Why isn't scheduling just a calculation an agent can do with a formula?

**Isn't this solved by an amortisation formula?**

For a clean new schedule with fixed terms, a formula does most of the work. Real scheduling rarely stays clean, and the formula answers only one part of the task.

**What else is the agent deciding?**

- **Which rule applies.** A rate change, a hardship arrangement and a contract variation each treat accrued interest or fees differently. - **Whether the request is allowed.** Policy may cap deferrals, require approval above a threshold, or forbid changes while a dispute is open. - **What to do with the old schedule.** Superseded lines must be closed, not left active beside the new ones. - **How to handle rounding.** Instalments are rounded to the cent, and the last line absorbs the difference. - **What to tell the customer.** The agent must state the new dates and amounts accurately.

**Where do agents fail?**

Typical failures include a plan that sums correctly but ignores an approval requirement, a new schedule created while the old one stays live, or a date placed on a non-business day when the policy rolls forward. Each of these is invisible to a formula check.

**What does that mean for an evaluation?**

A useful task pack separates the calculation from the decision. It scores whether the agent picked the right action, applied the right policy, left the records consistent, and communicated the result. A generated environment can hold the policies and linked state needed to test all of these together.