# SavingsMax Phase II — Tax Prep Module Scoping Addendum

**Author:** Rex (Tax Strategist)
**Date:** 2026-05-23
**Recipient:** Jimmie (primary) → Atlas (schema integration follow-up)
**Parent doc:** `Team Inbox/_Deliverables/Tax Planning Software - Scoping Doc.md` (Ledger, 2026-05-05)
**Status:** Scope-shift addendum; awaiting Jimmie's go/no-go on the business-model change before Atlas drafts schema integration

---

## 0. Why this doc exists

Ledger's 2026-05-05 scoping doc locked **SavingsMax Tax Strategy Planning** as a **planning-only** product:

> "Strictly planning — NO return preparation. No 1040/1120-S/1065 form generation; no IRS e-file integration; no signing as paid preparer. Each strategy hands off to client's CPA or EA for execution."

Today (2026-05-23) you asked about software that lets you **do taxes for your bookkeeping clients while tax planning for them as well throughout the year**. That collapses the planning + prep handoff into one J2-owned workflow. It is a different product, with different regulatory exposure, different software architecture, and different P&L economics.

This addendum scopes the **prep module** that bolts onto SavingsMax. It does **not** rewrite Ledger's planning scope — that work stands. It answers: *what changes when J2 owns the return, not just the strategy memo?*

---

## 1. The strategic case (why this is bigger than a feature)

**Standalone tax shops see the books once a year, in messy form, and prep is reactive.** You already keep books for these clients in QBO every month. That asymmetry is the entire product thesis:

| Lever | Standalone CPA | J2 (books + prep + planning) |
|-------|----------------|------------------------------|
| Intake | 6–8 weeks Jan–Mar of begging clients for docs | 11 months of clean books already on the shelf Jan 1 |
| Book-to-tax M-1 reconciliation | Painful, often surprises | Books were closed monthly with tax treatment in mind |
| Owner comp / QBI optimization | Discovered at year-end (too late to fix) | Monitored monthly, adjusted in real time |
| Estimated payments | Stale projections | Recalc nightly from actuals |
| Planning conversation | Disconnected from books — annual ritual | Continuous loop: every close updates projection |
| Turnaround | March–April | Drafts ready by Feb 15 |
| Client experience | Two relationships (bookkeeper + CPA) | One firm, one quote, one portal |
| Pricing leverage | Hourly or fixed prep fee | Bundle prep into the bookkeeping engagement → higher LTV per client |

**This is the J2 moat.** If we just buy TaxDome + UltraTax and prep returns like a normal CPA shop, none of the above shows up. The whole point of building rather than buying is to compress the books → prep → planning loop into one continuous workflow that no standalone shop can replicate.

If you don't actually want that loop and you'd be fine prepping in a standard tool, **buy TaxDome + Drake/UltraTax and stop here** — building is overkill. The rest of this doc assumes you want the moat.

---

## 2. The regulatory step-change (read this carefully)

Ledger's doc explicitly carved out "no return prep" because the moment you sign a return as paid preparer, three things change:

### 2.1 Practitioner credentials — LOCKED: PTIN + AFSP + CTP
- **PTIN alone is sufficient** to sign federal returns as a paid preparer. You have one. This is the federal floor since *Loving v. IRS* (2014).
- **AFSP (Annual Filing Season Program)** — voluntary IRS program, 18 hrs CE/year, free to join. Adds limited representation rights (revenue agents and customer service, on returns you signed only) and lists you in the IRS Federal Tax Return Preparer Directory. **Join before TY2026 season.**
- **CTP (Certified Tax Planner, AICTP)** — in progress. Industry credential, not IRS-recognized for practice rights, but anchors the SavingsMax planning positioning. This is the credential that differentiates the *strategy* layer; PTIN/AFSP cover the *prep* layer.
- **Practice-rights gap:** PTIN+AFSP cannot represent on IRS Appeals, IRS Counsel matters, or returns not personally signed. For SMB book in TX, this is low-frequency — refer out to a partner EA/CPA on case-by-case when it comes up.
- **State requirements:** Texas does NOT have a state CPA-like preparer registration for federal returns. (CA, NY, OR, MD do — relevant only if you prep returns for residents of those states, which is independent of where you sit.) Most J2 clients are TX → state-side regulatory burden is light.
- **EA path deferred** — not on the table currently; revisit after personal-side prerequisites are resolved.

### 2.2 Malpractice / E&O insurance
- Bookkeeping E&O ≠ tax prep E&O. Different policy class.
- Premium scales with: # returns signed, complexity (1040 vs. 1120-S vs. multi-state), highest single-client revenue exposure.
- Budget $3–8K/year for a 40-client book with mixed entity types. Get quotes before committing.

