The AI Ladder for Advisory Firms: Which Rung Is Your RIA On?

Most advisory firms can tell you which AI tools they've bought. Almost none can tell you what rung of the ladder they're standing on, and that's the difference between AI that saves an advisor twenty minutes and AI that lets a 10-person firm operate like a 25-person one.

The adoption numbers say firms are climbing. Per Schwab's 2026 RIA & AI Study (Logica Research, n=533), 63% of surveyed Schwab-custody RIAs now use AI in some capacity, and of those AI users, 82% rely on generative AI. But the same study found that only about one in ten AI-using RIAs have fully integrated AI into their strategy. Translation: nearly everyone is on the ladder, and almost everyone is stuck on the bottom two rungs.

The AI Ladder is the map. It has four rungs.

The four rungs of the AI Ladder

Rung 1: Chat

AI is used ad hoc, by individuals, in a browser tab. An advisor opens ChatGPT or Claude, pastes in a paragraph, gets a rewrite or a summary, copies it back out. There's no firm standard for which tool, no approved-tool list, and usually no policy.

• What it feels like: "We use AI" means "some of us use AI when we remember to."
• The value: individual productivity. A faster email here, a cleaned-up note there.
• The risk: this is where shadow AI lives. When use is ungoverned, client non-public information ends up in consumer tools that may retain and train on it, which is the exact Reg S-P exposure examiners now look for. It's no accident that 78% of RIAs have no written AI policy (ISS Market Intelligence, n=466, June 2025): a firm on the Chat rung has nothing to write a policy about yet.

Rung 2: Assist

AI is embedded in specific, named workflows, with a human reviewing every output before it leaves the firm. The firm has decided which tools are approved for which tasks.

• What it feels like: an AI note-taker drafts every meeting summary; AI drafts the first pass of client emails and market commentary; the advisor edits and approves.
• The value: real time back on defined, repetitive tasks. This matters because advisors spend roughly 80% of their time on non-client-facing work (Kitces, 2019). Assist chips directly at that block.
• The risk: manageable, because a human is the checkpoint. This is the first rung where a written AI policy and Rule 204-2 recordkeeping actually have something to govern. If you're here, the written AI policy guide is your next governance step.

Rung 3: Run

AI runs a workflow end-to-end, with humans at checkpoints rather than doing the execution. The work happens whether or not a specific person remembers to do it.

• What it feels like: a client meeting ends, and the transcript, the CRM log, the task list, and the draft follow-up email all generate automatically. The advisor approves at defined gates instead of building each artifact by hand.
• The value: capacity expansion. The workflow is now a system, not a person's habit. This is the rung where firms stop saying "AI saves me time" and start saying "we handle more clients without more headcount."
• The risk: it shifts. You're no longer worried about one advisor pasting into ChatGPT; you're worried about whether the system's checkpoints are real and whether its outputs are archived. Governance becomes design, not a memo.

Rung 4: AI Team

Multiple Run-level workflows are orchestrated together, firm-wide, with people supervising the system rather than operating each piece. AI handles the throughput of whole job functions (onboarding, quarterly reporting, compliance archiving, prospect nurture) under human oversight.

• What it feels like: the firm operates at a headcount it doesn't have on payroll.
• The value: structural. This is the rung where a small RIA's economics start to look like a much larger firm's.
• The reality: almost no small-to-mid RIA is here yet, which is exactly why the rung is worth naming. It's the destination, not next quarter's project.

A 10-question self-assessment: find your rung

Answer yes or no. Count your yeses.

1. Does your firm have a written, agreed list of approved AI tools?
2. Is there a written policy governing what client data can go into which tools?
3. Does at least one repeatable workflow (e.g. meeting notes) use AI on every instance, not just when someone remembers?
4. Is there a defined human-review checkpoint before AI output reaches a client?
5. Are AI outputs and their source material archived to meet Rule 204-2 retention?
6. Does any workflow move from step to step automatically (e.g. meeting to CRM log to task) without manual re-keying?
7. Can a workflow run correctly even if the person who "owns" it is out for a week?
8. Do two or more separate AI workflows hand off to each other?
9. Does a non-advisor (ops, an admin, or the system itself) trigger and run AI workflows day-to-day?
10. Does your firm measure the time or capacity each AI workflow returns?

