The tech-executive client with a four-page grant agreement, a fourteen-tab Carta export, a vesting calendar that crosses two open trading windows, and an AMT exposure that will quietly cost her $180,000 if she exercises the ISOs in the wrong calendar quarter is not, in 2026, a niche planning problem. She is the median client for the meaningful RIAs in Austin, Boulder, Bellevue, Cambridge, and Brooklyn. This lesson installs the AI-assisted exercise-modeling workflow that lets a single advisor handle thirty of her every January — reading every grant tranche, classifying ISO versus NQSO versus RSU versus ESPP, projecting the tax cost of each exercise-and-sale scenario, identifying the AMT crossover, screening Section 1202 QSBS eligibility under the post-OBBBA dual regime, locking the 83(b) window on restricted stock, structuring the 10b5-1(c) plan, and trapping the ESPP disqualifying-disposition error before it shows up on next year's W-2.

The Grant-Agreement Intake Pass — From PDF Stack to Structured Equity Picture

The intake pass is the single highest-leverage AI use in the equity-comp practice. The client arrives with four documents she has never opened: the original 2019 grant agreement for 12,000 ISOs at a $3.40 strike, the 2021 RSU grant for 8,000 units, the 2023 NQSO grant for 4,500 options at $22.10, and the ESPP enrollment summary showing six purchase periods of unsold shares accumulated since 2021. Some of the PDFs are scans. Some are native PDFs with searchable text. The vesting schedules use three different cliff-and-monthly cadences. The strike prices, fair-market-values at grant, current 409A valuations (for the still-private tranches), and current public trading prices (for the post-IPO tranches) sit in five different documents.

Run the stack through FP Alpha's extraction layer, Holistiplan's document module, or a structured-output prompt against an enterprise LLM with the firm's grant-agreement template loaded as the schema. The output is one normalized table: grant date, grant type (ISO / NQSO / RSU / ESPP), share count, strike or grant price, vesting cliff, vesting cadence, vested-to-date, unvested, expiration date, post-termination exercise window, plan-document section reference, and any 10b5-1 plan attached. Verification under the Cardinal Rule (L1 Ch2.3) is non-negotiable: source-system check against the Carta or Shareworks export, regulatory check on plan-document compliance, and client-fit check that the tranches you extracted are in fact the tranches the client owns. The extracted table becomes the input to every downstream modeling step in this lesson; an error here propagates into every Reg BI memo you produce for the next year.

Why Extraction Is the Bottleneck

Before extraction AI was deployable, a senior associate spent six to nine hours producing the same normalized table by hand for a single executive. That cost was the reason most advisors quoted equity-comp planning as an a-la-carte engagement at $4,500-$8,500 — the associate hours dominated the price. With FP Alpha-class extraction collapsing the work to under fifteen minutes and the verification pass adding another twenty, the engagement is now profitable at $1,200 and competitive at $2,400, which is why the equity-comp niche has become the most contested in independent-RIA prospecting for the 2025-2027 hiring cohort exiting tech IPOs and secondary tenders.

ISO Modeling, the AMT Crossover, and the Disqualifying-Disposition Decision

Incentive Stock Options under IRC §422 deliver two specific tax advantages relative to NQSOs and almost every other compensation form: the exercise itself is not a regular-tax event, and a qualifying disposition (stock held more than two years from grant and more than one year from exercise) converts the entire spread plus appreciation into long-term capital gain. The cost of those advantages is the alternative minimum tax. The spread at exercise — strike to fair market value — is an AMT preference item under §56(b)(3). For the 2019 ISO grant of 12,000 shares at a $3.40 strike now trading at $94, the exercise-without-sale event drops $1,087,200 onto the AMT preference line; for the dual-income household with $580,000 of W-2 income, the resulting AMT liability is in the $180,000-$220,000 range depending on the state of residence and the year's AMT exemption phase-out.

The AI workflow here is a forced-show-your-work calculation, executed via chain-of-thought prompting (developed fully in L3 Ch9 L1) against a model running the 2026 AMT formula. Inputs: regular taxable income, ISO bargain element at exercise, state of residence, capital gains and qualified-dividend layering, AMT exemption ($88,100 single / $137,000 MFJ for 2026, with the 28% rate kicking in at $239,100), and the phase-out thresholds. Outputs: regular tax, tentative minimum tax, AMT liability, AMT credit generated under §53 (which becomes recoverable in later regular-tax years), and the "AMT crossover" — the precise share count at which exercising one more ISO no longer reduces total federal tax over the planning horizon.

