A senior associate at a 200-household RIA in 2022 spent roughly three hours of focused work reading a single complex 1040 — the joint return for a 63-year-old executive with K-1 distributions from a private partnership, ISO and NQSO exercise activity, qualified dividend stacking, a Schedule E rental, an HSA, a backdoor Roth, and a charitable bunching event — and translating it into a tax-planning matrix the senior advisor could walk into the review meeting with. A senior associate at the same firm in 2026 runs the same 1040 through Holistiplan and gets the same matrix in 90 seconds. Trust binders that used to be a half-day read are extracted by Wealth.com or FP Alpha Estate Insights 2.0 in roughly two minutes into a structured map of agents, trustees, beneficiaries, distribution mechanics, GST provisions, and the digital-asset clauses RUFADAA-style provisions require. fpPathfinder's decision-tree library — the 60+ named planning flowcharts — now has an LLM-augmented conversational layer that lets the advisor type "client is 67, $1.8M IRA, married, charitably inclined, just inherited a non-spouse IRA from a sibling" and surface the QCD-vs-RMD-vs-inherited-IRA-10-year-rule decision tree pre-filtered for the situation. The category-changing productivity reveal is real. So is the silent corruption risk where an extraction error two layers upstream of a recommendation propagates through a Reg BI memo, an IPS update, and a client letter without anyone noticing. This lesson maps the four named planning/tax/estate AI tools, the workflows they collapse, and the precise places where extraction errors silently corrupt downstream recommendations.

The Three-Hour-to-90-Second Collapse

The advisor's pre-AI pre-meeting prep workflow for a tax-driven planning conversation was a known time sink. Pull the prior-year 1040 PDF. Extract line by line into the firm's planning template — line 1a wages, line 2a tax-exempt interest, line 2b taxable interest, line 3a qualified dividends, line 3b ordinary dividends, line 4a IRA distributions, line 4b taxable IRA distributions, line 5a pension distributions, line 5b taxable pension distributions, line 6a Social Security, line 6b taxable Social Security, line 7 capital gain or loss, line 8 other income from Schedule 1, line 9 total income, line 10 adjustments from Schedule 1, line 11 AGI, line 12 deductions (standard or itemized via Schedule A), line 13 QBI deduction, line 14 taxable income, line 15 tax (using the appropriate tax-rate schedule and accounting for qualified-dividend / capital-gain stacking), line 16 additional tax, line 17 total tax — and that was just page 1 of the 1040. Then the K-1s from the partnerships, the Schedule E rental activity, the Schedule D capital gain detail with the long-term / short-term split and the Section 1250 unrecaptured gain, the AMT calculation on Form 6251 if the ISO exercises triggered it, the QBI deduction worksheet on Form 8995 or 8995-A, the Form 8606 lines 6-15 for the nondeductible IRA basis and the backdoor-Roth pro-rata math, the Form 8889 for the HSA. Three hours, plus or minus.

Holistiplan's value proposition is that the same workflow takes 90 seconds. Upload the 1040 PDF (and the K-1s, the brokerage 1099s, the W-2s, the state return). The OCR-and-LLM-augmented extraction pulls every relevant line item into a structured tax-planning matrix. The output is the same matrix the senior associate built by hand in 2022, plus the planning-opportunity overlay — the bracket-fill room before the next cliff (24% to 32%, or 22% to 24% depending on filing status and income), the IRMAA two-year-lookback projection for Medicare in two years, the QCD eligibility line for the 70.5+ client, the tax-loss harvest threshold (the $3,000 ordinary offset and the carryforward stack), the charitable bunching opportunity for the standard-deduction client, the Roth conversion window calculation, and the AMT crossover for the ISO-exercise client.

The Holistiplan 10,000-firm install base reported in the 2026 Enterprise Advisory Board announcement is the scale that drives the category benchmark. The product extended into estate planning in 2025-2026, putting it adjacent to Wealth.com and FP Alpha on a subset of the extraction surface; the 2026 Enterprise Advisory Board mandate is to coordinate the cross-tool data flow for the largest enterprise RIA buyers. The L2 Ch4.1 lesson develops the full Holistiplan tax-memo workflow including the QCD election; the L2 Ch4.4 lesson develops the retirement income drawdown memo workflow that consumes the Holistiplan output downstream.

