Measuring AI ROI in Financial Advisors: A Simple Framework

The Adoption-Value Gap: Why 68% AI Usage Doesn't Equal Returns

Roughly two-thirds of advisory firms have already put AI to work, yet almost none can tell you what it's earning them. That gap between adoption and measurable value is where most firms quietly lose the ROI they assumed they were banking.

The numbers tell the story. According to Schwab's 2025 RIA Benchmarking Study, 68% of RIA firms now report using AI in some way. On paper, that's mainstream adoption — but adoption counts logins, not outcomes.

Kitces Research makes the problem concrete. In its study of AI notetakers, the most-adopted tool — Zoom AI Companion — ranks dead last in advisor satisfaction. Meanwhile, industry-specific tools like Jump and Zocks trail in adoption but lead in satisfaction.

The lesson is uncomfortable but clear: high adoption does not equal high value. Generic tools get adopted because they're easy to switch on, not because they finish the job. Advisors need CRM note capture for compliance, post-meeting task assignment, and client recap emails — not just a meeting summary they have to clean up anyway.

What's the actual prize here? Kitces finds that advisors spend more than one hour in prep and follow-up for every hour in client meetings. That hidden second hour is the untapped ROI opportunity — and it's measurable, unlike vague "efficiency" claims.

For most firms, the stakes are personal. The industry remains dominated by small shops — 92.8% employ 100 or fewer people, per the Investment Adviser Association's 2025 snapshot. In a firm that size, every reclaimed hour shows up directly on the owner's calendar and P&L.

So what should firms actually measure instead of adoption?

• Hours spent on a workflow before AI versus after — the baseline that makes savings calculable.
• Whether the tool completes the full workflow, from compliant CRM notes to follow-up tasks.
• Advisor satisfaction, which captures friction that usage logs hide.
• Review time — the hours spent checking AI outputs before they count as real work.

That last point matters more than most firms expect. Even Morgan Stanley, which reports over 98% of advisor teams using its AI assistant, requires advisors to review and adjust AI outputs before finalizing — a real cost that naive ROI math ignores.

The same discipline applies at the tool level. Morgan Stanley tested every use case against real-world scenarios before rollout, and that "test before you trust" approach is exactly what SkillzBench is building for agent skills: sandboxed benchmarks that run each skill with and without it, publishing the measured Lift so adopters see value before committing. Browsing and testing skills free on SkillzBench is the same principle applied to your own stack.

The rest of this article turns that principle into a step-by-step framework any firm can run.

The Time-Baseline Method: Your First ROI Number

The simplest ROI calculation you can run today requires nothing more than a stopwatch and your own hourly value. Start with the workflow you want to automate — for most advisors, that's meeting prep and follow-up. Kitces research found that advisors spend more than one hour in prep and follow-up for every hour spent in client meetings, which gives you a concrete, quantifiable baseline.

Here's the four-step calculation:

• Step 1 — Measure the baseline. Track total hours spent on prep, note-taking, CRM entry, and follow-up emails for a typical week before introducing any AI tool.
• Step 2 — Measure after adoption. After 30 days with the tool, track the same workflow again, including every minute you spend reviewing, correcting, or approving AI output.
• Step 3 — Convert hours to dollars. Multiply hours saved by your effective hourly cost (annual compensation divided by working hours).
• Step 4 — Subtract the true tool cost. Include subscription fees plus any integration or setup time, then compare against the dollar value of hours saved.

The math is easy; the honest inputs are not. The most common pitfall is forgetting to count human review time as a real cost. Morgan Stanley's widely cited AI deployment — where over 98% of advisor teams use the AI Assistant and follow-ups moved from days to hours — works precisely because advisors review and adjust AI outputs before anything is finalized, with daily compliance regression testing behind the scenes (vendor-reported figures). That oversight layer is a feature, not overhead you can wish away. If your "hours saved" number assumes the AI's output goes straight into client-facing emails or compliance-ready CRM notes untouched, you're measuring a fantasy.

