L3 Ch5 opens with estate workflow at the document-intake layer. The household uploads will, trust, power of attorney, healthcare directive. The AI extraction tool — Wealth.com or FP Alpha — auto-extracts agents, trustees, beneficiaries, distribution mechanics, GST-tax provisions, and flags the standard gaps. This lesson installs the population-level estate document intake discipline that feeds the L3 Ch5 L2 estate-gap-and-beneficiary-audit memo (next) and the L3 Ch5 L3 advanced vehicle decision tree (lesson after). The deliverable: structured extraction per household with flagged gaps; the next lesson's audit memo coordinates the extracted data with actual account titles and beneficiary forms. L3 Ch5 is the single biggest hidden-risk area in most practices — the estate plan was signed in 2011 and nobody has looked at it since.

The Extraction Tools — Wealth.com and FP Alpha

Two named tools dominate the 2026 estate document extraction category: Wealth.com (focused on estate documents end-to-end with AI extraction and household management) and FP Alpha (broader planning AI with Estate Insights 2.0 as the estate-specific module). Both apply LLM-assisted structured extraction to scanned or uploaded documents (will, trust, POA, healthcare directive). The output: JSON-structured fields per document with the named entities (agents, trustees, beneficiaries, executors, healthcare agents), the distribution mechanics, the GST-tax provisions, the digital-asset provisions (RUFADAA), the incapacity chain.

The L3 Ch5 L1 workflow runs the extraction per household at population scale. For a 200-household practice with 60-90 households having estate documents on file, the extraction batch surfaces document-quality issues (missing pages, scanning artifacts, illegible signatures), document-vintage issues (5+ year old documents flagged for review), and document-completeness issues (will but no trust, trust but no POA, POA but no healthcare directive). The senior advisor or paraplanner runs the batch and the AI extraction surfaces the structured output for each household.

The Extraction Data Fields

The structured-JSON output per estate document per household captures the following field categories.

Will Extraction

Executor (primary + alternate), testator name + capacity attestation, residuary clause (who gets the remainder), specific bequests (named beneficiaries + named property), guardianship provisions (minor children), funeral preferences, no-contest clause, in terrorem provisions, exclusion clauses (heir disinheritance), state-of-domicile + jurisdiction, date of execution + witness attestation, codicil history, anti-lapse provisions, per-stirpes vs per-capita language.

Trust Extraction

Trustee (initial + successors), grantor, trust type (revocable/irrevocable/testamentary), beneficiaries (income + remainder, primary + contingent), distribution mechanics (ascertainable standard / HEMS / discretionary / pure-discretionary), withdrawal-right beneficiaries, special-needs provisions, spendthrift provisions, dynasty provisions (GST allocation, perpetuities), trustee discretion language, distribution standard, accumulation provisions, decanting provisions, trust protector role, GST-exempt vs GST-non-exempt, state-of-situs.

POA Extraction

Power of Attorney type (financial / general / limited / springing / durable), agent (primary + alternate), powers granted (specific or broad), springing trigger event (incapacity definition + medical attestation requirement), revocation provisions, state-of-domicile, date of execution + witness/notary attestation.

Healthcare Directive Extraction

Healthcare agent (primary + alternate), life-prolonging-treatment preferences (DNR, no extraordinary measures, palliative care), specific medical directives (organ donation, autopsy preferences, religious considerations), HIPAA release authority, mental health treatment preferences (some states require separate authorization), state-of-domicile + jurisdiction-specific form.

The Standard Gaps Flagged

The L3 Ch5 L1 workflow surfaces standard gaps that nearly every estate document set has by year 5+ post-execution. The AI's gap-detection capability is one of the workflow's most valuable features.

Missing Successor Trustee

Trust documents from 2011-2015 era often name only the initial trustee (typically the grantor or grantor's spouse). When the trust becomes irrevocable upon grantor's death or incapacity, succession requires named successors. Missing successor leaves the trust without authority to act. The flag surfaces high-priority cases for L3 Ch5 L2 audit memo and attorney handoff.