### 2.3 IRS infrastructure
- **EFIN** (Electronic Filing Identification Number) — required to e-file. Application includes fingerprinting and IRS suitability check. ~45 days.
- **E-Services account** — for transcript pulls, CAF (Centralized Authorization File), 2848/8821 power of attorney workflow.
- **Form 2848 / 8821** — required to talk to IRS about each client. Storage and lifecycle management is a real workflow.

**None of this is exotic — it's the standard onboarding to operate as a paid preparer.** But it's a ~60-day setup plus the EA path if you want to be credentialed (3–6 months). Build can run in parallel.

---

## 3. Build-vs-integrate decision on the prep engine

This is the central architectural call. **Do not let Atlas build a form-generation engine from scratch** — that path is years of work and ongoing regulatory maintenance (form updates every year, every state, every tax law change). The choice is *which third party owns the form mechanics*.

| Option | What it is | Pros | Cons | Fit for J2 |
|--------|-----------|------|------|------------|
| **A. Aiwyn Tax (MCP)** | Tax engine accessible via API/MCP in our environment today | Programmatic prep — drops directly into Atlas workflow. No desktop app. Calls return PDF, jurisdiction list, namespace schema. | Coverage unknown — need to verify business returns (1120-S, 1065), state breadth, e-file capability, pricing. | **High potential if it covers business returns + state + e-file** — this is the cleanest architecture by far. **Needs evaluation before commitment.** |
| **B. ProConnect Tax Online API** (Intuit) | Cloud prep engine, has limited API exposure | Intuit-owned → integrates with QBO data primitives we already know. Full federal + state + e-file. | API is partner-tier; not all functions exposed; per-return pricing $40–80; UI is still Intuit's, not ours. | Medium. Loses some control over UX but eliminates form-mechanics risk entirely. |
| **C. UltraTax / Lacerte / Drake (desktop)** | Industry-standard prep packages | Mature, full coverage, every CPA's known quantity. | **No real API.** Integration = file imports / file exports. Defeats the continuous-loop moat. Desktop license model fights cloud-native architecture. | Low. Falls back to "buy TaxDome and stop" path. |
| **D. TaxDome + Drake (hybrid)** | TaxDome for client portal/workflow, Drake for prep | Fast deploy, well-trodden | Same as C — Drake is the engine, our system becomes a thin wrapper. Limited differentiation. | Low. |
| **E. Native form engine** | Build 1040 + 1120-S + 1065 + states from scratch in Atlas | Total control, total moat | Years of build, perpetual annual maintenance, regulatory risk on form correctness, no e-file path without IRS-approved software certification | **Reject.** |

**Rex's recommendation:** **Evaluate Aiwyn (A) first.** If it covers 1040 + 1120-S + 1065 + relevant states + e-file, that's the architecture. If it's federal/1040-only or no e-file, fall back to **ProConnect API (B)** for the prep engine while we build the surrounding workflow natively in Atlas.

**The evaluation is a 1–2 day task** (Riv can probe the Aiwyn MCP — confirm form coverage, state list, e-file mechanism, pricing per return). It's the gate on every other build decision and should run before Atlas writes a single line of schema for the prep module.

---

## 4. Workflow modules (what the prep side actually does)

Bolts onto SavingsMax's planning engine. Shared client master record, shared entity stack, shared QBO data feed.

### 4.1 Pre-season (Nov–Dec)
- **Tax engagement letter** generated per client (scope, fee, deadlines, e-sign)
- **Document checklist** auto-generated from prior-year return (every form they had last year, plus delta items detected in books)
- **Client portal** request: missing personal-side docs (W-2, 1099, K-1, brokerage 1099-B, mortgage int, etc.)

### 4.2 In-season (Jan–Apr)
- **Auto-draft from books:** P&L → Schedule C / 1120-S / 1065 line items via pre-mapped chart of accounts. Owner W-2 from payroll. Depreciation from fixed asset register. Distributions/draws from balance sheet.
- **Book-to-tax M-1 worktape:** every adjustment shown with source transaction. Reviewer-friendly.
- **Personal-side intake:** OCR + AI extraction on uploaded W-2/1099/K-1/brokerage. Low-confidence fields flagged for human review (this part already specified in Ledger's doc §2.5).
- **Cross-reference engine:** prior-year carryforwards (NOL, capital loss, §179, passive losses, AMT credit) auto-loaded from prior return.
- **Review queue:** every return goes to a 2-stage review. Preparer → reviewer → signer. Three roles even if Jimmie wears all three hats today.
- **E-sign + e-file:** client signs 8879; we transmit; we get acks; we send copies and payment vouchers automatically.

### 4.3 Post-season (May–Dec)
- **Hand the return data BACK to the planning engine** to update next year's baseline.
- **Actual vs. projected variance report** — how good were our projections? Tightens the model over time.
- **Year-round amendment / IRS-notice workflow** (this is real work; clients get CP2000s and we need a place to track them).