A second pitfall runs in the opposite direction: assuming that because a tool is widely used, it's creating value. Kitces found that Zoom AI Companion ranks first in adoption yet last in advisor satisfaction — evidence that high adoption does not equal high ROI. A generic notetaker that produces a meeting summary but doesn't complete the workflow (CRM capture, task assignment, client recaps) may save you fifteen minutes while leaving an hour of work untouched.

Run the calculation conservatively. If the Kitces benchmark says one meeting hour generates one-plus hour of prep and follow-up, and your AI tool cuts that by half after accounting for review time, a two-meeting day saves roughly one hour. Multiply across a full client calendar and the annual figure gets material — especially for the small firms that dominate the industry, where 92.8% employ 100 or fewer people and every recovered hour goes straight back into client capacity.

When you're ready to test whether a specific AI agent or skill actually delivers on a workflow before you pay for it, the same test-before-you-trust logic applies: benchmark the task with and without the tool. SkillzBench is building free sandboxed benchmarking and a free MCP connector so you can measure the lift yourself, not take a vendor's word for it.

Beyond Time Savings: Measuring Capacity Leverage and Workflow Fit

Time saved is only half the story. The more revealing ROI question for an advisory firm is what each advisor can now do with those recovered hours — and whether the tool actually fits the way your practice runs.

Adams Street Partners frames this as leverage, not replacement: the long-term value in wealth management "will not come from replacing advisors" but from "radically increasing their leverage." Instead of cutting headcount, advisors expand into tax optimization, estate coordination, and retirement income planning — all without adding staff.

That framing matters right now. According to the Schwab 2025 RIA Benchmarking Study, 78% of firms reported hiring in 2024, and recruiting staff ranked as the second-highest strategic priority. In a capacity-constrained industry — where most advisory firms employ 100 or fewer people — every recovered hour is potential revenue capacity, not just a cost line.

To calculate capacity leverage, take hours saved per advisor per month (from your time baseline), multiply by your average revenue per advisor hour, and compare against what you'd pay to hire equivalent capacity. The gap is your leverage ROI.

But raw capacity only materializes if the tool completes the full workflow. Kitces research on AI notetakers found that generic tools like Zoom AI Companion lead in adoption yet rank last in advisor satisfaction, while industry-specific tools like Jump and Zocks earn higher marks. Adoption alone can overstate value.

Score each tool against this workflow-fit checklist:

• Does it capture notes directly into your CRM for compliance purposes?
• Does it assign post-meeting tasks automatically, or does someone re-enter them?
• Does it generate and send client recap emails without manual assembly?
• Does the advisor still have to review and adjust outputs — and is that review time counted in your ROI?

Enterprise results illustrate the ceiling, with caveats. Morgan Stanley's vendor-reported case study claims over 98% advisor-team adoption and follow-ups that once took days now completing within hours. Treat those figures as directional, not benchmarkable — they're self-reported and come from a firm with resources most RIAs lack.

The practical lesson: test tools the way Morgan Stanley did, running every use case against real scenarios before rollout. SkillzBench is building exactly that kind of with-and-without benchmarking for agent skills — a free, sandboxed way to measure Lift before you commit a single advisor hour.

Test Before You Trust: Benchmarking Your AI Investments

The most expensive mistake advisors make with AI isn't buying the wrong tool — it's buying the right tool and never checking whether it actually improves the work. With 68% of RIA firms now using AI in some form, adoption has outpaced measurement, and that gap is where wasted budgets hide.

Morgan Stanley offers the enterprise-grade blueprint. Before rolling out its AI assistant to advisors, the firm ran every use case through an evaluation framework that measured model performance against real-world scenarios with expert feedback, paired with daily compliance regression testing and mandatory advisor review of outputs. That discipline helped drive over 98% advisor adoption, with document access jumping from 20% to 80% and follow-ups shifting from days to hours — though these figures are vendor-reported.

