Finance without the grind.

What manual work costs

One approach. Every system.

Two Planning Paths, One Deterministic Core

Every question is routed before a single row is touched. The router’s only job is deciding which planning path to use:

# Specialized Financial Path
def route(question, dataset):
    if dataset.is_sample:
        # LLM planner emits specialized FinancialAnalysisPlan
        return FinancialAnalysisPlan(
            metric="operating_expense",
            breakdown="cost_center",
            time_dimension="Q3_2026",
            calculations=["variance", "driver_waterfall"]
        )

Supported Analysis Types (Both Paths)

• total
• breakdown
• ranking
• trend
• variance
• driver_analysis
• waterfall
• completeness

The result: You can ask questions against your Postgres GL today, swap to Snowflake in six months, and the same question works without reconfiguration. Finance data is data; the analysis stays the same.

Columns labelled for you.

DimensionSelecting Dimension makes a column eligible for GROUP BY — e.g. "variance by cost_center" groups every row by this field.

The Audit Trail Every CFO Demands

SELECT cost_center,
       SUM(actual_amount) AS actuals,
       SUM(budget_amount) AS budget,
       SUM(actual_amount - budget_amount) AS variance
FROM general_ledger_entries
WHERE fiscal_quarter = 'Q3-2026'
GROUP BY cost_center
ORDER BY variance DESC;

Locked down.

Where we’re going.

1. Now (Shipped)COMPLETED
2. This QuarterIN PROGRESS
3. Next 6 MonthsPLANNED
4. Vision (12+ Months)NORTH STAR

What Makes This Different