The orthodontist with $720,000 of net practice income, two staff, and a five-year-out retirement horizon does not need a generic SEP-IRA recommendation; she needs an AI-driven model that compares the SEP, the Solo 401(k), the defined benefit plan, and the cash balance plan side-by-side on 2026 contribution maximums, annual administration cost, total tax benefit, plan-document complexity, and post-funding flexibility — and that surfaces the mega-backdoor Roth opportunity hiding inside whichever Solo 401(k) plan document the AI screens for the right after-tax-contribution-and-in-plan-Roth-conversion features. This lesson installs that workflow: from the QuickBooks export and W-2/Schedule C extraction to the four-plan comparative exhibit, to the recommendation memo, to the third-party administrator (TPA) handoff for the actual plan adoption.

The Business-Owner Baseline Pass — What the AI Reads First

Before the four-plan comparison runs, the AI extracts the household's business-and-personal baseline. Inputs: the most recent two years of Schedule C, Schedule K-1 (S-corp or partnership), W-2 (if the owner pays herself a salary through an S-corp), Form 1120-S or Form 1065 (for entity returns), Form 5500 or 5500-EZ if any plan already exists, the QuickBooks or Xero general-ledger export for current-year run-rate, the staff census (count, ages, compensation, hire dates, FTE/PTE status), and the owner's age, anticipated retirement date, current taxable income, marginal federal bracket, state of residence, and existing IRA / 401(k) / Roth balances elsewhere. Holistiplan reads the 1040 line items; FP Alpha reads the K-1; the firm's structured-output prompt against an enterprise LLM (per L2 Ch8 L2) normalizes the staff census and the GL export.

The output is the one-page baseline that drives every subsequent calculation: net self-employment earnings (or net W-2 earnings if S-corp), the §401(c) earned-income computation (net Schedule C minus half of SE tax), the maximum compensation cap for plan purposes ($350,000 for 2026 under §401(a)(17)), the annual additions limit under §415(c) ($70,000 for 2026, $77,500 with the §414(v) catch-up for age 50+, $81,250 with the §414(v)(2)(E) age 60-63 super-catch-up under SECURE 2.0), the employee elective-deferral limit under §402(g) ($23,500 for 2026 with $7,500 catch-up at age 50+ and $11,250 super-catch-up at age 60-63), and the defined-benefit annual benefit limit under §415(b) ($280,000 for 2026). These numbers are the entire grammar of the lesson; the AI's job is to apply them to the client's facts with chain-of-thought transparency (L3 Ch9 L1) so the math is auditable on every line.

The SEP-IRA — The Baseline Compare Everyone Else Beats

The SEP-IRA under §408(k) is the simplest plan: no plan document filed with the IRS, no annual Form 5500 (under most facts), no separate trust, and contributions made entirely by the employer. The 2026 contribution maximum is 25% of compensation (or 20% of net SE earnings for the sole proprietor after the §401(c) computation), capped at the §415(c) limit of $70,000. The owner contributes for herself and is required by the §408(k)(6) coverage rule to contribute the same percentage of compensation for every eligible employee — the employee who worked at least three of the prior five years and earned at least the §408(k)(2)(C) threshold ($750 for 2026).

For the orthodontist with $720,000 of net income (assume $670,000 of net SE earnings after the §401(c) computation) and two long-tenured staff earning $58,000 and $74,000, the SEP-IRA contribution math: 20% of $670,000 = $134,000, capped at $70,000 for the owner; the same 20% rate applied to staff = $11,600 + $14,800 = $26,400 of mandatory employer contributions for the staff. The owner gets a $70,000 contribution and a $96,400 total deduction (her $70,000 plus the $26,400 staff cost) at a 37% marginal rate, for $35,668 of immediate federal tax savings. Simple, fast, no Form 5500.

The SEP-IRA is the baseline; every other plan beats it on contribution capacity at the cost of administrative complexity. The AI's role is not to recommend the SEP-IRA — it is to put the SEP-IRA in the comparative exhibit so the client can see how much the other plans deliver in exchange for their complexity.

The Solo 401(k) and the Mega-Backdoor Roth Hidden Inside

The Solo 401(k) — also called a one-participant 401(k), Individual 401(k), or Uni-K — is the dominant plan for the no-employee or owner-and-spouse-only business. The 2026 contribution math: employee elective deferral of $23,500 (plus $7,500 catch-up at age 50+ or $11,250 super-catch-up at age 60-63 under SECURE 2.0), plus employer profit-sharing contribution of up to 25% of compensation (20% of net SE earnings for the sole proprietor), capped at the §415(c) limit of $70,000 ($77,500 with 50+ catch-up, $81,250 with 60-63 super-catch-up). The structure delivers more contribution capacity than the SEP-IRA at the same income level because the employee elective deferral is on top of the 25% employer share rather than a substitute for part of it.

