A central limit theorem for normalized sums of random variables that form an autoregressive integrated moving average (ARIMA) process is developed. The need for such a limit theorem is discussed in connection with modeling total compensation costs associated with insurance or medical claims.