Jump, Zocks, Holistiplan, FP Alpha, and Wealth.com solved the four most-trafficked AI surfaces in wealth — meeting capture, tax extraction, estate extraction, and the planning-CRM interface. The category benchmarks are real, the productivity claims (Zocks' 10+ hours per week, the Schwab 2026 doubled adoption) are not vendor marketing, and any mid-sized RIA that hasn't deployed all four by mid-2026 is making an unforced error. But the shelf vendors stop where the differentiation starts. The L5 strategist's question is not which shelf tool to buy next — it is where there is no Jump or Zocks yet, what proprietary workflow the practice can build, and how to build it without writing custom AI infrastructure. This lesson maps the five highest-leverage white-space categories where 2026 AI shelf vendors have not yet shipped — legacy planning for blended families, business-owner exit, multi-generational philanthropic stewardship, expat tax-and-estate, complex annuity laddering — and then installs the build architecture that lets a $500M-$5B RIA produce its own prompt library, RAG pipeline, and internal AI assistant on top of the existing enterprise LLM (Microsoft Copilot, OpenAI Enterprise, Google Gemini Enterprise) without becoming a software company.

Why Shelf Vendors Stop Where Differentiation Starts

The wealth AdvisorTech map Kitces publishes each March is a useful diagnostic for understanding where the shelf vendor economy invests and where it does not. Categories where the addressable market is 300,000+ US advisors and the workflow generalizes cleanly — meeting AI (Jump, Zocks), tax extraction (Holistiplan with its 10,000-firm install base and 2026 Enterprise Advisory Board), estate extraction (FP Alpha Estate Insights 2.0, Wealth.com), CRM AI (Salesforce FSC + Einstein, Wealthbox AI, Redtail Engage), prospecting (Catchlight, SmartAsset) — get well-funded vendor competition by mid-2026 because the unit economics support a $300-$1,200 per advisor per year SaaS model at scale.

Categories where the addressable advisor population is 5,000-30,000 and the workflow is high-complexity, high-customization, or high-regulatory-friction — legacy planning for blended families, business-owner exit at the lower-middle market, multi-generational philanthropic stewardship across CRT/CLAT/DAF/private foundation vehicles, expat tax-and-estate coordination, complex annuity laddering with NAIC Model #275 best-interest documentation — do not get well-funded vendor competition because the SaaS unit economics do not work at the smaller addressable market. The shelf vendor cannot recoup a $5M-$15M product investment on 8,000 advisors paying $1,500/year. The opportunity for the firm that builds its own workflow is exactly the gap.

The Mercer Capital and ECHELON Q3-Q4 2025 RIA M&A data make the financial case explicit. The premium-attribute lift of 0.5-1.5x on the 8x-10x top-quartile EBITDA band (with premium-top at ~11.6x) per L4 Ch8 L1 is exactly the lift that documented proprietary AI workflows produce in buyer diligence. A $30M-EBITDA RIA targeting a 1.0x premium lift through proprietary AI workflows is looking at $30M of incremental enterprise value — multiples of the build cost of any single proprietary workflow.

The Five White-Space Categories Where No Jump or Zocks Has Shipped

Legacy Planning for Blended Families

The American household has restructured. Roughly 16% of US children live in blended families, and the planning workflow for the 60-year-old client on a second marriage with three kids from the first marriage and two stepchildren from the second is qualitatively different from the textbook estate plan. The decisions cascade: separate property versus marital property, conflicting beneficiary designations across IRAs and life insurance, GST-tax allocation across natural and step-descendants, prenuptial-agreement integration with the revocable trust, SLAT funding when both spouses have separate-property exclusion to use, the surviving-spouse-with-stepchildren conflict of interest at trustee selection. FP Alpha and Wealth.com extract trust documents well; neither resolves the blended-family conflict of interest in trustee selection or the GST allocation across mixed-descendant classes. The proprietary workflow that does — RAG-powered, integrated with the firm's prior blended-family memos, locked to a CFP and an estate-attorney persona — is a $200K-$400K build that no shelf vendor will produce in 2026-2027.