For the same client, the crossover under 2026 rules typically sits between 35% and 55% of the vested-but-unexercised ISO position depending on baseline income. The recommendation memo, executed via Holistiplan's tax-projection module or RightCapital's tax-projection layer cross-checked against the AI's chain-of-thought output, ranks the four standard ISO exercise patterns: (1) exercise-and-hold for qualifying disposition with the AMT cost financed from outside cash, (2) exercise-and-immediate-sell as a disqualifying disposition (ordinary income on the spread, no AMT, no long-term-gain conversion), (3) staggered exercise across two or three calendar years to ride below the crossover each year, and (4) wait-and-exercise-only-in-a-low-income-year — the year of a sabbatical, a maternity leave, or a job change. The Reg BI Care Obligation under §240.15l-1(a)(2)(ii) requires the file to document the consideration of each.

The AMT Credit Recovery Projection

The AMT credit under §53 is the often-overlooked second half of the ISO calculus. AMT paid in the year of exercise generates a credit usable in later years when regular tax exceeds tentative minimum tax — typically the year of the qualifying-disposition sale, when long-term capital gain flows through the regular-tax track and the AMT track sits idle. A defensible exercise memo projects the credit's recovery path: which calendar year reabsorbs how much credit, the present-value impact, and the worst-case scenario in which the client never recovers the credit because she retires into a low-income state with low future taxable income. The model that pretends the AMT is a sunk cost is the model that makes the recommendation look worse than the math justifies; the model that pretends the AMT credit always recovers is the model that gets a Reg BI complaint when it doesn't. The AI's job is to show both scenarios with named assumptions.

NQSO, RSU, and the Withholding Double-Counting Trap

Non-qualified stock options under §83 and restricted stock units settled at vesting are mechanically simpler than ISOs but produce a different category of error: under-withholding at the corporate-payroll layer that leaves the client owing a six-figure shortfall on April 15. NQSO exercise produces ordinary income on the spread, reported on the W-2 Box 1 with a Code V flag in Box 12, with federal withholding at the supplemental rate (22% on the first $1 million of supplemental wages in 2026, 37% above). RSU vesting produces the same ordinary-income treatment on the fair-market-value of the shares delivered, with the same supplemental withholding rate.

The trap: a 24% marginal-bracket client whose RSU vest pushes her into the 32% or 35% bracket is under-withheld by 8 to 13 percentage points on the entire RSU spread. For a $620,000 RSU vest, the under-withholding is $50,000-$80,000. The AI workflow runs a year-end projection in October of the vesting year, identifies the under-withholding gap, and produces three correction paths: (1) a Q4 estimated-tax payment under §6654 to avoid the underpayment penalty, (2) a year-end W-4 adjustment to apply additional withholding from the client's regular salary (which counts as evenly paid throughout the year, even when concentrated in December), or (3) coordinated capital-loss harvesting against the unrealized losses elsewhere in the household to offset the tax bill. The memo references the prior-year safe harbor (100% of prior-year tax or 110% for high-income clients) and documents the chosen path.

RSUs introduce a second hidden trap: the broker's automatic sell-to-cover rate is almost always the 22% supplemental rate, not the client's actual marginal rate. The post-vest share count after the broker's sell-to-cover is the count the client actually owns, and that count is consistently overstated relative to what the client needs to fund the real tax bill. The advisor's job is to recommend an early supplemental sale — at the same trading window, ideally on the same day to lock the basis — that closes the gap. The AI memo prepares the client-facing email with the gap quantified.

QSBS Section 1202 Screening Under the Post-OBBBA Dual Regime

The single most valuable exclusion in the Internal Revenue Code for early-employee equity in C-corp startups is the qualified small business stock exclusion under §1202. The One Big Beautiful Bill Act signed into law on July 4, 2025 did not repeal the legacy regime; it created a second regime that runs in parallel for stock issued after that date. Every equity-comp client with founder shares, early-employee ISO/NQSO exercises that crossed to stock, or RSUs settled into stock from a private-stage employer needs a per-tranche QSBS screen, and that screen now requires explicit regime classification.

The legacy regime, applicable to stock originally issued on or before July 4, 2025, provides 100% gain exclusion (federal) after a five-year holding period, capped at the greater of $10 million per issuer or 10x the taxpayer's aggregated adjusted basis in the issuer's stock, with the issuing C-corporation required to have gross assets at or below $50 million at the time of issuance. The OBBBA regime, applicable to stock originally issued after July 4, 2025, introduces tiered exclusions: 50% at a three-year hold, 75% at a four-year hold, and 100% at a five-year hold; the lifetime cap rises to the greater of $15 million per issuer or 10x basis; and the corporate gross-asset threshold rises to $75 million at the time of issuance. Both regimes continue to require that the issuer be a domestic C-corporation in a qualified trade or business (excluding most professional services, banking, hospitality, farming, mineral extraction, and any business where the principal asset is the reputation or skill of one or more employees per §1202(e)(3)), and both require that the stock be acquired at original issuance from the corporation (or by gift or inheritance from someone who held qualifying QSBS).

