A team needs an AI solution for a standard supported task quickly and lacks a custom training dataset. Which approach best fits?

1. Estimate end-to-end workload cost using realistic usage and workflow behavior.
2. Evaluate customization after establishing a baseline and clear evaluation criteria.
3. Connect an approved enterprise data source with appropriate access controls.
4. Evaluate a suitable pre-trained model or API before commissioning custom training.