Business-Owner Exit at the Lower-Middle Market

The lower-middle-market business exit ($5M-$50M enterprise value, owner age 58-68, sale to private equity or strategic acquirer or ESOP or management buyout) is the highest-leverage single planning event in most clients' lifetimes. The full workflow spans 18-36 months pre-sale: QSBS Section 1202 eligibility tracking across the OBBBA dual regime (legacy 100% exclusion at 5-year hold with $10M/10x cap and $50M gross-asset threshold for stock issued on or before July 4, 2025; OBBBA tiered exclusions of 50%/75%/100% at 3/4/5-year holds with $15M/10x cap and $75M gross-asset threshold for post-7/4/2025 stock), installment-sale modeling under IRC 453, CRT funding pre-sale to defer-or-eliminate gain, 10b5-1 plan structuring under Rule 10b5-1(c) if the company is public, opportunity-zone reinvestment under IRC 1400Z-2, 83(b) election windows on restricted stock, post-sale liquidity event tax management, charitable bunching with a DAF in the sale year, and family-office-readiness assessment for post-sale stewardship. Holistiplan extracts the 1040; no shelf vendor orchestrates the 18-36-month pre-sale workflow. The proprietary build that orchestrates it — pulling cap-table data from Carta or Pulley, integrating with the trust attorney's draft documents, tracking the QSBS per-tranche classification — is the proprietary workflow that justifies a 0.5-1.0x M&A multiple lift by itself for a firm whose niche is business owners.

Multi-Generational Philanthropic Stewardship

The $30B-$50B annual flow of charitable assets through DAFs, CRTs, CLATs, private foundations, and supporting organizations is mis-served by every shelf tool except the DAF sponsors' own portals. The advisor's workflow — coordinating gift policy across a multi-generational family (G1 grandparent, G2 parent, G3 grandchild), allocating across multiple vehicles by tax-character (appreciated stock to DAF, ordinary-income generators to CRT, illiquid assets to private foundation), tracking 5% private-foundation minimum distribution requirement under IRC 4942, managing CRT income-tier accounting under IRC 664, advising on grant strategy across years, integrating with the family's overall estate-tax-exclusion-sunset planning — has no shelf product in 2026. The proprietary build is a multi-vehicle ledger plus an AI-drafted annual philanthropic strategy memo plus a grant-recommendation queue, integrated with Holistiplan for tax-character analysis and with the family's primary wealth platform (RightCapital, eMoney, or MoneyGuidePro).

Expat Tax and Estate Coordination

The US-citizen-abroad and the inbound foreign-national-in-the-US populations together represent roughly 9 million US tax filers with foreign-asset reporting (Form 8938, FBAR/FinCEN 114, Form 5471 for foreign-corporation interests, Form 3520 for foreign trusts), foreign-tax-credit coordination under IRC 901, treaty interpretation across US-UK, US-Canada, US-Germany, US-Australia, US-Singapore, US-Switzerland, US-Israel income-tax and estate-tax treaties, foreign-earned-income exclusion under IRC 911, PFIC reporting under IRC 1297, and qualified electing fund elections. No shelf wealth tool handles the cross-border coordination. The proprietary build is a country-pair-specific knowledge layer (RAG-powered against the firm's library of country-pair tax memos), plus a citizenship/residency status tracker, plus a coordinated tax-and-estate output that the firm's CPA partner and the country-pair counsel can both consume.

Complex Annuity Laddering

The dually-licensed advisor running complex annuity laddering — multi-year guaranteed annuities (MYGAs), fixed indexed annuities (FIAs), registered index-linked annuities (RILAs), single premium immediate annuities (SPIAs), deferred income annuities (DIAs), QLACs under IRC 401(a)(9) — combined into income floors and longevity-tail protection, mapped against the client's Social Security claiming strategy, Medicare IRMAA bracket management, and Roth conversion sequencing, has zero shelf tooling that produces a NAIC Model #275 best-interest documented recommendation. The proprietary build is a multi-product comparison engine plus a NAIC Model #275 documentation template plus a state-DOI-aware output (the 50-state matrix per L4 Ch6 L2). The build is straightforward; the regulatory documentation discipline is the differentiator and the supervisory protection.

Build Architecture for a Mid-Sized RIA — Without Becoming a Software Company

The temptation at every $500M-$5B RIA contemplating proprietary AI workflows is to hire engineers and start writing software. That path produces a software-company P&L, key-person risk on the engineering hires, and a maintenance liability that crowds out the advisory business. The actually-shipped 2024-2026 pattern for proprietary AI at the mid-sized RIA scale is to assemble — not write — on top of three layers the firm already owns or licenses.

