The Investment Policy Statement is the single most over-discussed and under-completed document in a 200-household advisory practice. Every textbook, every CFP curriculum, every compliance manual, every Kitces webinar reinforces that the IPS is the foundational governance artifact tying the household's goals to the portfolio's behavior — and yet, in May 2026, a random spot-check of mid-sized RIA client files routinely turns up IPSs that are missing, three years stale, copy-pasted from a template the advisor inherited from a predecessor in 2017, or simply blank. The reason is operational, not philosophical: drafting a defensible, household-specific IPS from scratch is a two-to-four hour exercise that an advisor with 200 households and a calendar of three review meetings a day cannot afford to repeat for every new engagement. This lesson installs the workflow that collapses that cost — from a Zocks or Jump discovery meeting transcript through a structured AI extraction pass to a full first-draft IPS the senior advisor can edit in twelve minutes — without surrendering the fiduciary judgment that makes the IPS defensible under the Investment Advisers Act of 1940 and the firm's ADV Part 2A.

What an IPS Actually Has to Contain in 2026

Before the prompt, the structure. An IPS that survives a 2026 SEC routine examination — and, more importantly, an IPS that actually governs the portfolio across a market drawdown, a life event, and a successor advisor inheriting the relationship — contains, at minimum, eleven sections. Skip any of them and the document is a marketing artifact, not a governance artifact. The eleven sections are: (1) household identification and the decision-makers named on the engagement letter, (2) the investment objective stated in plain English the client can read aloud without translation, (3) the time horizon broken into the planning horizon (life expectancy / joint life expectancy) and the income horizon (years to drawdown, years to legacy), (4) the risk tolerance language captured both qualitatively (the words the client used in discovery) and quantitatively (the standard-deviation tolerance band, the maximum acceptable drawdown, the Monte Carlo success threshold the household has implicitly accepted), (5) the asset allocation policy with explicit policy ranges per asset class (target weight, lower bound, upper bound), (6) the rebalancing thresholds and the rebalancing method (calendar, threshold, hybrid), (7) the tax considerations including the household's marginal bracket, NIIT exposure, state residency, IRMAA proximity, AMT exposure for ISO holders, and tax-loss harvesting policy, (8) ESG / values-aligned constraints including any sector or company exclusions, (9) prohibited holdings including any single-name limits and any inherited legacy positions the client has emotional or tax reasons not to sell, (10) liquidity needs broken into expected (annual distributions, RMDs, planned outflows) and contingent (emergency reserve, medical, capital-call obligations), and (11) the review schedule (annual full review, quarterly portfolio review, triggered review on life events).

Two further sections are optional but increasingly standard in 2026: a separate section on the household's experience with AI-touched workflows (which AI tools the firm uses on this household's data, which it does not, and the cross-reference to the firm's ADV Part 2A disclosure paragraph), and a fiduciary acknowledgment block reciting the advisor's Investment Advisers Act fiduciary duty and the Reg BI obligations applicable to any brokerage account in the household. Neither is required by rule. Both close audit findings before they open.

The Discovery Transcript as the Source of Truth

The traditional IPS workflow assumed the advisor would synthesize the eleven sections from memory, from a Risk Profile questionnaire output, from the planning software (RightCapital, eMoney, MoneyGuidePro) intake, and from whatever the associate captured in the CRM during the discovery meeting. By May 2026, that synthesis chain is obsolete for any practice that has adopted meeting AI. The Zocks or Jump transcript of the discovery meeting is the most complete, time-stamped, verbatim record of what the household actually said about their goals, their fears, their constraints, their values, and the legacy positions they will not sell — and it is therefore the highest-fidelity source for nine of the eleven IPS sections.

The two sections the transcript cannot fully populate are the tax considerations section (which needs Holistiplan's 1040 extraction and the state-residency confirmation) and the asset allocation policy (which needs the firm's investment committee's strategic ranges combined with the household's risk capacity, drawn from the RightCapital / eMoney plan). Both of those are layered in at the editing pass. The discovery transcript drives the first draft.