The Solo 401(k) is also the only plan that can host the mega-backdoor Roth: after-tax employee contributions up to the §415(c) gap between the employee/employer pre-tax contributions and the full $70,000 (or $77,500 / $81,250 with catch-up), followed by an in-plan Roth conversion or an in-service distribution-to-Roth-IRA. The mega-backdoor depends entirely on the plan document permitting (a) after-tax employee contributions and (b) either in-plan Roth conversions or in-service distributions. The AI's job is to screen the prospective Solo 401(k) plan document (or the prototype offered by Schwab, Fidelity, Vanguard, ETrade, Charles Schwab Personal Defined Benefit, or a TPA-administered custom document like Employee Fiduciary, MySolo401k, or Discount Solo 401k) and confirm both features before recommending the structure. The orthodontist would, in this lesson's facts, hit the $70,000 cap on pre-tax + 25%-employer math; the mega-backdoor is irrelevant for her because the cap is already maxed. For the consulting solo with $180,000 of net SE earnings who is under the cap, the mega-backdoor is the lesson — properly structured, it doubles the annual Roth contribution capacity.

The Solo 401(k) With Staff Trap

The Solo 401(k) is only "solo" if the business has no employees other than the owner (and the owner's spouse). The moment an eligible W-2 employee joins, the plan converts to a traditional 401(k) subject to ADP/ACP non-discrimination testing, top-heavy testing under §416, and the full Form 5500 filing burden. For the orthodontist with two staff, the Solo 401(k) is not the plan; she needs a traditional 401(k) with a safe-harbor provision (or a defined benefit / cash balance plan, below). The AI's job is to flag this gating condition before the client signs a Solo 401(k) plan document she will have to terminate within a year.

The Defined Benefit Plan — The Half-Million-Dollar Deduction

The defined benefit (DB) plan is the contribution-capacity heavyweight. Where the Solo 401(k) caps the owner at $70,000-$81,250 of total annual contribution, the DB plan funds a future annual benefit up to the §415(b) limit ($280,000 for 2026, indexed) — and the required annual contribution to fund that benefit, computed by the plan's actuary, can run $150,000 to $400,000+ depending on the owner's age (older owners need less time to fund the benefit, so the actuarial contribution is higher) and the discount-rate assumption. For the 57-year-old orthodontist five years from retirement, a DB plan funding a $250,000-per-year retirement benefit requires an annual contribution in the $310,000-$380,000 range based on current §417(e) interest-rate assumptions and the IRS-prescribed mortality tables.

The trade-off: the DB plan covers staff under the same plan-document benefit formula. The staff annual contributions are computed by the actuary and are typically 5%-15% of staff compensation depending on the benefit formula, which for the two staff at $58,000 and $74,000 translates to $6,000-$18,000 of additional annual cost. The plan document is more complex (a custom plan drafted by an ERISA attorney and a plan administrator), the annual Form 5500 is required, and the actuarial valuation is an annual cost ($3,500-$7,500 for a small-plan TPA). The deduction at a 37% marginal rate on a $340,000 owner contribution is $125,800 of federal tax savings per year — a multiple of the SEP-IRA or Solo 401(k) tax benefit and worth the administrative overhead for the high-income professional with a defined retirement horizon and stable cash flow.

The Cash Balance Plan — The Flexibility Version of the DB

The cash balance plan is a defined benefit plan dressed up to look like a defined contribution plan. The participant has a hypothetical "account balance" credited each year with a pay credit (a percentage of compensation) and an interest credit (a fixed rate, typically 4-5%, or a market-rate proxy). The plan's actual funded status is still measured under DB rules and the actuary computes the required annual contribution, but the participant-facing balance looks like a 401(k) statement, which dramatically improves staff perception and reduces the "where's my account" friction that classical DB plans generate.

For the high-income owner who wants DB-scale contributions but also wants the flexibility to combine the DB with a Solo 401(k) or a 401(k)-with-profit-sharing in the same year (the "DB plus DC" combo), the cash balance is the standard choice. The IRS limits the combined deduction under §404(a)(7), but the structural combo can push total annual contributions for the 50+ owner above $400,000 and the §415(c) DC piece plus the §415(b) DB benefit both stay within their separate caps. For the orthodontist this is the most-likely recommended path: a cash balance plan funding a $250,000 future annual benefit (annual contribution ~$320,000) plus a Safe Harbor 401(k) with profit-sharing for the staff (employee deferral + employer match + profit sharing within §415(c)).

The Three-Year Cash Balance Cliff

The DB and cash balance plans must be funded by the plan sponsor's tax-filing deadline (with extensions), and the IRS expects the plan to be "permanent" — a plan terminated within three years of adoption without a valid business reason (the owner gets sick, the business closes, a major industry change) is subject to retroactive disqualification. The AI's job is to flag the permanence expectation in the recommendation memo so the client doesn't adopt a DB plan as a one-year tax move and then disqualify the entire trust the following year.