Layer 1 — The Enterprise LLM

Microsoft Copilot (for the Microsoft 365-anchored firm), OpenAI Enterprise (for the Salesforce-anchored or independent firm), Google Gemini Enterprise (for the Google Workspace-anchored firm). All three offer enterprise contracts with Reg S-P 17 CFR Part 248-acceptable vendor terms (BAA-equivalent, data-residency, no-training-on-firm-data, SOC 2 Type II), GLBA Safeguards-acceptable security posture, and the API access required to build on top. The licensing cost at $30-$60 per seat per month for 50-200 internal users is $18K-$144K annually — a fraction of what custom infrastructure would cost.

Layer 2 — The RAG Vault

The retrieval-augmented generation (RAG) layer is what makes the LLM speak the firm's voice. The vault holds: every compliance-blessed IPS template, every standard Reg BI memo, every prior-approved disclosure paragraph, every approved client-facing concept memo (QCD, NUA, Roth conversion, RMD, 72(t), backdoor and mega-backdoor Roth flowchart per L3 Ch3 L3, SLAT/ILIT/CRT/CLAT/DAF decision tree per L3 Ch5 L3, 529-ABLE-SNT coordination per L3 Ch7 L2), the firm's WSPs, the ADV Part 2A current draft, the engagement letter, the fee schedule, the principal-review-queue exception log (sanitized), the AI Risk Register, the prior 12 months of Marketing Rule audit substantiation files, and the prior 24 months of approved client-facing communications by category.

The build pattern: use a vendor-managed RAG infrastructure (Pinecone, Weaviate, Microsoft Azure AI Search, or Google Vertex AI Search) rather than building a vector database from scratch. The cost is $20K-$80K annually for a firm-scale deployment. The data classification work — labeling which documents are firm-internal vs client-NPI vs public-sourced — is the work the CCO and CIO must do anyway under Reg S-P vendor oversight (L5 Ch3 L2 develops this in depth).

Layer 3 — The Prompt Library

The firm's prompt library is the proprietary asset. The L2 capstone (the 25-prompt advisor library) and the L3 capstone (the practice playbook with 10 workflows) are the starting point; the firm's prompt librarian (L5 Ch4 L1 develops the role) extends and maintains the library against the firm's specific niches.

The library structure: each prompt has a version number, an authoring date, a tested-with-clients-N flag (how many real-client uses informed the iteration), an owner (the senior advisor whose niche the prompt serves), a regulatory-citation footer (the Marketing Rule, Reg BI, FINRA Rule, or NAIC Model the prompt anticipates), a required-disclosure block, and a verification checklist. The library lives inside the RAG vault so the LLM has access to the firm's prompt patterns; the principal-review queue under Rule 2210 reviews the prompts themselves under the FINRA 2026 Annual Regulatory Oversight Report's framing of prompts-as-records under Rule 4511.

The Internal AI Assistant — Assembling the Three Layers

The internal AI assistant — the conversational interface advisors and paraplanners and CSAs use day-to-day — is the user-facing surface of the three-layer architecture. The assembly pattern: the enterprise LLM (layer 1) accesses the RAG vault (layer 2) via the prompt library (layer 3), with the principal-review queue under Rule 2210 + Marketing Rule 206(4)-1 sitting on top of every advisor-facing output before it becomes a client-facing artifact.

The user experience: advisor opens the assistant in the browser or in the CRM (Salesforce FSC, Wealthbox, Redtail) via embedded panel. The assistant presents persona options (junior advisor, senior advisor, compliance reviewer, CCO, estate attorney, business-owner-exit specialist, expat coordinator, philanthropic strategist) — the L3 Ch9 L1 persona engineering pattern extended. The advisor selects the persona, types or pastes the prompt (or selects from the prompt library), and gets output that draws from the RAG vault. The output goes through the principal-review queue under Rule 2210 if it is client-facing.

The output is logged with prompt + retrieved context + LLM response + advisor edits + reviewer signoff. The Rule 4511 retention pipeline absorbs every artifact. The Smarsh or Global Relay archive holds the record. The AI Risk Register tracks any incident under Reg S-P 17 CFR Part 248 IRP.