The mechanical setup before the meeting is simple and matters more than most advisors realize. Configure Jump or Zocks to capture the full meeting, mark the transcript as a discovery meeting (so the post-meeting routing pipeline knows to push it to the IPS-draft workflow rather than to the routine review pipeline), and confirm with the client at the top of the meeting that the recording is happening and that an AI tool will summarize and route action items. The client consent recital takes nine seconds, is required by most state two-party-consent jurisdictions, and is the disclosure the firm will lean on if the Reg S-P amendments' written-policies obligations are tested in a future exam. The meeting then runs as a normal discovery meeting; the IPS draft is a post-meeting artifact, not an in-meeting one.

The System Prompt for the IPS Draft

The prompt is the artifact the firm should standardize, version, and retain under FINRA Rule 4511 and SEC Rule 204-2. A representative working prompt in May 2026, used inside an enterprise LLM (Microsoft Copilot, OpenAI Enterprise, or the vendor-hosted IPS module inside Jump or Zocks), reads roughly:

"You are a senior investment adviser representative drafting a first-draft Investment Policy Statement for the household based on the attached discovery meeting transcript and the firm's IPS template. Produce the draft in eleven sections matching the firm's template. Quote the client verbatim where the client used language about risk tolerance, time horizon, legacy intent, prohibited holdings, ESG constraints, and liquidity needs — and mark each quote with the timestamp from the transcript. Where the transcript is silent on a required section, do not invent content; output 'INSUFFICIENT DISCOVERY — confirm with household' and list the specific questions the senior advisor must ask before the IPS can be finalized. Do not produce specific asset allocation percentages — output the household's risk tolerance and capacity language and flag the section for the investment committee to populate from the firm's strategic ranges. Do not produce tax bracket guidance — output the household's stated tax situation and flag the section for Holistiplan reconciliation. Include the fiduciary acknowledgment block from the firm's IPS template verbatim. Output the draft in Markdown with clear section headings; do not add commentary outside the IPS structure."

That prompt is doing four things at once that matter. First, it instructs the model to quote the client verbatim with timestamps — turning the IPS into an auditable artifact rather than a paraphrase whose fidelity the advisor would have to defend in a later disagreement with the household. Second, it explicitly forbids invention in the sections where the discovery transcript is silent — the LLM's strongest failure mode is plausible-sounding fabrication, and the "INSUFFICIENT DISCOVERY" flag converts a silent failure into a visible work item. Third, it removes the two highest-judgment sections (asset allocation percentages, tax bracket guidance) from the AI's purview entirely, reserving them for the investment committee and Holistiplan respectively. Fourth, it forces structured output (Markdown with the firm's eleven sections), which lets the next workflow step — the Wealthbox / Redtail / Salesforce FSC import — consume the draft cleanly without manual re-keying.

Walking Through a Real Draft — The Hendersons in May 2026

Consider the Hendersons — the recurring household from the L1 lessons. Married, 64 and 62. $2.4M between a traditional IRA, a Roth, a joint brokerage, and a 529 for two grandchildren. The discovery meeting was 53 minutes captured by Zocks. The AI-drafted IPS, run through the prompt above, returns in roughly forty seconds. The opening of the draft reads like this:

The objective section quotes Mrs. Henderson at timestamp 00:08:12: "We want to retire fully when Tom turns 67, travel for ten years, and leave something meaningful for the grandkids' education — not luxury, education." The transcript-derived objective: "Fund full retirement at age 67 for the primary earner and supplemental joint retirement income for the secondary earner; fund a ten-year discretionary travel allocation; fund education legacy for two grandchildren through 529 vehicles and a residual testamentary gift."

The time horizon section quotes Mr. Henderson at timestamp 00:14:45 acknowledging joint life expectancy of approximately 32 years for the surviving spouse based on the planning software's mortality assumption; the income horizon is the three years to Mr. Henderson's age-67 retirement plus the subsequent ten years of high-discretionary travel; the legacy horizon is the residual to the 529 beneficiaries and any testamentary distribution.

The risk tolerance section captures both the qualitative quote ("I lived through 2008 in this account; I do not want to do that again" — Mr. Henderson, 00:22:31) and the explicit AI-flagged item: "INSUFFICIENT DISCOVERY — quantitative risk tolerance not captured. Confirm with household: maximum acceptable single-year drawdown threshold; Monte Carlo success-probability floor; behavioral trigger above which household would request defensive repositioning." The senior advisor reads that flag and knows the follow-up to schedule: a 20-minute risk-tolerance recalibration call before the IPS is finalized.