Stale Beneficiary Designations

The will and trust may reference beneficiaries who have died, divorced, or otherwise had life-event changes that should have triggered beneficiary form updates on individual accounts (IRA, Roth, 401(k), 529, HSA, life insurance, annuity, TOD/POD brokerage). The L3 Ch5 L2 audit memo cross-references trust language against actual current beneficiary forms — gap-detection flags households for that audit.

No Digital-Asset Clause (RUFADAA)

Pre-2018 documents typically lack RUFADAA-style (Revised Uniform Fiduciary Access to Digital Assets Act) language granting executors/trustees access to digital assets (email, social media, online accounts, cryptocurrency). The gap leaves digital-asset access ambiguous — particularly important for cryptocurrency wallets, online business accounts, photos/documents stored in cloud services. Flag triggers attorney update for digital-asset clause.

No Incapacity Provision

Some trust documents define grantor incapacity narrowly (e.g., "permanent unconsciousness for 30 days") leaving the trust without authority during gradual cognitive decline. Modern incapacity provisions specify two-physician concurrence and broader functional-incapacity definitions. The flag identifies households needing incapacity-provision update.

Outdated POA Agent

POA agents named in 2011 may be inappropriate by 2026 (deceased, estranged, no longer trusted, divorced, geographic relocation). The POA itself may be 15+ years old and use outdated language (some banks reject POAs older than 5 years). The flag identifies households for POA renewal.

The GST-Tax Provisions Extraction

The GST (Generation-Skipping Transfer) tax extraction is a particularly nuanced field where the AI must identify GST-allocations made by the trust (which generation receives), GST-exemption status (whether the trust qualifies for GST-exempt treatment), and the implications for grandchildren/skip-person beneficiaries. The 2026 GST exemption amount (~$13.99M individual; ~$27.98M MFJ; pending sunset to ~$7M individual 2026 under prior law and ~$13.99M under post-2025-extension if applicable) drives the GST-tax planning. The L3 Ch5 L3 advanced vehicle decision tree (later lesson) builds on GST planning extensively.

Worked Example: A Full FP Alpha Extraction of the Henderson Revocable Trust

The Hendersons drop a 47-page scanned PDF into the firm's Wealthbox client portal: "Henderson Family Revocable Trust dated March 18, 2014, as amended by First Amendment dated June 2, 2019." The paraplanner routes the document into FP Alpha's Estate Insights 2.0 module. Extraction completes in roughly 90 seconds. The output is a structured JSON object the team reviews against the underlying document. Walking through the fields in order:

Trust identification: Type = revocable inter vivos; grantors = Robert M. Henderson and Margaret L. Henderson (joint); state-of-situs = Illinois; state-of-execution = Illinois; recorded amendments = First Amendment 2019-06-02 (changed successor trustee). Trustee chain: Initial trustees = Robert and Margaret jointly with rights of survivorship; First successor (per 2019 amendment) = daughter Sarah Henderson-Liu; Second successor = son Michael Henderson; Third successor = Northern Trust Company corporate trustee. Beneficiary structure: Income beneficiaries during grantors' lifetimes = grantors; remainder beneficiaries upon death of both grantors = Sarah (40%), Michael (40%), grandchildren per stirpes (20% combined). Distribution standard: The trust grants the successor trustee discretion to distribute principal "for the health, education, maintenance, and support" of remainder beneficiaries — the classic HEMS ascertainable standard under IRC §2041(b)(1)(A), which preserves the beneficiary's ability to serve as trustee without inclusion in the beneficiary's estate. FP Alpha flags this as HEMS and notes the favorable estate-tax characterization. GST allocation: The 2014 trust includes a GST-exempt sub-trust structure ("Grandchildren's Trust") with affirmative GST-exemption allocation language. The 2019 amendment did not modify GST provisions. FP Alpha flags the GST-exempt sub-trust as available for $20K in 2014-era exemption that should be checked against the Hendersons' lifetime GST-exemption usage on Form 709 history. Digital-asset provisions: None. The 2014 document predates Illinois RUFADAA adoption (2016); the 2019 amendment did not add digital-asset language. FP Alpha flags this as a critical gap. Incapacity provisions: Grantor incapacity defined as "written certification by two licensed physicians" — adequate. Trust protector: None named. Decanting: Illinois Trust Code §16.4 default decanting authority applies; no document-level override.

