Held-Away Account Monitoring: AI for Rollover Capture

Ask any RIA or wealth-management partner about their book and they'll quote AUM to the basis point. Ask them about their clients' total balance sheets and most can't tell you. That gap — between assets you manage and assets you know about — is where the next decade of organic growth lives for most firms.

Held-away accounts are the assets your clients hold somewhere else: a 401(k) at a current employer, a HSA at Fidelity, a CD ladder at a community bank, a $40k brokerage account left at the custodian where they started 20 years ago. Some of these become managed assets the moment a triggering event occurs (job change, retirement, account balance threshold). Most firms miss those triggers because they don't have visibility.

AI changes the visibility math. Here's how.

Why this is the highest-ROI AI workflow most firms haven't deployed

The math at a typical $1B AUM firm:

• 500-1,000 households served
• Estimated held-away assets per household: 30-150% of managed assets (varies hugely by demographic)
• Realistic rollover capture rate when triggered properly: 40-70%

If your firm captures even 20% of the held-away pool over 5 years through better trigger detection, that's 6-30% AUM growth from existing relationships. Compounded by retention (clients with all their assets at one firm don't leave), this is the most underbuilt growth lever in wealth management.

Yet most firms don't track it.

What "held-away monitoring" means in practice

Three pieces:

1. Discovery: what does the client actually have?

Clients tell you held-away information in lots of places that don't normally feed your CRM:

• Onboarding documents (often a balance sheet)
• Tax documents shared during planning (1099s from other custodians)
• Notes from review meetings ("oh and there's the old 401(k) from when I was at ABC Corp")
• Email mentions
• Estate planning conversations

An AI pipeline that ingests these sources (with explicit client authorization) into a structured held-away record is the foundation.

2. Triggers: what events should fire a rollover conversation?

The classics:

• Job change (held-away 401(k) becomes rollover-eligible)
• Reaching age 59½ (in-service distributions become an option)
• Retirement (full 401(k) rollover window)
• Custodian closure or notable event (fee changes, capability changes)
• Tax-bracket changes (Roth conversion windows)
• RMD eligibility on held-away IRAs
• Hitting an AUM threshold that justifies a planning conversation

Each trigger requires a specific kind of intelligence to detect. Some triggers come from public records (employment changes via LinkedIn). Some come from custodian disclosures. Some come from CRM data (a note saying "John mentioned he's thinking about retirement").

3. Action: who reaches out and how?

Once a trigger fires, the right pattern is:

• AI generates a personalized advisor talking-point brief
• Advisor decides whether to reach out (some triggers are too soft to act on)
• Outreach happens through normal channels (advisor email, phone call)
• Outcome tracked back to the trigger record

The AI does not make the rollover recommendation. That's a fiduciary act subject to DOL/SEC oversight. The AI surfaces the opportunity and supplies context.

Compliance considerations

Held-away monitoring sits on a fiduciary line. Two important considerations:

Reg BI / fiduciary obligation on rollover recommendations. Recommending a rollover from a 401(k) to your firm's managed IRA is a fiduciary act subject to Reg BI (for BDs) and adviser fiduciary standards. The recommendation must be in the client's best interest, not the firm's. Documentation matters — what alternatives were considered, why the recommendation, what fees vs the alternative.

The AI doesn't make the recommendation. It surfaces the opportunity and pulls together the data the advisor needs to make and document the recommendation responsibly.

Client authorization for data ingestion. If the AI pipeline ingests client-shared documents, the client needs to have authorized that use. Standard client onboarding language should cover it; verify with your compliance team.

The build

Four data sources:

CRM: Redtail, Wealthbox, Salesforce FSC. Pull all notes mentioning external accounts, employers, financial events.

Document repository: wherever you store client onboarding docs, tax returns, planning documents. Ingest text. NLP extract structured "held-away records": account type, custodian, approximate balance, last-mentioned date.

Public signal: LinkedIn change-of-employment monitoring (with client consent or via public records), regulatory filings if relevant.

Custodian data: for cases where the client has signed permission for data feeds to be established from their other custodians (some firms offer this as an option).

Output:

Trigger dashboard: advisor-facing view of every triggered opportunity with priority, suggested action, and the data backing it.

Advisor brief: for each trigger, a one-page brief the advisor can read in 3 minutes before deciding to reach out.

What this looks like at scale

For a 10-advisor RIA:

• Pipeline ingests ~3,000 client touchpoints/month (notes, emails, doc uploads)
• Detects 15-40 triggers/month worth advisor attention
• Of those, advisor acts on 5-15
• Rollover conversion rate (when advisor reaches out) typically 30-60%

3-8 new rollovers per month at typical RIA average account size = $1-5M of new AUM per month. Annualized: $12-60M new AUM from a workflow that was previously zero.

Build cost: $30k-$80k depending on integrations + ongoing $2k-$5k/month.

The ROI math here is some of the cleanest in advisor-facing AI work. Most CFOs approve this engagement after one meeting.

Where it gets stuck

Three patterns of failure:

Compliance signoff for client-data sources. Some firms have rigid limits on what client data can flow into AI pipelines. Plan for the CCO conversation as part of scoping; sometimes the privacy posture requires architectural adjustments.

Trigger over-firing. A pipeline that flags every employment change as a rollover opportunity will be ignored within a month. Tuning triggers to fire only on advisor-grade signals matters.

Advisor adoption. Some advisors don't want a trigger dashboard. Build the workflow so it integrates with where the advisor already lives (their email, their CRM task list) rather than another tool to check.

How to start

The minimum-viable held-away workflow at a firm not currently doing this:

Phase 1 (4 weeks): Discovery layer only. AI ingests existing CRM notes and uploaded docs, builds structured held-away records. Advisors see what their clients actually hold for the first time.

Phase 2 (next 4 weeks): Add one trigger — usually "age 59½ approaching" because it's high-confidence and Reg BI-clean to discuss.

Phase 3 (next 8 weeks): Add 3-5 more triggers based on which surface the highest conversion opportunities for your specific book.

Three months in, you have a working held-away pipeline driving advisor conversations that weren't happening before. That conversion math is what makes this the deployment we recommend most often for RIAs above $500M AUM.

If you're at that scale and not doing this, it's almost certainly the highest-ROI AI investment available to you.