The eight prior L2 lessons all produced an artifact: an IPS draft, an IPS update redline, a reconciliation breach report, a quarterly commentary, a personalized concept memo, a difficult-conversation message, an onboarding checklist, a Reg BI rollover memo. Each artifact was useful to the senior advisor at the moment it was produced. The decisive operational question — the one that separates "advisor uses AI" from "practice uses AI" — is what happens to those artifacts in the five minutes after the senior advisor closes the laptop. Does each artifact get manually re-keyed into Wealthbox by a paraplanner? Does the trade ticket get manually typed into Orion Eclipse? Does the Reg BI rationale paragraph get manually copy-pasted into the household record? Or does each artifact arrive in the firm's CRM, planning software, trading platform, and compliance archive in a shape the downstream systems can ingest cleanly, without re-keying, without transcription error, without losing the source-data fidelity, and without the senior advisor's twelve-minute editing pass being followed by a paraplanner's forty-five-minute re-entry cycle? This lesson installs the discipline of structured output — JSON, tables, Markdown, schema-conforming — that lets every prior L2 lesson's artifact land where it needs to without human re-keying.

Why Structured Output Is the "Practice Uses AI" Discipline

An AI that produces prose paragraphs delivers value once — when the senior advisor reads them. An AI that produces structured output (JSON conforming to the CRM's import schema, a Markdown table the planning software's import tool consumes, a tagged-paragraph artifact that the firm's Smarsh archive routes correctly by document type) delivers value across the entire workflow chain: senior advisor reads it, paraplanner imports it, CRM stores it, planning software references it, trading platform consumes the rebalance instructions, compliance archive retains it with correct metadata, the next quarterly review pulls it back as context, the M&A buyer's diligence team finds it cleanly. The structured-output discipline is the operational difference between a tool that saves the senior advisor twelve minutes and a workflow that saves the practice nine person-hours per onboarding.

The 2026 advisor practice convention treats every AI-generated artifact as a candidate for structured output. The questions for each artifact: (a) which downstream systems consume it (Wealthbox / Redtail / Salesforce FSC for CRM; RightCapital / eMoney / MoneyGuidePro for planning software notes; Orion Eclipse / 55ip / Aladdin for trading; Smarsh / Global Relay for archive; the Reg BI evidence log); (b) what schema do those systems expect (CRM custom-field structure, planning-software note tagging, trading-platform rebalance-instruction format, archive document-type metadata); (c) how does the AI's output map to the schema (JSON object with the right keys, Markdown table with the right column headers, tagged content blocks the import tools can route).

JSON for CRM Custom-Field Imports — Wealthbox, Redtail, Salesforce FSC

The three dominant 2026 wealth-management CRMs each expose custom-field structures the AI's output can populate. Wealthbox's API and custom-field model, Redtail Engage's structured fields and CRM activity log, and Salesforce Financial Services Cloud (FSC) with Einstein's integration capabilities all accept JSON-structured input that maps to specific CRM records and activities. The discipline: the AI's prompt requires structured-JSON output conforming to the firm's documented CRM custom-field schema, with each artifact type producing a JSON object with predictable, mappable fields.

IPS Draft JSON Output Example

The L2 Ch5 L1 IPS draft output becomes the firm-defined JSON object: { "household_id": "[CRM-record-id]", "ips_edition": 1, "draft_date": "2026-05-21", "objective": "[verbatim from prompt output with timestamp citations]", "time_horizon_planning": "32 years joint life expectancy per RightCapital", "time_horizon_income": "3 years to retirement + 10 years discretionary travel", "risk_tolerance_qualitative": "[verbatim client quote with timestamp]", "risk_tolerance_quantitative": "INSUFFICIENT DISCOVERY — flag for follow-up", "asset_allocation_policy": "IC REVIEW REQUIRED — IPS template ranges 50/50 to 60/40", "rebalancing_thresholds": "hybrid: annual full + 5% drift threshold", "tax_considerations": "HOLISTIPLAN RECONCILIATION REQUIRED — household-stated context [transcript timestamp]", "esg_constraints": "no tobacco, no private prisons; direct-holding-only", "prohibited_holdings": [{"ticker": "[carved-out]", "shares": 412, "rationale": "legacy from father", "concentration_disclosure": "accepted by household 2026-05-15"}], "liquidity_needs_expected": [...], "liquidity_needs_contingent": [...], "review_schedule": "annual full + quarterly portfolio + triggered on life events", "ai_tool_disclosure": "drafted with [vendor/model version] under firm ADV Part 2A", "fiduciary_acknowledgment": "[firm template paragraph]" }