The full extraction surfaces four advisor-action items the L3 Ch5 L2 audit memo will pick up: (1) the missing RUFADAA digital-asset clause — Robert holds a meaningful cryptocurrency position at Coinbase that the successor trustee cannot lawfully access without express authority; (2) the 2014 corporate trustee designation (Northern Trust) should be reconfirmed because trustee fee schedules and minimum thresholds have evolved; (3) Sarah Henderson-Liu's 2019 successor designation precedes her 2023 relocation to Singapore — the practical capacity to serve from Singapore should be discussed; (4) the GST-exempt sub-trust allocation needs reconciliation against the family's Form 709 history at the CPA.

Extraction Error Categories the Reviewer Watches For

FP Alpha and Wealth.com extraction quality has improved materially through 2024-2026, but the reviewer cannot treat the JSON output as ground truth. Five error categories recur and the paraplanner's review explicitly checks each. Boilerplate confabulation. Where a document section uses standard estate-planning language, the extractor occasionally pattern-completes from training data even when the actual document contains a non-standard variant. The fix: every HEMS / spendthrift / dynasty / GST clause is verified against the document's quoted text in the extraction's source-snippet field. Codicil and amendment drift. Where the document set includes a base instrument plus one or more amendments, the extractor may surface the base-document field without applying the amendment's override. The fix: the extraction explicitly identifies all amendments and the reviewer walks each amendment against the affected fields. State-law misalignment. The extractor identifies state-of-execution but may not flag a state-of-domicile mismatch (a Florida resident with an Illinois-executed trust raises situs questions). The fix: the firm's intake captures current state-of-domicile separately and the reviewer compares. OCR artifacts in scanned documents. Older scanned-from-paper documents produce occasional OCR errors that propagate into extracted fields (a named successor "Sarah Henderson-Liu" rendered as "Sarah Henderson-Lui"). The fix: any named-individual field is verified against the document image, not the OCR text. Missing pages. A 47-page upload may be missing pages 23-24 (the signature page or a key distribution-mechanics page) due to scanning error. The extractor produces output anyway. The fix: the reviewer confirms page count and signature-page presence before accepting extraction.

The Population-Scale Extraction Batch and Triage

For the 200-household practice with 75 estate-document-bearing households, the annual extraction batch is structured as a four-week motion. Week 1: upload and pre-processing. The operations team pushes existing document sets into FP Alpha; new uploads from client portal accumulate; the batch is initialized. Week 2: extraction run. All 75 households extract in parallel; the JSON outputs land in a shared workspace with each household tagged. The CCO receives a control-total report (number of documents extracted, extraction quality scores, flagged anomalies). Week 3: paraplanner review. The senior paraplanner reviews each household's extraction against the five error categories, corrects misextractions, and produces a triage tag (Green: no gaps; Yellow: minor gaps to address at next review; Red: high-priority gaps requiring attorney handoff in current cycle). For a typical practice, Green is ~30% of households, Yellow is ~50%, Red is ~20%. Week 4: handoff to L3 Ch5 L2 audit memo + attorney coordination. Red and Yellow households flow into the audit-memo workflow; Red households additionally trigger an attorney handoff coordinated through the firm's preferred estate-attorney panel. The 220 annual estate-review hours from the L3 Ch1 L1 audit compress to ~66 hours through this four-week batch (a 70% leverage), recovering 154 hours for higher-margin work.

The Handoff Diagram and Three-Tier Verification

Source-system: scanned/uploaded estate documents (PDFs typically); extraction-tool output (Wealth.com or FP Alpha JSON); document-image quality verification; document-version verification (is this the most recent edition? are codicils included?); jurisdiction confirmation (state-of-execution may not match state-of-domicile post-relocation). Regulatory: state-specific estate law (UPC vs non-UPC states; community-property vs equitable-distribution states); federal estate tax thresholds (2026 exemption amount; sunset to lower threshold in 2026 if pending legislation applies); GST-tax exemption thresholds; RUFADAA adoption status by state; state-specific healthcare-directive language requirements; HIPAA release standards. Client-fit: documented client preferences from prior meetings (Zocks/Jump transcripts); life-event triggers since document execution (births, deaths, marriages, divorces); IPS estate-planning section; family-structure context; charitable orientation; business-succession considerations; special-needs household status.

