Why should you care?

Executives want why-did-revenue-dip answers without waiting for analyst SQL and slide decks.

See it live — copy this example

Paste into an ASP.NET Core 8+ / AgentNest project, then run with dotnet run (set your API keys in user-secrets).

// AgentNest.Analytics/MetricsNarratorAgent.cs
public sealed class MetricsNarratorAgent(IChatClient chat, IWarehouseQuery sql)
{
    public async Task<string> ExplainWeeklyAsync(Guid tenantId, DateOnly week, CancellationToken ct)
    {
        var rows = await sql.QueryAsync("""
            SELECT region, SUM(revenue) rev FROM fact_sales
            WHERE tenant_id = @t AND week = @w GROUP BY region
            """, new { t = tenantId, w = week }, ct);
        return (await chat.GetResponseAsync([
            "Explain revenue by region using only this table:\n" + rows.ToMarkdown()
        ], cancellationToken: ct)).Text;
    }
}

What happened?

• MetricsNarratorAgent runs parameterized SQL against warehouse, converts rows to markdown table, asks LLM to narrate — numbers come from SQL not model memory.
• Follow the steps below — typing the code yourself is the fastest way to learn.

Practice next

1. Use read-only warehouse credentials with row-level tenant filter.
2. Register MetricsNarratorAgent in analytics API.
3. Schedule weekly brief job per tenant CEO dashboard.
4. Add anomaly detection step before narration.
5. Export brief as PDF attachment in Monday email job.

Remember

Analytics agents narrate SQL-grounded metrics. Never let LLM generate unrestricted warehouse SQL. Show figures alongside narrative for trust.

Monday CEO brief

CEO opens AgentNest analytics dashboard Monday 7am.

Outcome: MetricsNarratorAgent explains regional revenue shift with SQL-backed numbers.