Scoring:

• 0 to 2 yeses → Chat. Start by picking one workflow and one approved tool. Don't try to fix everything at once.
• 3 to 5 yeses → Assist. You have human-checked AI in real workflows. The climb is to make one of them run on its own with checkpoints.
• 6 to 8 yeses → Run. You have systematized workflows. The climb is orchestration: connecting them and handing the triggers to your team or the system.
• 9 to 10 yeses → AI Team. You're operating AI as infrastructure. The work now is governance-at-scale and measurement.

The self-assessment gives you a rung. For a deeper, department-by-department read on where AI would return the most time in your specific firm, run the AI Bottleneck Scorecard.

How to climb a rung (without falling off)

The mistake firms make is trying to jump from Chat to Run by buying more tools. Rungs are climbed by systematizing one workflow at a time, not by accumulating software. The path is always the same: pick the workflow that wastes the most advisor time, govern it (approved tool, data rule, review checkpoint), then automate the hand-offs inside it until it runs on its own.

Big firms and small firms climb differently. A billion-dollar RIA climbs by buying integration and dedicated ops headcount; it can afford to put a person in charge of each rung. A firm at the industry-median size of around 8 employees (IAA Industry Snapshot) climbs by choosing fewer workflows and automating them deeper. Depth on three workflows beats shallow tool sprawl across ten. The small firm's advantage is that it can move a single workflow from Assist to Run in weeks, not quarters, because there's no committee between the decision and the change.

That "automate workflows, not roles" principle is its own framework, AI Leverage, and it's the operating discipline behind every rung above Chat. If you don't yet know which workflow is costing you the most, the AI Bottleneck Scorecard is the fastest way to find it.

Frequently Asked Questions

What are the four stages of the AI Ladder?

Chat, Assist, Run, and AI Team. Chat is ad hoc individual use in a browser. Assist is AI embedded in named workflows with a human reviewing every output. Run is AI executing a workflow end-to-end with humans only at checkpoints. AI Team is multiple workflows orchestrated firm-wide, with people supervising the system rather than operating each part. The rungs describe how systematized your AI use is, not how many tools you own.

How do I diagnose which rung my firm is on?

Use the 10-question self-assessment above and count your yeses: 0 to 2 is Chat, 3 to 5 is Assist, 6 to 8 is Run, 9 to 10 is AI Team. The questions test for governance (approved tools, data rules), workflow integration (does AI run on every instance), and autonomy (does it run without a specific person). A firm that owns expensive tools but answers "no" to questions 3 through 7 is still on the Chat rung.

What does each rung look like in a real RIA workflow?

Take meeting notes. On Chat, an advisor occasionally pastes notes into ChatGPT to clean them up. On Assist, an approved AI note-taker drafts every meeting summary and the advisor reviews it. On Run, the meeting automatically produces a transcript, a CRM log, a task list, and a draft follow-up, with the advisor approving at gates. On AI Team, that meeting workflow connects to onboarding, reporting, and compliance archiving so the whole client lifecycle runs as one supervised system.

What's the concrete next step to climb a rung?

Pick the single workflow that wastes the most advisor time, not the flashiest one. Govern it (approved tool, written data rule, defined review checkpoint), then automate the hand-offs inside it one at a time until it runs on its own. Climb by deepening one workflow, not by buying more software. Tool sprawl keeps firms stuck on Chat.

How do large RIAs climb compared to small firms?

Large firms climb by buying integration and assigning dedicated ops headcount to each rung. Small firms, where the industry median is about 8 employees, climb by choosing fewer workflows and automating them deeper, which they can do faster because there's no layer between the decision and the change. Depth on three workflows beats shallow coverage across ten.

How long does it take to move up a rung?

For a small-to-mid RIA, moving a single workflow from Assist to Run typically takes weeks, not quarters, once the workflow is chosen and governed. Moving the whole firm up a rung takes longer because it means doing that for several workflows. The firms that stall are the ones trying to climb everywhere at once instead of finishing one workflow before starting the next.