did measure, 68% reported 25% more efficiency in the workflows they targeted.
That last figure is the pattern. Where firms do measure AI, the headline result is efficiency, which is what most AI tools in wealth were bought to deliver. Efficiency is a legitimate return. It is just not the return a growth leader is asked to report to a board.
What is wrong with measuring AI in hours saved?
Nothing, as far as it goes. Meeting-capture and documentation tools do real work. Kitces Research found that the average advisor spends more than an hour on meeting preparation and follow-up for every hour spent in client meetings, so there is a lot of time to give back, and advisor-specific tools rate higher in satisfaction than generic ones.
The limit is arithmetic. Hours saved can never return more than the cost of the hours. If a tool frees two hours a week for each of 12 advisors, over a 48-week working year that is about 1,150 hours. Whether those hours are worth anything depends entirely on what they become. If they become more client conversations, the return is growth. If they become a shorter week, the return is a quality-of-life benefit that no renewal committee can find on a statement. Either way, it does not compound: the second year saves the same hours as the first.
This is not a case against efficiency tools. Most firms should run one. A firm can run AmpUp for Wealth Management and a meeting-capture tool side by side; the two come out of different budgets and answer different questions. It is a case against using the efficiency number to justify a growth purchase, or the reverse.
What is the right unit for AI ROI at an advisory firm?
Assets per ask. The firm's growth already runs mostly through introductions from existing clients, and the gap between willing clients and actual asks is large and well measured. Research by Advisor Impact with Schwab Advisor Services and Texas Tech University, published in 2010, found 91% of clients comfortable referring their advisor, 29% who had done so in the past year, and 2% who had referred because the advisor asked. The study is old, and the exact figures will have moved. The shape has not: the ask is the scarce event.
So the ROI model runs on inputs a firm can actually observe: how many additional asks the advisors make, what share of asks produce an introduction, what share of introductions become clients, and the assets an introduced household brings. Everything else, including hours saved, is secondary. The worked example below starts at introductions, because the ask-to-introduction rate varies too much by firm to assume; it is exactly what the 90-day proof further down measures.
A worked example, with every assumption labeled. Take a firm with 12 advisors, $1.5B in assets and roughly 600 households, so about $2.5M per household. Suppose AI support produces one additional introduction per advisor per quarter: 48 introductions a year across the firm. Assume a 1% advisory fee for the revenue column.
| Case | Introductions that become clients | Assets per new household | New assets | Organic growth added | Annual revenue at 1% |
|---|---|---|---|---|---|
| Low | 1 in 4 (12 households) | $1M | $12M | 0.8 points | $120,000 |
| Middle | 1 in 3 (16 households) | $1.5M | $24M | 1.6 points | $240,000 |
| High | 1 in 2 (24 households) | $2.5M | $60M | 4.0 points | $600,000 |
The revenue column understates the return. It is recurring: a household that joins this year pays the fee next year too, so each year's cohort stacks on the last. And introduced clients introduce, which is the compounding no efficiency gain has. None of the cases is easy. For scale, the median RIA added about 2.1% to its client count in fiscal 2025, which at a 600-household firm is roughly 13 households; even the low case would roughly double that. If it happened, it would move the firm's organic growth rate by most of a point, against a median of 4.8% for firms over $250M in Schwab's 2026 RIA Benchmarking Study. Set your own inputs and compare the result with the annual cost of whatever you are evaluating. Our own pricing model is on the pricing page.
Is there evidence that AI coaching increases advisor-driven growth?
Honest answer: the evidence that the levers matter is consistent, though correlational, and the evidence that any particular AI product moves them is young. Start with the levers. In Schwab's 2026 RIA Benchmarking Study, firms with a documented client referral plan generated 1.6x more new client assets than firms without one, and only 44% of firms over $250M have such a plan. That is a correlation, not proof of cause. The other lever is reinforcement: without it near the moment of use, referral coaching tends to fade within weeks, which is the gap AI support for the ask is meant to close.
The product evidence is thinner across the whole category, and that includes us. AI support for the ask is new, we know of no published causal studies yet, and a vendor case study is not a controlled trial. The industry backdrop also argues for caution. Form ADV data compiled by Paithos Research shows the median RIA grew regulatory assets 14.5% in fiscal 2025 while its reported client count rose just 2.1%, and nearly a third of firms grew assets while adding no reported clients at all. Market returns are masking slow client growth, which makes it easy for any tool to look effective in a good year.
So ask every vendor, including us, for the same evidence. Counts of asks made, not features used. Introductions that followed those asks. And the assets each introduction produced, traced in the firm's own systems. A vendor who can only show logins and satisfaction scores is measuring adoption, not growth.
How do you run a 90-day proof of AI ROI?
The firm's own numbers will settle the question faster than any study. A 90-day proof needs no new system of record and no firm-wide rollout.
- Pick one segment of the book. A group of advisors, or a household tier such as clients with adult children, large enough to produce introductions in a quarter. Keep a comparable segment as the reference.
- Baseline the ask rate first. For the 90 days before the proof, count asks and introductions in both segments as best the firm can reconstruct them. A rough baseline beats none.
- Instrument every ask. During the proof, record each ask, the client, the reason it was the right moment, and what happened next. Software can do most of this; a shared log works for a pilot.
- Read introductions, not sentiment. Advisor enthusiasm is not the outcome. Count introductions received and first meetings held, and start tracing assets as accounts fund.
- Decide on the numbers. At day 90, compare introductions per advisor across the segment against the baseline and the reference segment. Assets will lag introductions by months, so introductions are the leading signal the renewal decision rests on.
What should you measure, and what should you ignore?
For a growth purchase, measure the left column and ignore the right. The right column is not wrong in general; it is the wrong evidence for this decision.
| Measure | Why it matters | Ignore | Why it misleads |
|---|---|---|---|
| Asks made | The behavior the investment is meant to change | Logins and active users | Adoption is not growth |
| Introductions received | The leading indicator of new assets | Notes or summaries generated | Output volume says nothing about outcomes |
| Assets traced to introductions | The number that reaches the organic growth rate | Adoption dashboards | They report usage, not results |
| Openings surfaced versus acted on | Shows where growth is being left on the table | Hours saved, for a growth purchase | Right number, wrong question |
Every row on the left returns to one figure, the firm's organic growth rate. That is the number the Growth Office is built to show: from the conversations where referral moments surface, through each ask, to the assets it produced. How RIAs track and measure referral activity covers the measurement in more depth.
Frequently asked questions
What is the ROI of AI for a wealth management firm? It depends on the unit. Measured in hours saved, the return is real but capped at the cost of the time freed. Measured in introductions and the assets they bring, it compounds. In a worked example for a 12-advisor, $1.5B firm, one extra introduction per advisor per quarter would add 0.8 to 4.0 points of organic growth, depending on conversion and household size.
Is there evidence that AI coaching actually increases advisor-driven growth? The evidence that managing the referral channel matters is consistent, though correlational: Schwab's 2026 study found firms with a documented referral plan generated 1.6x more new client assets. Causal evidence for any specific AI product is still young. The firm's own 90-day proof, measured in introductions, is the most reliable evidence available today.
How long before AI shows ROI at an advisory firm? Introductions are visible within a quarter. Assets lag by months as accounts fund and transfers complete. Judge a pilot at 90 days on introductions, and judge the full return on traced assets over the following year.
Should we stop measuring hours saved? No. Keep it for efficiency tools, where it is the right measure. Just do not use it to justify a growth purchase, or judge a growth tool by it.