FP Alpha and the Estate Insights 2.0 Benchmark

The estate-extraction problem is structurally harder than the 1040 problem. A 1040 is a standardized form with named lines; the OCR-and-extract pattern is well-defined. A revocable trust is a 30-to-100-page bespoke legal document drafted by an attorney for a specific client, with article and section numbering that varies across attorneys, with defined terms that vary across documents (one trust's "HEMS" provision is in Article 4.2(b), another's is in Article 6 with different language), with embedded references to other documents (the pour-over will, the schedule of beneficiaries, the GST allocation memo), and with conditional logic embedded in the trust language (the marital-deduction formula clause, the QTIP election, the disclaimer trust mechanics). The 2022-era OCR product gave you a searchable PDF. The 2026 LLM-augmented extraction product gives you a structured map.

FP Alpha's Estate Insights 2.0 — the revamped estate planner launched in 2025-2026 — is the category benchmark. The product extracts agents (POA agent, healthcare-directive agent, HIPAA-release contact), trustees (initial, successor, co-trustees, the trustee removal-and-replacement mechanism), beneficiaries (primary, contingent, with percentages, with conditional clauses for predeceased / disclaimer / minor-beneficiary trusts), distribution mechanics (mandatory, discretionary, HEMS-standard, ascertainable-standard, special-purpose), GST-tax provisions (allocation, dynasty mechanics, taxable-distribution triggers), digital-asset clauses (RUFADAA-style provisions), incapacity provisions, and the standard estate-gap flags (missing successor trustee, stale beneficiary, no digital-asset clause, no incapacity provision, outdated POA agent). The output is a structured estate map plus an estate-gap audit and an attorney-handoff memo with the standard "not legal advice" disclosure. The L2 Ch4.3 lesson develops the full FP Alpha workflow; the L3 Ch5 chapter develops the end-to-end estate-and-beneficiary audit pipeline.

Wealth.com as the Estate Extraction Alternative

Wealth.com competes with FP Alpha in the estate-extraction category and overlaps on the core capability: upload will, trust, POA, healthcare directive, get back a structured map plus a gap analysis. The Wealth.com positioning emphasizes the client-facing portal layer — the household-level estate plan visualization the advisor walks the client through in the review meeting — and the integration with the advisor's CRM and planning software. The Wealth.com referral network (advisors using Wealth.com get referrals from clients onboarded through Wealth.com's direct-to-consumer channel) is an adjacent dynamic; the Marketing Rule 206(4)-1(b) testimonial / endorsement mechanics on advisor referrals through Wealth.com need the same disclosure rigor as any third-party referral channel and the firm's WSPs should address. The L1 Ch3.3 lesson covers the prospecting / marketing AI category where the referral-network mechanics live; this lesson focuses on the extraction capability.

The advisor evaluating FP Alpha vs. Wealth.com in 2026 is choosing between two products with substantially overlapping extraction capability and meaningfully different adjacent product surfaces (client portal, referral network, planning-software integrations). The Kitces AdvisorTech map's March 2026 update lists both in the estate AI cluster. The buying decision is firm-specific; the operational discipline is identical — verify the extraction under the Cardinal Rule (L1 Ch2.3) tier-one source-system check before the downstream estate gap memo or attorney handoff goes anywhere.

fpPathfinder Decision Trees Plus the Conversational Layer

fpPathfinder is the category leader for planning decision-tree flowcharts — the 60+ named flowcharts covering Roth conversions, RMDs, Social Security claiming, Medicare IRMAA, beneficiary-designation review, NUA elections, QCD election, 72(t) substantially equal periodic payments, backdoor and mega-backdoor Roth, 529-to-Roth under SECURE 2.0 §126, the inherited IRA 10-year rule, the small-business-owner retirement plan selection, the QSBS Section 1202 dual-regime decision under the post-OBBBA July 2025 framework, the §1014 step-up basis analysis for surviving spouses, and the rest. The pre-2024 fpPathfinder workflow: open the relevant flowchart, walk through the decision nodes by hand, and apply to the client. The 2026 LLM-augmented conversational layer: type the client situation in plain English, get the relevant flowchart pre-filtered with the decision nodes contextually framed for the client.