The Wealthbox import tool reads the JSON, populates the household's custom fields, creates the IPS document record, links it to the household record, sets the next-review-due field based on the review schedule, and flags the INSUFFICIENT DISCOVERY and IC REVIEW REQUIRED items as advisor follow-up tasks. The paraplanner's manual re-entry time drops from forty-five minutes to zero; the senior advisor's edits flow through the same JSON updates to the same downstream systems.

Reg BI Rollover Memo JSON Output Example

The L2 Ch7 L2 rollover memo output becomes: { "household_id": "...", "memo_date": "2026-05-21", "rollover_type": "401(k) to Traditional IRA", "household_age": 58, "household_72t_status": "below_59.5", "plan_balance": 1200000, "alternative_1_leave_in_plan": { "evaluated": true, "considerations": [...], "recommendation": "rejected", "rationale": "..." }, "alternative_2_new_employer_plan": { "evaluated": true, "applicability": "not_applicable", "rationale": "..." }, "alternative_3_roll_to_ira": { "evaluated": true, "considerations": [...], "recommendation": "accepted", "rationale": "..." }, "alternative_4_take_cash": { "evaluated": true, "tax_cost_analysis": {...}, "recommendation": "rejected", "rationale": "..." }, "fee_comparison": { "plan_share_class_avg_bps": 8, "ira_share_class_avg_bps": 38, "horizon_years": 25, "lifetime_cost_differential_estimate": "..." }, "fund_comparison": "...", "surrender_penalty_analysis": "no §72(t) trigger; direct rollover mechanism; no plan surrender charges", "net_benefit_narrative": "...", "recommendation": "Roll to Traditional IRA at [firm] Custodian", "conflict_disclosure": "standing AUM-fee structure per ADV Part 2A and engagement letter", "regbi_citation": "§240.15l-1(a)(2)(ii) Care Obligation", "reviewer_signoff": "[senior advisor name + CCO name + date]" }

The structured object lets the firm's Reg BI evidence log (typically a Wealthbox custom-object, a Redtail custom field set, or a Salesforce FSC custom record type) capture the full memo with structured queryability. The CCO's quarterly review can pull all rollover memos with "fee_comparison.ira_share_class_avg_bps > 50" to triage the highest-fee-differential cases. The M&A buyer's diligence team can query the rollover memo dataset to confirm Reg BI documentation completeness across the book.

Markdown Tables for Planning Software Notes — RightCapital, eMoney, MoneyGuidePro

Planning software platforms (RightCapital, eMoney, MoneyGuidePro) accept structured note content with embedded tables. The AI's output for planning workflows (Roth conversion analysis, RMD calendar, Social Security claiming analysis, IRMAA coordination) produces Markdown tables that the planning software's note-import tool consumes cleanly. Example for a Roth conversion analysis:

| Year | Pre-Conversion AGI | Recommended Conversion | Post-Conversion AGI | Federal Bracket | IRMAA Bracket | Conversion Tax | Lifetime LT Benefit |
|------|---------------------|------------------------|---------------------|-----------------|---------------|----------------|---------------------|
| 2026 | $148,000 | $96,000 | $244,000 | 24% | Tier 2 | $23,040 | $58,000 |
| 2027 | $151,000 | $94,000 | $245,000 | 24% | Tier 2 | $22,560 | $54,000 |
| 2028 | $155,000 | $90,000 | $245,000 | 24% | Tier 2 | $21,600 | $50,000 |

The planning software's note-import tool parses the Markdown table, populates the household's planning scenario with the year-by-year conversion plan, links the scenario to the senior advisor's Reg BI rationale memo, and triggers the next-review calendar. The structured-table format means the household-facing concept memo, the planning software scenario, the trading-platform conversion-execution instructions, and the Smarsh archive all draw from the same source of truth without manual reconciliation.

