Audit Scenarios/Financial Forecast Generation

Financial Forecast Generation Audit Trail

Every revenue forecast, budget projection, and earnings guidance input your AI agent submits, revises, or rejects: logged, hashed, and independently verifiable. Forecasts are scored against post-period actuals, so reliability reflects whether the numbers held up.

Request pilot access Methodology for auditors

AI agents are generating forecasts. Most have no audit trail.

Finance teams use AI agents to draft revenue forecasts, budget projections, and the numbers that feed earnings guidance. When the audit committee or an external auditor asks who produced a material forecast and how it performed against the real outcome, most teams cannot answer with evidence.

Under SOX and the PCAOB's focus on management estimates, a forecast that feeds the financial statements needs a record that survives that question. Sigmodx closes the loop by scoring agent forecasts against post-period actuals.

What gets logged for every decision

The decision

Submit, revise, or reject. Recorded with timestamp, agent ID, forecast type, period, value, and stated confidence interval.

The input fingerprint

A SHA-256 hash of the forecast inputs: period, model version, and references to source data. Raw customer records and proprietary model weights stay in your environment.

The rationale

The agent's stated reason, recorded as an immutable string. "Q3 revenue forecast from weighted pipeline and seasonality model v4. Material to quarterly guidance."

Post-period actuals

When the period closes, the actual value is recorded and the forecast error is computed. Accuracy and calibration error feed the reliability signal automatically.

How the agent's state is determined

Human review and outcomes
Finance reviewers assess submitted forecasts. When the period ends, actuals are recorded. Five signals are computed: forecast accuracy, calibration error, revision rate, materiality flag rate, and human override rate.
State assignment
Forecast accuracy below 80% or revision rate above 25% moves the agent to LIMIT. Above 40% revision rate or 15% human override rate moves it to BLOCK.
State changes are logged
Every state change and the reason for it is written to the append-only log. Material forecasts are routed for immediate review regardless of queue position.

What you give your auditors

Your verification string

SIGMODX-FORECAST-E54ECE50-D331D86614C1

Your auditors submit this string to sigmodx.com/verify. They receive cryptographic confirmation that the audit record for the attested period is intact and unaltered.

They get proof. They don't get your data.

You retain all underlying forecast source records internally and produce them directly to auditors if required. Sigmodx is the proof layer.

Integration is one afternoon

Install the SDK

pip install sigmodx

npm install @sigmodx/sdk

Add three lines to your agent

Direct SDKLangChainMCP

pip install sigmodx

from sigmodx import SigmodxClient

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

input_hash = client.hash_inputs({
    "forecast_period": "2026-Q3",
    "pipeline_value": 4200000,
    "model_version": "rev-forecast-v4",
})

result = client.submit_forecast_decision(
    decision_type="submit",
    input_hash=input_hash,
    rationale="Q3 revenue forecast from weighted pipeline and seasonality.",
    forecast_type="revenue",
    forecast_period="2026-Q3",
    forecast_value=4200000,
    confidence=0.78,
    is_material=True,
)

pip install sigmodx sigmodx-integrations

from sigmodx import SigmodxClient
from sigmodx_integrations import SigmodxCallbackHandler, FORECAST_GENERATION_CONFIG

client = SigmodxClient(api_key="...", agent_id="...")
handler = SigmodxCallbackHandler(
    client=client,
    config=FORECAST_GENERATION_CONFIG,
)

# Every tool call is automatically logged to Sigmodx
agent_executor.invoke(
    {"input": "Generate Q3 2026 revenue forecast"},
    config={"callbacks": [handler]},
)

pip install sigmodx-mcp

SIGMODX_API_KEY=your-key \
SIGMODX_AGENT_ID=your-agent-uuid \
sigmodx-mcp

Claude Desktop / Cursor config:

{
  "mcpServers": {
    "sigmodx": {
      "command": "sigmodx-mcp",
      "env": {
        "SIGMODX_API_KEY": "your-api-key",
        "SIGMODX_AGENT_ID": "your-agent-uuid"
      }
    }
  }
}

Decision events appear in your org dashboard immediately
No infrastructure changes. No vendor access to your source data.

Agent API reference →

Available now

The financial forecast generation scenario is live in production. Pilot access is available for Q3 2026. 90 days, no charge, no contract.

Request pilot access →See how verification works →