### 4.4 Continuous (year-round, the moat)
- **Monthly close → tax projection refresh.** Every Ledger close pushes new actuals; tax engine recalcs nightly. Projection dashboard always current.
- **S-Corp comp drift alerts.** If owner's YTD distribution/W-2 ratio crosses a planned threshold, flag it to Jimmie *while it's still fixable*.
- **Estimated payment auto-adjustment.** Q1/Q2/Q3 vouchers recalculated from actuals, not stale projections.
- **Strategy queue.** Planning items from SavingsMax's strategy library appear as tasks tied to specific deadlines on each client.

---

## 5. Data feeds — what prep needs FROM the books

Most of this Ledger already specified in §2.3 of the parent doc. Delta items for the prep module specifically:

| Data | Cadence | Source | Notes |
|------|---------|--------|-------|
| Trial balance, mapped to tax form lines | Monthly + final | QBO via Coefficient | Chart of accounts must include tax-line mapping (one-time setup per client) |
| Owner W-2 (S-corp), box-by-box | Annual + YTD | Rippling / payroll system | Already inside the Rippling pipeline Riv is building |
| Fixed asset register with current-year additions/dispositions | Annual + as-incurred | QBO + asset detail spreadsheet | This is where §179 / bonus / cost seg elections live |
| Owner distributions vs. compensation (S-corp) | Monthly | QBO equity accounts + payroll | Drives S-corp reasonable comp monitor |
| Loan basis tracking (S-corp shareholder loans, partnership outside basis) | Annual | QBO + manual worktape | Critical for K-1 losses; standalone shops miss this constantly |
| 1099 source data | Annual | Already in Tally's flow | Eliminates a major intake step |
| AP/AR aging detail | Annual at YE | QBO | For cash-basis adjustments if applicable |

**Implication for the books side:** every client's chart of accounts needs a **tax-line mapping** as part of onboarding. This is a Ledger workflow change. Once mapped, prep-from-books is near-automatic. **Without it, you re-do account-to-line classification every January.** Mapping is one-time per client.

---

## 6. Pricing model — needs a rethink

Ledger's doc set SavingsMax planning pricing at **12% of estimated tax savings per strategy**. That model doesn't extend cleanly to return prep, which is a compliance service rather than a value-driven one. Options:

| Model | How it works | Trade-off |
|-------|--------------|-----------|
| **Bundle prep into bookkeeping retainer** | Monthly retainer increases $200–500/mo; prep is "included" | Highest LTV, smoothest revenue. Anchors client to J2 long-term. Compresses standalone-CPA competition. |
| **Flat fee per return** | $800–1,500 for 1040, $1,200–2,500 for 1120-S/1065 | Industry-standard. Easy to quote. |
| **Hybrid** | Bookkeeping retainer covers entity return + owner 1040; complex personal stuff (multi-state, K-1 cascade, RE) billed flat | Most common in the bookkeeping → tax bundle market. Probably right answer. |

**Rex's recommendation:** bundle the entity return + owner 1040 into the bookkeeping retainer (raise the monthly), flat-fee complex add-ons. This makes the "we own the whole stack" pitch real. Defer SavingsMax planning's 12%-of-savings model **for the planning strategies that are above and beyond compliance work** — those are still incremental value, and 12% is still defensible.

This is your call, not mine — but flagging that the prep side needs its own pricing logic.

---

## 7. Schema integration — preview for Atlas

The planning engine schema (Ledger §2) already nailed the heavy lifting (client master, entity stack, financial data, personal data, carryforwards). The prep module adds these tables/extensions:

| Object | What it tracks |
|--------|----------------|
| `tax_engagement` | One per client per tax year. Engagement letter, scope, status (engaged → in-prep → in-review → filed → amended), fee terms, due date. Extends Atlas `projects`. |
| `tax_return` | One per return per engagement. Form type (1040, 1120-S, 1065, state code), filing status, status (draft → reviewed → e-signed → transmitted → accepted → rejected), e-file submission ID, acceptance ack, refund/balance due. |
| `tax_document` | All source docs (W-2, 1099-NEC, 1099-DIV, 1099-B, K-1, brokerage, mortgage int, etc.). OCR extraction fields per type. Confidence scores. Linked to engagement. Extends Atlas media table. |
| `tax_line_mapping` | Chart-of-account-line → tax-form-line. One per client. One-time setup, reused every year. |
| `tax_review_queue` | Two-stage review state machine. Preparer / reviewer / signer. Audit trail. |
| `tax_projection_snapshot` | Time-series projection per client. Lets us show "projection in March vs. October" — proves the continuous-loop value. (Note: this serves BOTH the planning and prep modules — single table.) |
| `irs_authorization` | 2848/8821 per client. Effective dates. CAF tracking. Renewal workflow. |
| `tax_notice` | CP2000s, audit letters, etc. Long-running open-item queue. |

**Existing extensions:**
- `clients` table: add `tax_filing_status`, `state_of_residence`, `dob_taxpayer`, `dob_spouse`, `efin_assigned`, `cpa_handoff` (for clients we don't prep for, just plan).
- `entities` table (per Ledger §2.2): add `tax_year_end`, `entity_election_date` (S-corp), `ein`, `tax_books_method` (cash/accrual).
- `projects` table: tax engagements ride on this; status states extend.