The ESG / values-aligned constraints section captures Mrs. Henderson at timestamp 00:38:55: "I'd really prefer not to own tobacco companies, and Tom feels strongly about not owning private prison operators." The AI flags it as a hard constraint and adds: "INSUFFICIENT DISCOVERY — household values screening preference not benchmarked to specific index family. Confirm whether household accepts MSCI ESG screened index exposure as proxy for direct exclusion or requires direct holding-level prohibition." That distinction is the difference between a $0 implementation cost (swap to an ESG-screened ETF in the equity sleeve) and a multi-month direct-indexing SMA onboarding. The senior advisor needs to surface the trade-off explicitly before drafting the asset allocation policy.

The prohibited holdings section captures a legacy: Mr. Henderson disclosed at 00:41:08 that his late father's industrial holding — 412 shares of a mid-cap manufacturer — has $87,000 of long-term appreciated basis and that "I'm not selling that one while my mother is alive." The AI converts that to a prohibited-holding entry: "Single-position carve-out: 412 shares [TICKER]; sale prohibited until household instructs otherwise; position excluded from rebalancing logic; concentration risk disclosed and accepted by household at discovery on [date]."

The asset allocation section is correctly left blank with the flag: "POPULATE FROM INVESTMENT COMMITTEE STRATEGIC RANGES — household risk tolerance and capacity language above suggests target range 50/50 to 60/40 equity/fixed; defer to IC strategic range adjusted for ESG constraint and prohibited-holding concentration." The investment committee — or in a solo practice, the advisor wearing the IC hat — completes that section against the firm's strategic ranges in twelve minutes. The advisor's judgment is preserved exactly where it matters; the AI's pattern-matching is leveraged exactly where it cuts forty minutes.

Rebalancing Thresholds and the Orion Reconciliation Handoff

The rebalancing thresholds section deserves explicit attention because it is the section that drives the next lesson (L2 Ch5 L3, IPS-to-Trade Reconciliation Against Orion). The AI-drafted IPS should, per the firm's IPS template, populate the threshold method as either calendar (annual or semi-annual full rebalance regardless of drift), threshold (rebalance any asset class when drift exceeds X percentage points from target, where X is typically 3-5 for major asset classes and 1-2 for satellite sleeves), or hybrid (annual full rebalance plus threshold-triggered interim rebalances on major-class drift greater than the threshold). The choice belongs to the investment committee and the household's tax sensitivity; the AI documents the choice.

The Orion Eclipse interlock matters because the rebalancing thresholds in the IPS become the input parameters for the rebalancer's drift alerts and the trade-ticket generation logic. If the IPS says "rebalance equity sleeve at 5% drift from target," Orion Eclipse needs that threshold loaded into the household's profile. If the household additionally has the prohibited-holding carve-out for the legacy single position, Orion needs that exclusion loaded so the rebalancer doesn't try to sell the position when total equity drifts above target. The IPS-to-Orion handoff is the operational reason the IPS matters in the trading workflow — and the reason a stale or missing IPS produces the rebalancing breaches the next lesson teaches you to detect.

Tax Considerations and the Holistiplan Handoff

The tax considerations section is the second AI-deferred section. The discovery transcript will rarely capture the precise marginal bracket, the NIIT exposure, the state residency nuance for a household with multi-state retirement income (a New York retiree wintering in Florida is a different IPS than a New York retiree staying put), the IRMAA bracket proximity, the AMT exposure for an ISO-holding executive household, or the carryforward losses sitting on the prior-year Schedule D. All of those come from the Holistiplan 1040 extraction (L2 Ch4 L1) and from the household's CPA's prior-year return. The AI-drafted IPS should capture the household's verbal description of their tax situation from the transcript, flag the section for Holistiplan reconciliation, and explicitly leave the bracket-management policy paragraph blank for the senior advisor to populate after the tax workflow has run.

The tax-loss harvesting policy is the one tax sub-section the IPS template should force a binary decision on: does the household authorize tax-loss harvesting in the taxable brokerage account, with the standard wash-sale-avoidance constraints, or does it not. The discovery transcript may or may not surface that conversation; if it doesn't, the AI flags it as INSUFFICIENT DISCOVERY and the senior advisor adds it to the follow-up call agenda.