The example from the lead: "client is 67, $1.8M IRA, married, charitably inclined, just inherited a non-spouse IRA from a sibling." The conversational layer surfaces three relevant flowcharts — the QCD eligibility flowchart (the 70.5+ threshold is approaching but not yet hit), the inherited-IRA 10-year rule flowchart for the sibling inheritance (non-EDB classification under SECURE 2.0 unless the sibling qualifies for an EDB exception; the annual-RMD-during-10-year-window analysis), and the bracket-fill Roth conversion flowchart for the pre-RMD client — with the decision nodes contextually framed. The advisor reviews the framing, applies professional judgment, and selects the recommendation. The fpPathfinder workflow output integrates with the planning software and feeds the Reg BI documentation.

Where Extraction Errors Silently Corrupt Downstream Recommendations

The category-changing productivity reveal of planning/tax/estate AI is real. So is the silent-corruption risk where an extraction error two layers upstream of a recommendation propagates through a Reg BI memo, an IPS update, and a client letter without anyone noticing. The L1 Ch2.2 hallucination lesson catalogs sixteen wealth-specific patterns; the L1 Ch2.3 Cardinal Rule installs the verification protocol; this section maps the planning/tax/estate-specific corruption patterns the protocol is designed to catch.

The Misread 1040 Line

Holistiplan OCR-and-extracts every line of the 1040, but an unusual filing (a Schedule D with a Section 1250 unrecaptured gain, a Schedule E with a passive-loss carryforward, a Form 4972 lump-sum-distribution election, a Form 2210 underpayment penalty calculation) can be silently misread. The downstream Roth conversion recommendation built on an extracted AGI of $148,000 when the actual AGI is $148,400 produces a bracket-fill calculation $400 different from optimal. The error is small per artifact and large per book — across 50 households in a Q4 Roth conversion screen, the cumulative misallocation can run into five figures of long-term client value. The Cardinal Rule tier-one source-system check requires opening the actual 1040 PDF and confirming the extracted lines against the source.

The Misread K-1

K-1s are structurally messier than 1040s — varying line numbers across partnership types, footnote-based disclosures of Section 199A QBI components, embedded references to supplemental schedules, and frequent passive-activity loss complications. Holistiplan's K-1 extraction has improved materially in 2025-2026, but the error rate on complex partnership K-1s remains higher than on 1040s. The downstream consequence is wrong cash-flow assumptions in the RightCapital or eMoney plan, wrong tax-projection numbers in the Reg BI memo, and wrong recommendations on the partnership-distribution timing for the partner-client. Cardinal Rule tier-one check: open the K-1 PDF, confirm the extracted figures, and verify any footnote-based disclosure the model may have missed.

The Misread Trust Language

Wealth.com and FP Alpha extract trust language into structured fields, but the extraction can confabulate provisions that don't exist (the hallucinated Article 4.2(b) HEMS clause discussed in L1 Ch2.2) or miss provisions that do (the disclaimer-trust mechanic embedded in the marital-deduction formula clause). The downstream consequence is wrong estate-gap audit findings and wrong attorney-handoff memos. Cardinal Rule tier-one check: every quoted trust provision in the AI extraction goes through the source document, and the section reference must match the actual document.

The Stale Beneficiary Form vs. the Trust

The classic estate-gap pattern is the trust that says one thing and the beneficiary form on the IRA that says something else — and the beneficiary form controls. The AI extraction reads both documents independently and may flag the inconsistency, but may also miss it if the beneficiary form predates the trust restatement. The downstream consequence is an estate-gap audit that misses the actual gap (the 10-year-stretch IRA bypassing the trust's intended distribution mechanic). Cardinal Rule tier-one check: cross-reference the trust against the actual beneficiary form on every retirement account, brokerage TOD/POD, life insurance, annuity, and 529.