**This pairs with Aiwyn's `tax_namespace_schema` if we go that route — their schema defines what they need; ours defines what we hold. Atlas needs Aiwyn's contract before finalizing.**

---

## 8. Phasing — recommended sequence

Ledger's doc proposed planning-side phasing (Phase 1 federal-only → Phase 4 HNW). For prep, here's the parallel track:

| Phase | Scope | Duration estimate | Gate |
|-------|-------|-------------------|------|
| **0** | Aiwyn evaluation (Riv) + AFSP enrollment (Jimmie) + EFIN application | 2–4 wks | Go/no-go on engine choice; AFSP enrolled; EFIN submitted |
| **1** | Books-to-tax mapping for top 5 S-corp clients + entity return prep (1120-S) using chosen engine, internal-only (no client portal yet) | 6–8 wks | First return drafts in Feb 2027 for TY2026 |
| **2** | Add owner 1040s for those 5 clients + client portal + e-sign + e-file | 6–8 wks | First full bundled returns Mar 2027 |
| **3** | Roll out to all clients in book + multi-state + complex K-1 cascades | TY2027 season | Whole book prep-ready by Jan 2028 |
| **4** | Notice/amendment/audit workflow, advanced HNW (8606, F2555, F3520, etc.) | Ongoing | — |

**Bias toward the "full-architecture-with-gate" model** (per your stated preference): build the data model and engine choice end-to-end before Phase 1 ships, so Phase 2/3 don't require schema rewrites. Your supervision cost: ~weekly check-in with Atlas during Phase 0–1 (likely Nov 2026 – Feb 2027), heavier in Phase 1 first-return reviews (Jan–Feb 2027).

---

## 9. What's automated vs. manual (preempting the obvious question)

**Fully automated (one-time setup):**
- Tax-line mapping per client chart of accounts
- EFIN, 2848, engagement letter templates

**Fully automated (recurring):**
- Monthly close → tax projection refresh
- Entity return draft from books (1120-S / 1065 from clean GL)
- Estimated payment recalculation from YTD actuals
- S-corp comp drift detection
- E-file submission + acknowledgment capture
- Carryforward tracking year-over-year

**Manual (recurring, can't be automated):**
- Personal-side doc review (OCR will be 90% — the last 10% is judgment)
- Two-stage review of every return (this is the credential layer — software can't sign for you)
- Client communication on complex items
- IRS notice responses

**Manual (rare):**
- Aiwyn schema updates if they rev the API
- Annual rule-config updates (brackets, contribution limits — Ledger flagged this)

**Tools ruled out (and why):**
- Drake/UltraTax/Lacerte: no API path → moat-destroying
- Native form engine: build cost too high, regulatory risk too high
- TaxDome standalone: it's a workflow shell, not a prep engine — doesn't replace this build

---

## 10. Credential decision — RESOLVED 2026-05-23

**Locked stack: PTIN + AFSP + CTP (in progress). Jimmie is sole signer.**

Implications for the build:
- `tax_review_queue` is single-signer — no external partner handoff in the loop. Simpler state machine.
- Malpractice insurance class is non-credentialed paid preparer + planning advisory. Get quotes during Phase 0.
- First signed return possible TY2026 season (no exam timeline gating).
- IRS Appeals / non-signed-return representation refers out case-by-case to a partner EA/CPA (low frequency for TX SMB book; not a structural blocker).

**EA path is deferred** — revisit after personal-side prerequisites are resolved. Do not plan around it.

---

## 11. Next steps

1. **Jimmie** — AFSP enrollment + EFIN application kickoff (parallel to engine evaluation).
2. **Riv** — Aiwyn Tax MCP evaluation (form coverage, state list, e-file path, pricing). 1–2 days.
3. **Ledger** — Tax-line mapping template for the QBO chart of accounts. Pilot on Afton Electric.
4. **Atlas** — Schema integration design draft (extends SavingsMax planning schema with §7 of this doc); `tax_review_queue` modeled as single-signer.
5. **Pixel** — SavingsMax brand kit (already flagged in Ledger §9.3) extended to prep deliverables (engagement letter, return cover sheet, e-sign templates).

---

**End of addendum.**
**Parent doc remains canonical for the planning side. This doc adds the prep module on top.**