A municipal courts office asks whether an LLM can replace its statistical epidemic-style models for forecasting case volumes. What is the most accurate practitioner response?

The accurate practitioner response is no: generative LLMs are weak substitutes for purpose-built quantitative forecasting models when municipal courts need calibrated case-volume projections. Forecasting for staffing and docket planning requires statistical rigor, uncertainty intervals, and methods tuned to time-series structure. LLMs generate fluent text; they do not reliably replace calibrated epidemic-style or classical forecasting systems for numeric planning. Claiming any generative chat model is a drop-in replacement for calibrated forecasts overstates LLM capability and risks bad resource decisions. Asserting LLMs always outperform time-series models on numeric projections is unsupported and exam-wrong. Believing polite prompting eliminates forecast error confuses conversational UX with statistical validity. Exam caveat: generative models may narrate or explain forecast outputs, but narration is not the forecasting engine. Operational check: keep official volume forecasts on validated time-series or statistical pipelines with backtesting; if staff use an LLM, limit it to drafting explanations of already-computed figures, require numeric citation from the forecasting system, and never let chat output alone drive overtime or courtroom scheduling.