"Replacing Coding with Fine-Tuning"

Is it possible to train a model with fine-tuning to achieve overfitting and replace coding?
**Reason for overfitting: To shut down the creativity part. To create a structure that produces the same results all the time, just like in coding.
When a new request comes, just train the business logic and change the behavior of the code.

For example, a model that calculates amounts from a request according to a certain mathematical model.
If a promo code is added for the customer, the model should apply a discount of %x on the cart amount.