The article nicely shows how simple it really is to build a rag system. The simplicity is part of why rag is so popular, but the hard part is

1. determining a good rag use case for your business (hopefully something more inspiring than chat with our internal knowledge system or customer service chat bot).

2. tuning each part of the system to optimize for your use case. In this case, you need data and some way to benchmark how good your system is and whether it's getting better.

I'd be curious to hear people's experience successfully deploying rag systems that had positive business impact.