The Missing RUFADAA Digital-Asset Clause

Digital-asset access under the Revised Uniform Fiduciary Access to Digital Assets Act (RUFADAA-style state statutes) requires explicit authority granted by the principal or testator. A pre-2017 trust often lacks the clause. The AI extraction may flag "no digital-asset clause" correctly, but the absence of a flag does not mean the clause exists — the model may have missed it in a non-standard location. Cardinal Rule tier-one check: search the document for the digital-asset language; if not found, the flag stands.

The Incapacity-Chain Gap

The incapacity provision — the POA, the healthcare directive, the trust's incapacity-of-trustee mechanic — is the most-frequently-broken chain in older estate plans. The AI extraction reads each document independently; the consistency check across documents (the POA agent is the same as the healthcare-directive agent is the same as the successor trustee, OR the chain is explicitly tiered) is human work. Cardinal Rule tier-three client-specific check: verify the chain works for this household given the family structure (blended family, adult child as agent, distant successor trustee, etc.).

The fpPathfinder Conversational-Layer Context Error

The fpPathfinder conversational layer surfaces flowcharts based on the natural-language client description. If the description is missing a material fact (the client is also a Form 4 insider with 10b5-1 windows constraining transaction timing; the client has a §409A deferred-comp election in place; the client's spouse is a covered employee under §162(m)), the surfaced flowcharts may miss the binding constraint. Cardinal Rule tier-three check: confirm the client-specific facts that bind the recommendation are in the description AND the flowchart logic absorbs them.

The IRC Section Citation in the AI Output

The planning AI tool's generated memo or client-facing summary cites IRC sections. The L1 Ch2.2 hallucination catalog flagged the §408(d)(6) vs. §408(d)(2) read with §72(e)(8) miscitation for backdoor-Roth pro-rata. The same risk applies to QCD (§408(d)(8), often confused with §408(d)(2)), NUA (§402(e)(4)), RMD age (§401(a)(9), the SECURE 2.0 73-vs-75 trap), 72(t) substantially equal periodic payments, the QSBS §1202 dual-regime issuance-date classification post-OBBBA July 2025, and the §663(b) 65-day election for trust distributions. Cardinal Rule tier-two regulatory check: every IRC citation opens to the actual code text.

The Disclosure and Supervisory Layer for Planning/Tax/Estate AI

The planning/tax/estate AI tools touch client NPI in the most-sensitive form — tax returns, K-1s, trust documents, beneficiary information, household financial detail. Reg S-P 17 CFR Part 248 (with the May 2024 amendments — 30-day breach notification, written IRP, vendor oversight) governs the vendor relationship. The ADV Part 2A disclosure must name the tools and the data categories touched; an off-cycle ADV amendment trigger fires when a new tool is added that materially changes the data-handling profile (L5 Ch7.6 develops the ADV diff workflow). The FINRA Rule 3110 reasonable-design supervisory framework (per the 2026 FINRA Annual Regulatory Oversight Report) extends to the planning/tax/estate AI deployment — written supervisory procedures naming acceptable tools, prohibited data categories, approval workflow, training, and incident response (L4 Ch3 develops the WSP language).

The Marketing Rule 206(4)-1 layer attaches to every AI-generated client-facing artifact downstream of the extraction — the tax-memo client summary, the estate-gap client letter, the Roth conversion recommendation memo, the IPS update narrative, the attorney-handoff cover letter. The 2024-2025 AI-washing enforcement pattern (Delphia, Global Predictions, and the broader 2025 cluster) means that any marketing claim about the tool's AI capability — "our firm uses AI-powered tax planning" — needs the substantiation file and the "clear and prominent" disclosure standard under Rule 206(4)-1(d). The L4 Ch7 chapter develops the full Marketing Rule strategy for an AI practice.

The Downstream Workflow Collapse

The end-to-end downstream workflow when the planning/tax/estate AI extraction is operating cleanly: the senior associate uploads the prior-year 1040, the K-1s, the trust, the will, the POA, and the healthcare directive to the firm's planning AI stack (Holistiplan for tax, Wealth.com or FP Alpha for estate). The extraction produces the structured matrix and the structured estate map within minutes. The fpPathfinder conversational layer surfaces the relevant decision-tree flowcharts. The senior advisor runs the Cardinal Rule three-tier check on the extracted data and the surfaced flowcharts. The Reg BI memo for the Q4 Roth conversion screen drafts from the verified data. The estate-gap audit memo drafts from the verified estate map. The IPS update drafts from the surfaced planning opportunities. The CRM activity log records the pipeline. The Smarsh archive captures the chain. The client-facing summary letters draft from the verified outputs with the Marketing Rule 206(4)-1(d) "clear and prominent" disclosure language baked in.

