Start With the Model. Harness What You Cannot Afford to Guess.
Every team is arguing model versus code one ticket at a time, and most of them are arguing the wrong thing. Start with the model. Let it run in a loop until the check passes. Then build the environment it works in, and write down the few actions it cannot take on its own. That is the harness, and its length is set by risk, not by how much the vendor charges.