Forecaster Agent

WHAT IT DOES

Connects to multiple data sources (databases, APIs, spreadsheets) and builds forecasting models that explain their reasoning. Uses ensemble methods — ARIMA for trend, seasonal decomposition, gradient boosting for non-linear signals. But the differentiator is triangulation: it pulls correlated signals from multiple sources (web traffic predicts revenue, job postings predict growth, support tickets predict churn) and weights them into the forecast. Monitors actuals vs predictions in real-time. When reality diverges from forecast, auto-recalibrates and explains what changed.

WORKFLOW

1. Ingest historical data from multiple sources
2. Decompose trends + seasonality
3. Identify correlated external signals
4. Build ensemble model
5. Generate forecast with confidence intervals
6. Monitor actuals
7. Detect drift
8. Recalibrate
9. Explain divergence

SKILLS

INTEGRATIONS