{
    "action": "tool_call",
    "reasoning": "No evidence is in hand, and the draft numbers weren't backed by any query I can see. I'll re-ground the answer by looking at the raw Amount formats across all five monthly tables before ranking anything.",
    "tool": "query_sql",
    "args": {
        "sql": "WITH u AS (SELECT 'Jan' AS m, \"Company Name\" AS company, Sector, Stage, Amount FROM raw.\"Indian Startups _ Funding  Investors Data January 2022\" UNION ALL SELECT 'Feb', \"Company Name\", Sector, Stage, Amount FROM raw.\"Indian Startups _ Funding  Investors Data February 2022\" UNION ALL SELECT 'Mar', \"Company Name\", Sector, Stage, Amount FROM raw.\"Indian Startups _ Funding  Investors Data March 2022\" UNION ALL SELECT 'Apr', \"Company Name\", Sector, Stage, Amount FROM raw.\"Indian Startups _ Funding  Investors Data April 2022\" UNION ALL SELECT 'May', \"Company Name\", Sector, Stage, Amount FROM raw.\"Indian Startups _ Funding  Investors Data May 2022\") SELECT m, COUNT(*) AS n_rows, COUNT(Amount) AS n_amount, SUM(CASE WHEN regexp_matches(Amount, '^\\$[0-9,]+(\\.[0-9]+)?$') THEN 1 ELSE 0 END) AS clean_dollar, SUM(CASE WHEN Amount ILIKE '%undisclosed%' THEN 1 ELSE 0 END) AS undisclosed, SUM(CASE WHEN Amount IS NULL OR trim(Amount) = '' THEN 1 ELSE 0 END) AS blank FROM u GROUP BY m ORDER BY m",
        "purpose": "Profile Amount text formats per month to see how many are cleanly parseable dollar values",
        "source": "raw"
    }
}