Hey everyone! I forgot which channel is best suited for this. But I’m still working on how to be a better AI Prompt engineer. I have 2 prompts, one where I started the idea with Claygent and have tried to adapt and change as I went, the other I started with ChatGPT and then worked my way there. Both prompts ask for generally the same thing but I’m consistently getting different answers for each as I sample them. The info I’m looking for is niche for my industry, it’s a company’s asset portfolio (property management) so this data isn’t easily obtainable, and has to be inferred a lot of the time. I’m needing help from someone who is much more experienced and versed in this than I am to check out my prompts, get an understanding of what I’m looking for, and how to build a more robust version to use. Any and all help would be amazing, if this isn’t the right channel. Point me in the right direction and I’ll post it there! Thank you!
Each unique prompt will generate unique results each and every time you run it. Keep iterating the prompt and testing across the models until you get the results you're looking for. Give explicit instructions for what you're looking for, where to look, etc ask the LLM to help you iterate and refine the prompt against the results you're looking for, ask other models their feedback on how to refine the prompt, copy/ paste 10x prompt results into Claude & ask: "these are the results I'm getting, but I want 'x' results, help me refine the prompt." And keep going until you get the results you want. You'll need to now your industry, ICP, segment, prospect pain points like the back of your hand, #HI is critical.
In this scenario, I’m usually using GPT-4o Mini. Sometimes, there’s no data available, so there’s no way the AI can know something. In those cases, you need to look for data sources.