The AI screen: ingest the cap-table or Carta export, classify each stock-issuance tranche by date relative to July 4, 2025, pull the issuer's gross-asset history from the 409A reports or a forensic accounting estimate, confirm trade-or-business qualification, project the hold-period status as of the anticipated sale date, and produce the per-tranche exclusion calculation under the applicable regime. The output is a one-page QSBS exhibit that names each tranche, its regime, its hold status, the projected exclusion percentage, and the lifetime cap consumption. The Reg BI file documents the regime determination — this is the single most-likely future audit trigger as the IRS issues post-OBBBA guidance.

QSBS Stacking and the Non-Grantor Trust Multiplier

Sophisticated planning multiplies the §1202 cap by gifting QSBS to non-grantor trusts before the qualifying sale. Each non-grantor trust is a separate taxpayer for §1202 purposes and gets its own $10 million / $15 million cap per issuer. A founder with a projected $80 million qualifying-stock sale can, with careful pre-sale planning (and attention to the basis-tracking rules and step-transaction doctrine), stack the cap across the grantor, the grantor's spouse, and multiple non-grantor trusts (SLATs, dynasty trusts) created and funded years before the sale. The AI's role is to model the stack and produce the attorney-handoff memo — the legal opinion on each tranche's qualification and the structure of the gifts has to come from a tax attorney with §1202 experience. The line between "AI generates the model" and "AI generates legal advice" is bright and is the line the lesson installs.

83(b) Elections, Restricted Stock, and the 30-Day Window That Cannot Be Reopened

The §83(b) election lets a recipient of restricted property (typically restricted stock issued at founding or in a private financing) include the spread in income at grant rather than at vest, fixing the future appreciation as long-term capital gain. The election is filed with the IRS within thirty days of the property transfer; the window is statutory, the deadline is hard, and the cost of missing it on a tranche that subsequently appreciates 50x is the difference between a $400 ordinary-income inclusion and a $2 million ordinary-income event at vest. The 2025 IRS-published Form 15620 standardized the election format, and the AI workflow takes the grant date, the share count, the grant fair-market-value, the post-vest spread projection, and the holder's facts and produces a draft Form 15620 with the supporting cost-basis attachment, the certified-mail tracking record, and the W-2 / 1099 reconciliation memo for the year of grant.

The advisor's role is to surface the election decision the moment a private-company employee or founder mentions a restricted-stock award, run the comparative model (election vs. no-election under three appreciation scenarios), document the recommendation in the Reg BI file, and queue the 30-day calendar tickler. The AI does the math, the draft, and the calendar; the human owns the decision.

10b5-1 Plans and the Rule 10b5-1(c) Cooling-Off Period

For the executive who sits inside the company's quarterly trading blackout for thirty weeks a year, the Rule 10b5-1(c) plan is the only lawful path to a scheduled sell-down. The 2022 SEC amendments imposed (a) a 120-day cooling-off period between plan adoption and the first trade for officers and directors (30 days for non-officers), (b) prohibition on overlapping 10b5-1 plans for the same class of securities subject to limited exceptions, (c) a limit of one single-trade 10b5-1 plan in any twelve-month period, (d) a certification that the adopter is not aware of material non-public information at the time of adoption, and (e) Form 4 / Schedule 13D disclosure of plan adoption. The 2026 enforcement environment continues to scrutinize plan modifications and the timing of adoptions relative to material announcements.

The AI workflow takes the client's vesting schedule, projected liquidity needs, concentration percentage (cross-referenced to L3 Ch6 L2 on the concentrated-stock decision memo), tax-exposure profile, and corporate trading-window calendar and produces a plan blueprint: tranche-by-tranche sale schedule, cooling-off-period compliance check, projected proceeds, projected tax cost by tranche, integration with any open ISO-exercise plan, and the documentation pack for the corporate insider-trading compliance officer to review and approve. The blueprint is not the 10b5-1 plan itself — that is a legal document executed between the client and the broker-dealer or trustee — but the blueprint compresses what was a four-week back-and-forth into a single working session.

ESPP and the Disqualifying-Disposition Trap That Bites Every January

The Employee Stock Purchase Plan under §423 looks like a small benefit and is, in fact, the most reliably under-reported and mis-classified tax item on the executive's 1040. The qualifying-disposition rule (stock held more than two years from the offering date and more than one year from the purchase date) converts the discount into ordinary income capped at the offering-date discount, with the remaining gain as long-term capital gain. A disqualifying disposition (sale before either holding requirement is met) converts the entire spread between purchase price and sale price into ordinary income reported on the W-2, with any post-purchase loss as a capital loss.