Tagged Markdown for Archive Routing — Smarsh, Global Relay

Smarsh and Global Relay accept tagged-Markdown documents with metadata blocks that route the document to the correct retention bucket, apply the correct retention policy, attach the correct supervisory-review chain, and surface the document during future searches. The AI's output for compliance-bearing artifacts includes a metadata block at the top:

---
document_type: reg_bi_rollover_memo
household_id: [...]
senior_advisor: [...]
ccoreviewer: [...]
retention_policy: rule_4511_5yr_practical
created_date: 2026-05-21
ipsedition_referenced: 1
ai_tool: [vendor/model_version]
ai_prompt_version: senior_advisor_v3.2
related_records: [household_id, ips_edition_1, discovery_transcript_id]
---

The Smarsh / Global Relay archive ingests the tagged document, applies the rule_4511_5yr_practical retention policy, links to the related records, and surfaces the document during any future exam or M&A diligence query. The L4 Ch3 supervisory architecture's principal-review queue uses the metadata tags to route the artifact correctly. The tagged-metadata discipline scales document management from hundreds of artifacts per year to thousands per month without operational degradation.

Structured Output for the Reg BI Evidence Log

The Reg BI evidence log is the practice's contemporaneous documentation of every recommendation, the alternatives considered, the household-best-interest rationale, the conflict disclosure, and the signoff chain. Structured output lets the evidence log function as a queryable database rather than a paper trail. A 200-household firm in active practice generates hundreds of recommendations per quarter — IPS-aligned rebalance recommendations, Roth conversion recommendations, rollover recommendations, beneficiary recommendations, allocation recommendations following life events. Each one is a Reg BI-documented artifact. The evidence log's structured queryability enables: (a) the CCO's quarterly sampling for principal review, (b) the L4 Ch5 ROI dashboard's per-recommendation completeness metrics, (c) the next quarterly review pulling the household's prior recommendations as context, (d) the M&A buyer's diligence query confirming Reg BI documentation completeness across the book, (e) the exam-defensibility query showing all recommendations for any specific period.

The Prompt Discipline — Schema as Constraint

The prompt that produces structured output is itself a discipline. Representative working prompt for the IPS draft (extending L2 Ch5 L1): "... Output a JSON object conforming to the firm's IPS schema (version 3.1): { "household_id": [from CRM record], "ips_edition": [integer], "draft_date": [ISO 8601], "objective": [string], "time_horizon_planning": [string], ... }. Use the verbatim client quotes from the transcript with timestamp citations in the appropriate fields. Use the INSUFFICIENT DISCOVERY flag value where the transcript is silent on a required field. Use the IC REVIEW REQUIRED flag value for asset_allocation_policy. Use the HOLISTIPLAN RECONCILIATION REQUIRED flag for tax_considerations bracket specifics. Output ONLY the JSON object — no prose commentary outside the object. Validate the JSON for schema compliance before returning."

The schema-as-constraint discipline produces output that is (a) deterministically importable into downstream systems, (b) versionable as the schema evolves, (c) validatable for completeness and correctness before downstream processing, (d) auditable as part of the firm's Reg BI evidence log, (e) substantiable in an exam because the schema demonstrates the firm's documented operational discipline.

Schema Versioning and the Firm's Operational Evolution

The firm's CRM custom-field schemas, planning-software note structures, archive metadata vocabularies, and Reg BI evidence-log schemas all evolve. The firm versions each schema (CRM-field-schema v3.1, planning-note-schema v2.4, archive-metadata-schema v1.7, regbi-log-schema v4.2) with documented update rationale and effective dates. The system prompts (L2 Ch8 L1) reference the schema versions; the few-shot exemplars reference schema versions; the prior-version artifacts retain their schema-version metadata for backward-traceability.