You don't need Morgan Stanley's budget to copy the method. The core move is simple: run the same task with and without the AI tool, then compare outputs side by side.

• Pick one workflow with a known baseline — advisors spend more than one hour on prep and follow-up for every hour of client meetings, per Kitces research, making meeting workflows ideal test cases.
• Run the task manually and time it, then run it with the AI tool under identical conditions.
• Score both outputs on accuracy, completeness, and compliance readiness — not just speed.
• Count review time as a real cost. If the AI output needs heavy editing, your savings shrink fast.
• Repeat across several tasks before scaling. A tool that wins on one workflow can lose on another.

This with-and-without comparison matters because adoption alone can deceive. Kitces found the most-adopted AI notetaker, Zoom AI Companion, ranked last in advisor satisfaction — a direct warning that usage metrics can overstate value.

The same test-before-trust logic applies to AI agent skills, the instruction packages powering tools like Claude Code and Cursor. SkillzBench is building free, sandboxed benchmarking that runs every skill with and without it and publishes the Lift in percentage points — the same A/B discipline Morgan Stanley applied, automated. Because published research has found roughly 26% of agent skills carry at least one vulnerability, SkillzBench also security-scans each one and pins every version to a commit SHA, so advisors can extend their AI stack without extending their risk. Skills, proven — browse the free directory and see the Lift for yourself before you deploy anything.

Your 30-Day ROI Action Plan

Knowing where AI fits your practice is one thing; proving it pays is another. A 30-day action plan turns vague enthusiasm into a number you can defend to yourself, your partners, or a compliance officer.

Start with the workflow advisors know best: meeting prep and follow-up. Research shows advisors spend more than one hour on prep and follow-up for every hour in client meetings, making it the highest-leverage place to begin (https://www.kitces.com/blog/ai-notetakers-client-meeting-for-financial-advisors-adoption-satisfaction-trends-research-productivity/). Pick that single workflow and commit to measuring it — don't roll out AI across the whole practice at once.

Weeks 1–2: Log your baseline. Track every hour spent preparing agendas, pulling client data, writing recap emails, and entering notes into your CRM. No AI yet — just honest numbers. Most small firms, which make up the vast majority of the industry, feel every saved hour directly (https://www.investmentadviser.org/industry-snapshots/).

Weeks 3–4: Run the with/without benchmark. Handle the same tasks with your AI tool and log hours again. This mirrors what Morgan Stanley did before deploying AI firmwide: testing every use case against real-world scenarios with expert feedback before rollout (https://openai.com/index/morgan-stanley/). Then calculate two figures:

• Time-baseline ROI: hours saved per week multiplied by your hourly cost, minus the tool's monthly price.
• Capacity leverage: what you do with reclaimed hours — more client meetings, deeper planning, new service lines — without adding headcount.

Be honest about hidden costs. Count the time you spend reviewing and correcting AI outputs; Morgan Stanley's advisors review AI drafts before finalizing, and skipping oversight in your math overstates savings. Also score satisfaction on the full workflow — CRM capture, task assignment, client recaps — not just adoption. The most-adopted generic notetakers rank lowest in advisor satisfaction, proof that usage alone doesn't equal value (https://www.kitces.com/blog/ai-notetakers-client-meeting-for-financial-advisors-adoption-satisfaction-trends-research-productivity/).

When you're ready to automate more of the workflow, agent skills can handle the remaining steps — and you shouldn't have to guess which ones are safe. SkillzBench is building a free directory of agent skills, each benchmarked with and without the skill so you can see the measured Lift before you trust it, plus a free MCP connector to plug vetted skills into your agents. Browse the skills, test them against your own baseline, and join the waitlist to get benchmarking results and security scans as they launch. Your 30-day number becomes the benchmark everything else has to beat.