The trap: the executive who participates for six purchase periods at a 15% discount on the lookback rule, never sells, and then liquidates all shares in a single panic sale during a market drawdown — the disqualifying disposition on all six lots produces an ordinary-income event the payroll system never anticipated and the W-2 never captured, because the sale happened after the calendar year of the purchase. The 1099-B from the broker reports a capital-gain-or-loss number that is mechanically wrong for tax purposes (the cost basis reported is the purchase price, not the corrected basis that includes the ordinary-income inclusion). The client and the CPA find out in April, and the under-withholding penalty under §6654 plus the surprise tax bill produce the highest single-issue Net Promoter Score destruction in the equity-comp advisor's book.

The AI workflow runs an annual ESPP holding-status sweep in October: every ESPP tranche the client owns is classified as qualifying, disqualifying-if-sold-now, or qualifying-after-date-X. The memo recommends sales by tranche, identifies the W-2 correction the payroll team needs to make if any disqualifying disposition has already occurred mid-year, and produces the CPA-handoff schedule with the corrected cost basis on each lot. The principal-review queue (L3 Ch10 L2) catches any memo that misses a disqualifying tranche, and the Smarsh archive retains the recommendation under FINRA Rule 4511.

The Integrated Equity-Comp Recommendation Memo and the Reg BI File

The deliverable for an equity-comp engagement is one document: the integrated recommendation memo. It pulls the extracted grant table, the ISO AMT-crossover model, the NQSO/RSU withholding projection, the QSBS per-tranche exclusion exhibit, the 83(b) election determinations on any restricted-stock awards, the 10b5-1 plan blueprint, and the ESPP holding-status sweep into one named-client artifact. It documents the considered alternatives under the Reg BI Care Obligation (the four standard ISO patterns, the three NQSO/RSU withholding paths, the QSBS sale-timing options, the 10b5-1 cadence choices, and the ESPP sale schedule), the client-specific rationale, the fee impact, and the human reviewer signoff. It cross-references the IPS for concentration-policy compliance (L2 Ch5 L3), the estate-vehicle decision if QSBS gifting is in play (L3 Ch5 L3 on SLAT / ILIT / CRT / CLAT / DAF), the charitable strategy if appreciated-stock gifting is the disposition path (L3 Ch8 L1), and the Marketing Rule disclosure language under Rule 206(4)-1 if any AI-generated capability claim appears in the engagement letter or pitch deck.

The Smarsh or Global Relay archive captures the memo, the underlying prompts, the AI outputs, the human edits, and the signoff trail under FINRA Rule 4511 and SEC Rule 204-2. The CCO's principal-review queue under FINRA Rule 2210 handles the client-facing version. The annual update — re-run every January as the new W-2 arrives, the new 1099-B clears, the new grant agreements get filed, and the new vesting events post — keeps the file current for the next exam cycle.

Key Takeaways

• Extraction is the bottleneck and the leverage. FP Alpha or Holistiplan-class extraction converts a six-hour senior-associate intake into a fifteen-minute pass and turns the equity-comp engagement from a $4,500 a-la-carte project into a $1,200 productized service.
• ISO modeling lives or dies on the AMT crossover. The AI's job is to force-show-its-work via chain-of-thought, project the AMT credit recovery under §53, and rank the four standard exercise patterns in the Reg BI file under §240.15l-1(a)(2)(ii).
• RSU and NQSO vesting under-withhold by 8-13 points for the high-bracket client. The October projection plus the Q4 estimated payment, W-4 adjustment, or coordinated loss harvest closes the gap before the §6654 underpayment penalty attaches.
• QSBS §1202 now runs under two coexisting regimes. Legacy (stock issued on or before July 4, 2025): 5-year hold, 100% exclusion, $10M/10x cap, $50M gross-asset threshold. OBBBA (stock issued after July 4, 2025): 50%/75%/100% at 3/4/5-year holds, $15M/10x cap, $75M threshold. The Reg BI file documents the regime determination per tranche.
• The §83(b) election is a 30-day statutory window that does not reopen. Form 15620 (standardized in 2025) is the AI-drafted deliverable; the calendar tickler is the operational safeguard.
• Rule 10b5-1(c) requires a 120-day cooling-off for officers and directors (30 days for non-officers), prohibits overlapping plans, limits single-trade plans to one per twelve months, and demands MNPI-free certification at adoption.
• ESPP disqualifying dispositions are the highest-NPS-destruction event in the equity-comp practice. The October holding-status sweep, the W-2 correction memo, and the corrected-basis CPA handoff are the operational discipline that prevents the April surprise.
• One memo, one Reg BI file, one Smarsh artifact. The integrated recommendation memo is the deliverable; the prompts, outputs, edits, and signoff trail are the record under FINRA Rule 4511 and SEC Rule 204-2.