The 2026 convention: schema changes go through a CCO-reviewed change-control process; the change-control documentation is itself part of the Smarsh archive; the L4 Ch3 supervisory architecture handles the schema-change rollout; the L4 Ch5 ROI dashboard tracks schema-change frequency and downstream-system integration health. Schema versions and prompt versions are linked artifacts in the firm's operational asset library.

Cross-System Integration — Wealthbox to RightCapital to Orion Eclipse to Smarsh

The practical operational power of structured output is the end-to-end workflow integration it enables. A household life event triggers an IPS update workflow (L2 Ch5 L2); the AI produces a JSON-structured IPS update output; Wealthbox ingests the household-update record; RightCapital's note-import tool ingests the planning-context update; Orion Eclipse's profile-update API ingests the rebalancing-threshold and prohibited-holdings changes; Smarsh ingests the tagged-Markdown signed IPS document with full metadata; the L4 Ch3 pre-use review queue receives the artifacts for principal review; the L4 Ch5 ROI dashboard captures the workflow-completion metric. End-to-end: from discovery-update transcript to firmwide system update in approximately 30 minutes of advisor time (senior advisor editing pass + signoff) plus automated downstream propagation.

Without structured output: the same workflow takes 4-6 paraplanner hours of re-keying across systems, introduces transcription errors at each handoff, breaks the source-data fidelity chain, undermines the Reg BI evidence log, and produces the firm's scattered AI use rather than practice-integrated AI use.

The Household-Facing Implication and the Strategic Implication

The household-facing framing about structured output is implicit rather than explicit — households don't care about JSON. But they do care about (a) consistent communication that reflects their actual relationship across all touchpoints, (b) fast turnaround on questions and decisions because the firm's systems are integrated, (c) accuracy in the firm's records about their household, (d) continuity through advisor turnover because the firm's systems retain the relationship knowledge structurally. Structured output enables all four.

The strategic implication for the practice is the L4 Ch8 valuation framework. A 200-household firm with structured AI workflow integration is worth materially more in an M&A transaction than a firm with equivalent AUM but scattered tool use, because the buyer can integrate the acquired book without rebuilding the operational workflow, can demonstrate Reg BI documentation completeness to their CCO and regulators, can leverage the codified house voice and few-shot library, and can plug the schema-conforming workflow into their existing operational stack. The premium-tier multiple (top-quartile RIAs roughly 8x-10x adjusted EBITDA per Mercer Capital and ECHELON Q3-Q4 2025 data, with the highest-rated transactions reaching ~11.6x at the premium top) depends on demonstrable operational maturity — and structured output is the operational maturity signal at the workflow integration layer.

Key Takeaways

• Structured output is the operational difference between "advisor uses AI" and "practice uses AI" — JSON for CRM custom-field imports, Markdown tables for planning-software notes, tagged-Markdown for archive routing, schema-conforming output for the Reg BI evidence log.
• The three dominant 2026 CRMs each accept JSON-structured input: Wealthbox via API and custom-field model, Redtail Engage via structured fields and CRM activity log, Salesforce Financial Services Cloud + Einstein via custom-object integration.
• Planning software (RightCapital, eMoney, MoneyGuidePro) accepts Markdown tables for year-by-year conversion analyses, RMD calendars, Social Security claiming scenarios, IRMAA coordination plans.
• Smarsh and Global Relay accept tagged-Markdown documents with metadata blocks that route documents to correct retention buckets, apply correct retention policies, attach supervisory-review chains, and enable search and surfacing.
• The Reg BI evidence log functions as a queryable database when populated by structured-output AI workflows — enables CCO sampling, ROI dashboard metrics, prior-recommendation context retrieval, M&A diligence queries, exam-defensibility queries.
• Schema versioning is the discipline that lets the firm evolve operationally while preserving backward-traceability; schemas, prompts, and few-shot exemplars are linked artifacts in the firm's operational asset library; CCO-reviewed change-control process under L4 Ch3.
• End-to-end cross-system integration (Wealthbox to RightCapital to Orion Eclipse to Smarsh) compresses a 4-6 paraplanner-hour re-keying workflow into 30 minutes of advisor time plus automated downstream propagation; the L4 Ch8 valuation framework treats this operational maturity as a premium-tier multiple driver.