A finance operations team must apply an exact published tax-code lookup table with no ambiguity. There is no need to infer patterns or generate content. Which initial approach is most appropriate?

1. Use deterministic business rules and assess AI only if an additional uncertain task requires it.
2. Replace every lookup with a generative model because AI always improves exact rule execution.
3. Train an unsupervised model and let its clusters redefine the published rules.
4. Allow the model to invent missing codes instead of handling invalid input.