The Comparative Exhibit, the TPA Handoff, and the Reg BI File

The deliverable is the four-plan comparative exhibit: one row per plan (SEP-IRA, Solo 401(k), DB, Cash Balance + Safe Harbor 401(k)), columns for owner contribution, staff cost, annual administration cost, federal-and-state tax savings at the client's marginal rate, plan-document complexity score, post-funding flexibility score, and the multi-year cumulative tax savings on a five-year projection. The AI's chain-of-thought output makes every number traceable to a 2026 IRS limit (cited in §401(a)(17), §402(g), §414(v), §415(b), §415(c)) and to a client-specific input (the actual net SE income, the actual staff census, the actual age). The recommendation memo selects one plan, documents the alternatives considered under Reg BI Care, surfaces the conflict of interest (the advisor's AUM-fee or planning-fee implication of a larger funded plan), and produces the TPA handoff package — the plan-document features needed, the actuary-input request, the recordkeeper recommendation, and the staff-notice and Summary Plan Description timing.

The Smarsh or Global Relay archive captures the memo, the prompts, the AI outputs, the human edits, and the signoff trail under FINRA Rule 4511 and SEC Rule 204-2. The CCO or designated principal reviews under FINRA Rule 2210 (the client-facing version) and FINRA Rule 3110 (the supervisory layer). The annual revisit — re-run as Schedule C income shifts, staff turnover changes the census, the actuary updates the funding requirement, and the §415 limits index up — keeps the plan recommendation current and the Reg BI file defensible.

Cross-Tested Profit-Sharing, New Comparability, and the HCE Concentration Problem

The plan-selection conversation gets more interesting when the business has multiple highly-compensated employees (HCEs) and the owner wants to maximize her own allocation relative to the rank-and-file. The §401(a)(4) general nondiscrimination test allows cross-testing (also called "new comparability") that measures benefits rather than contributions for testing purposes. Older owners benefit because the same dollar of contribution translates to a larger projected benefit for them than for younger employees — letting the plan pass non-discrimination while allocating disproportionately to ownership.

For the 58-year-old dental-practice owner with three associate dentists in their 30s and 40s plus seven hygienists and front-office staff in their 20s and 30s, a cross-tested profit-sharing 401(k) with age-weighted allocation can deliver the owner ~$70K of profit-sharing contribution plus the $23.5K employee deferral + $7.5K catch-up while keeping staff cost at ~3-5% of staff compensation. The plan-document cost is meaningful (custom TPA design) but the marginal benefit on top of a Solo 401(k)-equivalent is $30K-$60K of additional annual owner contribution — meaningful but not transformative.

The HCE concentration trap: if the firm's HCE-to-NHCE ratio is unfavorable (many HCEs, few NHCEs), the plan may fail ADP/ACP tests and trigger corrective distributions or QNECs (qualified non-elective contributions to NHCEs). The Safe Harbor 401(k) design (basic, enhanced, or QACA) avoids ADP testing by guaranteeing minimum employer contributions to NHCEs; the QACA variant adds auto-enrollment and accelerated vesting. The AI's plan-document screen confirms the chosen design and the projected staff cost under the design.

Key Takeaways

• The four-plan comparison is one AI-driven exhibit, not four separate analyses. SEP-IRA, Solo 401(k), DB, and Cash Balance + Safe Harbor 401(k) side-by-side on owner contribution, staff cost, admin cost, tax savings, complexity, flexibility, and five-year cumulative tax savings.
• 2026 limits drive the math. §401(a)(17) comp cap $350,000; §402(g) elective deferral $23,500; §414(v) catch-up $7,500 at 50+ and $11,250 super-catch-up at 60-63 under SECURE 2.0; §415(c) DC limit $70,000 ($77,500 / $81,250 with catch-ups); §415(b) DB benefit $280,000.
• The SEP-IRA is the baseline. Simple, no Form 5500 in most cases, mandatory pro-rata for eligible staff under §408(k)(6). Beaten by every other plan on capacity at the cost of complexity.
• The Solo 401(k) hosts the mega-backdoor Roth if the plan document permits after-tax contributions and in-plan Roth conversions or in-service distributions. The AI screens the plan document before the recommendation lands.
• The Solo 401(k) breaks the moment an eligible W-2 employee joins. The plan converts to a traditional 401(k) with ADP/ACP testing, §416 top-heavy testing, and full Form 5500. Flag before adoption.
• The DB plan funds annual benefits up to the §415(b) limit ($280,000). Required annual contributions for the 57-year-old owner often run $310,000-$380,000+. The actuarial valuation is the annual cost and the "permanent plan" expectation is the gating condition.
• The cash balance plan is the DB-with-flexibility and the standard recommendation for high-income owners wanting DB-scale contributions plus a paired DC plan. The combined-deduction limit under §404(a)(7) caps total but the structural pairing pushes total contributions well above $400,000 for 50+ owners.
• One exhibit, one memo, one Reg BI file. The TPA handoff, the Form 5500 schedule, the staff-notice timing, the Marketing Rule disclosure, and the CCO signoff complete the package. Archived under FINRA Rule 4511 and SEC Rule 204-2.