How did you make such an amazing gen AI app? I’m using an agentic workflow with many steps and many LMs. I tried doing some prompt engineering for my text to sequel app, but couldn’t get better than 20%. Text to sequel is a difficult problem. Our current workflow has six different steps. Wow. Can you walk me through it? The first thing we do is we categorize each incoming request. Does it have to do with data? Does it have to do with sequel? This way we’re not answering questions like how many ducks can dance on a pin. Then we have a context enrichment agent which provides another level of information. This is something I need. I noticed all the best solutions out there add a layer of human expertise. It really helps performance. We’re able to get usable answers in complex scenarios. The next step is generating the sequel. And we use different LMs because we find different LMs have different strengths. Gonna cost you a bit more. And the sequel they return isn’t perfect. So we have an air correcting agent that goes and validates and fixes any broken sequel. And finally a synthesizing agent that brings all that stuff together into a final answer. Amazing work, but I just need to demo this to my executives. So can you help me fix my prompt?