Choosing the Right AI Model: Effective Methods and Best Practices

What’s the best way to choose the right AI model for a use case? I tried testing small samples across different models and comparing results. Does this approach work, or is there a better method?