Liquidity Needs and the Review Schedule

The liquidity needs section is where the discovery transcript typically over-delivers and the AI extraction shines. Clients describe their liquidity needs in narrative form throughout discovery — "we'll need $80,000 next spring for the kitchen renovation," "Tom's quarterly estimated taxes run about $14,000," "I'd like a year of cash sitting aside in case one of us gets sick" — and the AI extracts each as a structured liquidity line item with the dollar amount, the timing (one-time, recurring, contingent), and the funding source (taxable brokerage, money market, HELOC). The senior advisor then assigns each line item a funding source from the household's cash and short-duration sleeves and the section is complete.

The review schedule is the simplest section and the most important to set correctly because it drives the firm's calendar. The default 2026 firm convention is: annual full review meeting (calendar month set by household preference), quarterly portfolio review (CRM-driven activity, no client meeting required absent material event), and triggered review on any life event from the household's update notification or any material market event (defined by the IPS itself as a single-quarter drawdown exceeding the household's stated maximum drawdown tolerance). The AI populates the default and the senior advisor confirms or adjusts.

The Twelve-Minute Editing Pass and the Sign-Off Chain

The AI produces the draft in approximately forty seconds. The senior advisor's editing pass takes approximately twelve minutes when the prompt is well-designed, because most of the senior advisor's time is spent on three discrete activities: (1) reading the AI's verbatim quotes and confirming they reflect what the household actually meant (this is judgment, not extraction, and it is the advisor's fiduciary duty under the Investment Advisers Act), (2) populating the asset allocation policy from the investment committee's strategic ranges, and (3) reviewing each "INSUFFICIENT DISCOVERY" flag and either populating the section from advisor knowledge or routing the question to the follow-up call agenda. The edited draft goes to the household for review (with the AI-tool-use disclosure paragraph standardized in the firm's ADV Part 2A), to the household's signature via DocuSign, and into the Wealthbox / Redtail / Salesforce FSC client record under the IPS document type. The prompt, the AI output, the senior advisor's edits, the household signature, and the signoff date are all retained in Smarsh under Rule 4511 and SEC Rule 204-2; the firm's L4 Ch3 principal-review architecture handles the queueing.

The 90-second client explanation, refined across hundreds of advisor-household conversations and now standardized in firm playbooks: "I'm sending you a draft Investment Policy Statement based on our discovery meeting. We use an AI tool to pull the structure from our meeting transcript so I can focus on your specific situation rather than the formatting — but the policy choices, the allocation ranges, the tax assumptions, and every word about your risk tolerance reflect my judgment as your fiduciary. Please read it carefully, especially the sections where I've quoted you directly, and let me know if any of it doesn't sound like what you meant. Once we agree on it, the IPS governs how your portfolio behaves between our meetings."

Key Takeaways

• An IPS in 2026 contains eleven required sections — household identification, objective, time horizon, risk tolerance, asset allocation policy, rebalancing thresholds, tax considerations, ESG / values-aligned constraints, prohibited holdings, liquidity needs, review schedule — plus two optional-but-standard sections (AI tool use disclosure and fiduciary acknowledgment).
• The Zocks or Jump discovery transcript is the highest-fidelity source for nine of the eleven sections; the asset allocation policy comes from the investment committee's strategic ranges and the tax considerations come from the Holistiplan 1040 extraction.
• The prompt is the artifact — version it, retain it under FINRA Rule 4511 and SEC Rule 204-2, and design it to (1) quote the client verbatim with timestamps, (2) flag silent sections as INSUFFICIENT DISCOVERY rather than inventing content, (3) defer asset allocation percentages to the investment committee, and (4) output structured Markdown the CRM can consume.
• The senior advisor's twelve-minute editing pass is where fiduciary judgment under the Investment Advisers Act of 1940 enters: confirming the verbatim quotes reflect intent, populating the asset allocation ranges, and resolving every INSUFFICIENT DISCOVERY flag.
• The IPS feeds the next two lessons: L2 Ch5 L2 (Update IPS After Life Event) reuses the same prompt structure with new facts; L2 Ch5 L3 (IPS-to-Trade Reconciliation Against Orion) consumes the rebalancing thresholds and prohibited-holdings sections as inputs to the rebalancer's drift logic.
• The household consent and the AI-use disclosure paragraph in ADV Part 2A are the regulatory anchors — without them, the AI-drafted IPS is a Reg S-P risk and a Marketing Rule risk; with them, it is a defensible, time-saving fiduciary artifact.