The pre-AI version of this end-to-end workflow took two-to-three days of associate work. The 2026 AI-augmented version takes a few hours of advisor and associate time, end-to-end. The Schwab 2026 RIA Benchmarking Study's "AI adoption more than doubled vs. 2023" finding captures this collapse in aggregate. The Zocks "10+ hours per week saved" figure is the meeting-AI slice; the planning/tax/estate AI slice is comparable on a different workflow. The L2 Ch4 chapter develops the per-vertical planning workflows; the L3 Ch2 lesson develops the 50-household Roth conversion screen that consumes the Holistiplan output across the book; the L3 Ch5 chapter develops the end-to-end estate audit pipeline.

Key Takeaways

• Holistiplan is the category benchmark for tax extraction with a 10,000-firm install base (per the 2026 Enterprise Advisory Board announcement). The 1040 extraction collapses three hours of associate work into 90 seconds and produces the structured tax-planning matrix plus the planning-opportunity overlay (bracket-fill room, IRMAA two-year-lookback projection, QCD eligibility, tax-loss harvest threshold, charitable bunching, Roth conversion window, AMT crossover).
• FP Alpha's Estate Insights 2.0 is the category benchmark for estate extraction. The product extracts agents, trustees, beneficiaries, distribution mechanics, GST provisions, digital-asset clauses (RUFADAA-style), incapacity provisions, and the standard estate-gap flags. Output: structured estate map plus estate-gap audit plus attorney-handoff memo.
• Wealth.com is the estate-extraction alternative with the client-facing portal layer and the referral-network dynamic (Marketing Rule 206(4)-1(b) testimonial / endorsement implications addressed in firm WSPs and L1 Ch3.3). Operational discipline is identical to FP Alpha.
• fpPathfinder is the planning-flowchart category leader — 60+ named decision trees — now with an LLM-augmented conversational layer that surfaces relevant flowcharts from natural-language client descriptions (QCD, RMD, Social Security claiming, IRMAA, backdoor / mega-backdoor Roth, inherited IRA 10-year rule, NUA, QSBS dual-regime post-OBBBA July 2025, §1014 step-up, and the rest).
• Extraction errors silently corrupt downstream recommendations in eight named patterns: misread 1040 line, misread K-1, misread trust language, stale beneficiary form vs. trust, missing RUFADAA digital-asset clause, incapacity-chain gap across documents, fpPathfinder conversational-layer context error from missing material facts, and IRC section citation hallucination in the AI-generated memo. The L1 Ch2.3 Cardinal Rule three-tier verification protocol is the operational defense.
• Reg S-P 17 CFR Part 248 governs the vendor relationship (May 2024 amendments: 30-day breach notification, written IRP, vendor oversight). ADV Part 2A disclosure names the tools and data categories; an off-cycle amendment fires when a new tool materially changes the data-handling profile (L5 Ch7.6 develops the ADV diff workflow). FINRA Rule 3110 reasonable-design supervision extends to the deployment (L4 Ch3 develops the WSP language).
• Marketing Rule 206(4)-1 attaches to every AI-generated client-facing artifact downstream of the extraction — tax-memo client summary, estate-gap letter, Roth conversion recommendation, IPS update narrative, attorney handoff. The "clear and prominent" disclosure standard under 206(4)-1(d) applies; the 2024-2025 AI-washing enforcement pattern is the structural warning.
• The end-to-end workflow collapses from two-to-three days of associate work to a few hours when planning/tax/estate AI runs cleanly with the Cardinal Rule verification reflex in place. The L2 Ch4 chapter develops the per-vertical workflows; L3 Ch2 develops the 50-household Roth conversion screen; L3 Ch5 develops the end-to-end estate audit pipeline.