← Audit scenarios

GL Entry Review Audit Trail

Every journal entry reviewed by an AI agent — logged, hashed, and independently verifiable. Segregation of duties violations are auto-blocked. Your data stays in your environment.

The problem

PCAOB AS 2201 amendments require evidence that automated controls over financial reporting operated effectively — including AI agents that review journal entries. SOD violations caught by an agent need an immutable, independently verifiable record.

What gets logged

  • Input fingerprint (SHA-256 of JE reference, account, amount, anonymized poster)
  • Decision with flag subtype and severity (approve, flag, or block)
  • Reliability signals: false positive rate, false negative rate, SOD detection rate
  • Period attestation with SIGMODX-GL verification string

SOD violations

When the same person created and approved an entry, the agent records a segregation of duties violation and the entry is auto-blocked — regardless of agent ALLOW/LIMIT/BLOCK state. The block is hashed and verifiable at /verify.

Integration

from sigmodx import SigmodxClient

client = SigmodxClient(api_key="...", agent_id="...")

input_hash = client.hash_inputs({
    "je_ref": "JE-2026-0441",
    "account": "4100-001",
    "amount": 50000,
    "poster_hash": "abc123"
})

result = client.submit_gl_decision(
    decision_type="flag",
    input_hash=input_hash,
    rationale="Round number entry. Manual review recommended.",
    flag_subtype="round_number",
    flag_severity="medium",
    entry_amount=50000,
    gl_account_code="4100-001"
)

Available for Q3 2026 pilot