Finance Architecture & Specs

Institutional-grade platform with an immutable double-entry ledger, OCR receipt extraction, bank-transaction sync across 15,000+ endpoints, multi-signal bank reconciliation scoring, a payroll engine with NACHA direct deposit and paystub generation, tax form generation (W-2/W-3/941/1099/ACA) with IRS FIRE/MeF-format e-file packaging, and asset depreciation (straight-line, declining-balance, MACRS).

System & Performance Metrics

Ledger Security

Strictly Immutable (no edit or delete on finalized entries)

Bank Integrations

15,000+ endpoints

AI Receipt OCR

<3s extraction with schema-validated output

XBRL Compiler

Generates valid US-GAAP-tagged XBRL XML from a visual node graph (SEC EDGAR transmission rolling out via a filing-partner integration)

Payroll Posting

Atomic per pay run — the whole run posts or none of it does

Currencies

160+ with temporal exchange-rate storage

Match Confidence

Weighted multi-signal scoring (amount, date, reference, description) with multi-candidate ranking

1. The Immutable Double-Entry Ledger

The ledger is the mathematical foundation of Aeion Finance — every transaction must balance before it's allowed to post.

Core validation rule: the sum of all debit lines in a transaction must equal the sum of all credit lines. If they don't match, the whole transaction is rejected before anything touches the books — no partial posts, no "fix it later."

Immutability enforcement:

  • No edit or delete operations are permitted on finalized ledger entries — this is enforced at the database level, not just in application code
  • Corrections require a new reversing journal entry
  • Every entry carries who created it, when, and an immutable sequence number
  • The audit trail can't be retroactively altered — not by an admin, not by a bug, not by anyone

Atomic transactions: all journal entry postings execute inside a single database transaction. If any line fails validation, the entire batch rolls back — you never end up with half a journal entry sitting in your books.

2. AI Receipt OCR Pipeline

Vision AI extracts structured financial data from a photographed or uploaded receipt.

Process:

  1. Upload a receipt image (from mobile or desktop)
  2. AI extraction validates the output against a strict schema before anything is accepted
  3. Structured data comes back: merchant, date, total, tax, currency, suggested GL category, and line items

What you get back:

  • Merchant / vendor name
  • Date (normalized)
  • Total amount paid, including tax
  • Tax amount
  • Currency
  • Suggested GL category
  • Line items (description, quantity, amount) where the receipt has them

Integration: extracted data flows directly into the accounts-payable workflow — upload a photo, and a bill is ready to review with the fields already filled in. Works equally well for a production department's on-set purchases as for routine office expenses.

Fallback: if AI extraction is unavailable, the pipeline surfaces an error rather than accepting an unverified receipt — nothing enters the books without either AI verification or a human's sign-off.

3. Bank Connections

Bank connectivity runs through a standard OAuth flow, giving you read access to transactions without ever handling your bank password directly.

Connection flow:

  1. You launch the bank-connect flow from the admin UI
  2. You authenticate directly with your bank (or search + select it) in a secure, hosted window
  3. Aeion receives a long-lived access credential in exchange for a one-time authorization code
  4. That credential is encrypted (AES-256-GCM) before it's ever written to the database

Transaction sync:

  • Transactions are pulled on connection and kept current via real-time webhook push
  • Each transaction is mapped into Aeion's internal transaction record
  • Account data includes name, mask, type, subtype, and available/current balances (tracked in cents for precision)

4. Bank Reconciliation Engine

SignalPointsDescription
Amount40 maxFull points on an exact match; partial credit scaled by how close it is within tolerance
Date30 maxSame day scores highest; partial credit on a sliding scale out to a 3-day window
Reference15Check number or merchant ID cross-referenced against the ledger entry's reference
Description15Significant-word overlap between the bank description and the ledger entry's description

5. Payroll Engine

TypeCalculation
RegularHours up to 40/week at the standard rate
OvertimeHours beyond 40/week at 1.5× the standard rate
Shift DifferentialNight/weekend premiums applied automatically per policy
TaxRateEmployer Match
---------------------------
Federal IncomeProgressive brackets (10–37%)None
Social Security6.2% up to the wage base6.2%
Medicare1.45% (no limit)1.45%
Additional Medicare0.9% over $200kNone
StateVaries by stateVaries
LocalVaries by localityNone
FUTA0.6% up to the wage base
SUTAVaries by state

6. Tax Form Generation

FormDescription
W-2Employee annual earnings & tax withholding
W-3Transmittal of W-2 forms
941Quarterly federal tax return
940Annual FUTA tax return
1099-NECNon-employee compensation
1099-MISCMiscellaneous income
ACA 1094/1095Health coverage reporting

7. XBRL SEC Compiler

Compiles a visual node graph into SEC EDGAR-compliant XBRL XML.

Workflow:

  1. Entity, context (instant/duration period), GL-value, and calculation nodes are wired together on a canvas, with each fact node tagged to a US-GAAP concept (e.g., us-gaap:Assets)
  2. The compiler walks the graph and emits real xbrli:context blocks and <us-gaap:...> fact elements with proper contextRef/unitRef/decimals attributes
  3. The output is a complete, schema-valid XBRL instance document — correct namespaces (xbrli, us-gaap, dei, iso4217), units, and DEI entity-identification facts — ready for an EDGAR filing package

Any US-GAAP concept string can be tagged on a fact node — the compiler doesn't ship a fixed taxonomy library, so coverage is as broad as the tags your CFO or accounting team applies. Direct EDGAR transmission rides the same filing-partner integration as the other e-file formats and is rolling out; the platform never fabricates a submission confirmation it hasn't received.

8. Supporting Capabilities

Asset depreciation:

  • Straight-line: (cost − salvage value) / useful life
  • Declining balance / double-declining balance: book value × (2 / useful life)
  • MACRS: IRS half-year-convention percentage tables across the standard 3/5/7/10/15/20/27.5/39-year property classes, for tax reporting alongside your book-basis method
  • Tracks original cost, salvage value, useful life, accumulated depreciation, and current book value
  • Auto-generates monthly depreciation journal entries

Late fee automation:

  • Configurable penalty schedules per customer or invoice
  • Supports flat fee, percentage, and tiered structures
  • Applies automatically to overdue AR invoices

Recurring invoicing:

  • Configurable schedules (daily, weekly, monthly, custom)
  • Supports subscription, retainer, and milestone billing
  • Auto-generates AR invoices and posts them to the ledger

Fiscal period management:

  • Handles monthly/quarterly/yearly period closing
  • Prevents posting to closed periods
  • Generates closing entries for year-end

Year-end close:

  • Closes all revenue/expense accounts to retained earnings
  • Generates closing journal entries
  • Archives prior-year data

Finance dashboard:

  • Live CFO command-center metrics
  • Cash-flow projections with runway analysis
  • AR/AP aging reports
  • Budget vs. actual variance

Composes with the rest of the OS