The Deliverable and Handoff to L3 Ch5 L2

The L3 Ch5 L1 deliverable per household: (1) structured-JSON extraction of all four documents (will, trust, POA, healthcare directive) with named field capture; (2) the flagged standard gaps with high-priority indicators; (3) document-quality and document-vintage flags (recommended-update timeline); (4) the source-system + regulatory + client-fit checkpoint logs; (5) the senior-advisor or paraplanner review narrative; (6) the artifact archive package in Smarsh with tagged-Markdown metadata.

The L3 Ch5 L2 (next) cross-references the extraction against actual account titles, current beneficiary forms across every account type (custodial brokerage TOD/POD, IRA, Roth, 401(k), 529, HSA, life insurance, annuity), RUFADAA-style digital-asset provisions, and incapacity chain. Together, the L3 Ch5 L1 + L3 Ch5 L2 produce the household-level action-item list and the attorney-handoff memo.

For a 200-household practice with 60-90 estate-document-bearing households, the L3 Ch5 L1 batch extraction is typically the once-per-year deep-review motion (or triggered on client document upload). The L3 Ch5 L2 audit memo follows immediately for any household with flagged gaps. The L3 Ch1 L1 audit's 220-hour annual estate-document review motion (with 0.7 leverage = 154 recoverable hours) is largely captured here through extraction + audit + attorney handoff workflow.

Key Takeaways

• Wealth.com and FP Alpha (Estate Insights 2.0) are the two named extraction tools dominating the 2026 estate-document-AI category. LLM-assisted structured extraction of will / trust / POA / healthcare directive into JSON-structured fields.
• Extraction data fields: Will (executor, residuary, specific bequests, guardianship, no-contest, per-stirpes vs per-capita). Trust (trustee + successors, grantor, type, beneficiaries primary + contingent, distribution mechanics HEMS/discretionary, dynasty/GST provisions, decanting, trust protector). POA (type — financial/durable/springing, agent + alternate, powers, springing trigger). Healthcare Directive (healthcare agent, life-prolonging preferences, HIPAA release, organ donation, jurisdiction-specific).
• Standard gaps flagged: missing successor trustee, stale beneficiary designations (cross-reference to L3 Ch5 L2 audit), no digital-asset clause (RUFADAA pre-2018), no incapacity provision, outdated POA agent (15+ years old; some banks reject pre-5-year POAs).
• GST-tax provisions extraction identifies GST-allocations, exemption status, implications for skip-person beneficiaries. 2026 GST exemption ~$13.99M individual / $27.98M MFJ (pending sunset framework). Drives L3 Ch5 L3 advanced vehicle decision tree.
• L3 Ch5 L1 population scale: typical 200-household practice has 60-90 estate-document-bearing households; batch extraction is once-per-year or triggered on upload; document-quality issues, vintage issues, completeness issues surface for review.
• Three-tier verification: source-system (extraction quality, document version, jurisdiction); regulatory (state estate law UPC/non-UPC, federal estate tax + sunset, GST-tax, RUFADAA state adoption, HIPAA standards); client-fit (prior preferences, life-events, IPS estate section, family structure, charitable orientation, business-succession, special-needs).
• The handoff to L3 Ch5 L2: extraction + flagged gaps feeds the estate-gap-and-beneficiary-audit memo that cross-references trust language against actual account titles, beneficiary forms across all account types, RUFADAA digital-asset provisions, incapacity chain.
• Recovered hours: L3 Ch1 L1 audit's 220-hour annual estate motion × 0.7 leverage = ~154 recoverable hours largely captured through L3 Ch5 L1 extraction + L3 Ch5 L2 audit + attorney handoff workflow. Material practice-economic value plus material client-protection value.