{
    "action": "tool_call",
    "reasoning": "I have good evidence on AI usage patterns across demographics and work behaviors. I've found some interesting patterns: 1) Engineering & science (72.19%) nearly matches Software & IT (74.80%) in AI adoption, 2) Age shows 30-44 year-olds (60.38%) using AI MORE than 18-29 year-olds (58.38%), which contradicts the 'digital native' assumption, 3) Only 4.68% let AI do most/all of their work despite 55% using AI, and 4) 78.70% hand off NONE of their tasks fully to AI. The age finding seems most counterintuitive - let me verify this pattern is robust by looking at it alongside education and occupation to see if there's a confounding factor, and also check income patterns which I haven't explored yet.",
    "tool": "query_sql",
    "args": {
        "sql": "SELECT Question, Response, \"Income: Under $25K\", \"Income: $25K–$50K\", \"Income: $50K–$75K\", \"Income: $75K–$100K\", \"Income: $100K+\" FROM raw.polling_on_ai_usage_jul_2026 WHERE Question = 'Any AI Use' AND Response = 'Yes'",
        "purpose": "Check if AI usage correlates with income in expected ways, to contextualize the surprising age pattern",
        "source": "raw"
    }
}