Case Study — Building the Business-Owner Exit Workflow

A $1.2B RIA in Austin, Texas, with a 40% concentration in business-owner clients, decided in Q2 2025 to build its proprietary business-owner-exit workflow. The team: one senior advisor (the niche owner), one paraplanner, the Chief AI Officer (a former associate advisor promoted into the role per L5 Ch4 L1), the AI Compliance Specialist, an outside-counsel relationship for the QSBS regime classification, and a part-time Pinecone-deployment consultant. Budget: $280K all-in across nine months. Outputs by end of Q1 2026: a 60-prompt sub-library covering pre-sale planning (T-36 to T-12), sale-year tax management, and post-sale stewardship; a cap-table-data ingestion pattern (Carta and Pulley integration); a QSBS per-tranche classification engine handling the OBBBA dual regime; a CRT/installment-sale/opportunity-zone modeling output that the firm's CPA partner consumes; and a NAIC Model #275-compliant annuity-floor recommendation memo for the post-sale liquidity event.

The financial outcome through Q4 2025: 14 business-owner exit engagements run through the proprietary workflow, average revenue per engagement $85K, total revenue $1.19M against $280K build cost — 4.25x payback on the build inside nine months. The buyer-diligence outcome (the firm is not for sale but ran a shadow diligence to test): the proprietary workflow scored 9 of 10 on the L4 Ch8 L1 premium-attribute checklist, supporting a 1.0-1.5x multiple lift at any future exit. The firm's $14M adjusted EBITDA times a 1.0x lift is $14M of incremental enterprise value, supporting the proprietary-build investment by an additional 50x of the original $280K.

The Build-or-Buy Decision Framework

Not every niche workflow justifies a proprietary build. The decision framework is four questions: (1) Is there a shelf vendor producing acceptable output today? If yes, buy. (2) If no shelf vendor today, is one likely to ship in 12-18 months at acceptable quality and price? If yes, wait. (3) If no shelf vendor today and none imminent, does the firm's niche concentration justify a $200K-$500K build amortized over 24 months? If yes, build. (4) If yes to build, can the firm assemble on the three-layer architecture (enterprise LLM + RAG vault + prompt library) without writing custom infrastructure? If yes, build via assembly; if no, partner with a specialty vendor or defer.

The 2026 winning portfolio for a $1B-$5B RIA running this discipline is roughly: 75-85% of AI value from shelf tools (Jump/Zocks/Holistiplan/FP Alpha/Wealth.com/Salesforce FSC + Einstein/Smarsh) deployed via the L4 Ch1-Ch5 strategy, 15-25% from 2-4 proprietary workflows in the firm's defended niches. The discipline to keep the proprietary build to 2-4 workflows — rather than 12 — is the same discipline L5 Ch2 L2 will develop next on piloting, scaling, and killing.

Key Takeaways

• Shelf vendors solved the four high-volume surfaces — meeting AI, tax extraction, estate extraction, CRM AI — and stopped where the addressable advisor population drops below 30,000. The five white-space categories — blended-family legacy planning, business-owner exit, multi-generational philanthropic stewardship, expat tax-and-estate, complex annuity laddering — have no Jump or Zocks in 2026.
• The build architecture for a mid-sized RIA is three-layer assembly, not custom software. Enterprise LLM (Microsoft Copilot or OpenAI Enterprise or Google Gemini Enterprise) + RAG vault (Pinecone, Weaviate, Azure AI Search, or Vertex AI Search) + prompt library — assembled on $50K-$250K of infrastructure cost, not $5M of custom-build cost.
• The RAG vault holds the firm's compliance-blessed content: IPS templates, Reg BI memos, approved client-facing concept memos, WSPs, ADV Part 2A, engagement letter, prior 12-month Marketing Rule audit substantiation, principal-review exception logs (sanitized), and prior 24 months of approved client-facing communications by category.
• The prompt library is the proprietary asset: versioned, dated, tested-with-clients-N flagged, owner-attributed, regulatory-citation footered, required-disclosure embedded, verification-checklist enabled. The principal-review queue under Rule 2210 reviews prompts-as-records under the FINRA 2026 Annual Regulatory Oversight Report's Rule 4511 framing.
• The build-or-buy decision is four questions: shelf vendor today? Shelf vendor imminent? Niche concentration justifies $200K-$500K? Assembly possible on the three-layer architecture? The 2026 winning portfolio is 75-85% shelf-tool value, 15-25% from 2-4 proprietary builds in defended niches.
• The Austin case study: $280K business-owner-exit build returned $1.19M of revenue in nine months (4.25x payback) and scored 9/10 on the L4 Ch8 L1 premium-attribute checklist, supporting a $14M-$21M M&A enterprise value lift on the firm's $14M adjusted EBITDA at a future exit — a 50x return on the build cost in the M&A scenario.