@Phil L I really like how you thought about the quality definition. Something for you to think about it: how will you measure each of these, especially at scale Success: same question -- how do you measure things like reuse signal? Overall, I think you are on the right path by wanting to meausre the right things. But it is also important to think "how" you will measure them. And the answer to that includes two things 1. what is the exact metric that you want to measure. what is the mathematical formula 2. is it feasible to measure it at scale

@Karthik Gunasekar Hey Karhtik, I recommend trying rewriting the prompt based on the new principles we learned for writing good prompts. ANd then see if the output changes drastically?

@Pushkar Chhibber Thanks for sharing. One other thing an AI PM could think of doing is to use the feedback from all users of a company to develop their own templates for different doc types. This way, over time the response is more "acceptable' even for first time users.

Thanks @Aditya Gupta for sharing the link. Hey @Vedanshi Murarka all the details of the activity and the learning are captured in the module that Aditya has linked in his comment. Let me know if that helps. Also, if you haven't used Skool before, here is a quick video that walks you through the platform and shows you how to use the platform for all the tasks in the course.

@Peculiar Ediomo-Abasi - I like this idea a lot, since it really solves a very big pain point for users. Can you talk more about how you are thinking about "does the data exist" Are you saying we should look at past supply and demand trends and use that for forecasting future demand? If yes, then can't that be solved by simple rules? In other words, past trends might show things like: 1. Nov is the busiest month. We need 20% extra syringes 2. Last 3 years, we had 30% extra paracetamol, so let's order 30% less. Is this what you're referring to?

@Manasa Shetty This is a problem that is very close to my heart. I became a dad last year and I saw first hand how much support a new mother and her baby needs. One assumption of yours that I would challenge is the ease of capturing data on the baby's routines. New moms don't have that much time and often capture this data in a dairy or not at all. This makes giving customised advice very difficult. Another aspect to think of is how we can capture better quality data - maybe we connect this to a wearable device capturing mother's vitals and using that to share insights and recommendations with her. As a PM, I would also focus on ensuring we have the right controls in place to manage use case where incorrect recommendations are shared. This is a big risk, especially when dealing with medical advice. Lastly, I would add there are enough sources of structured data that can help train the LLM. So that will not be a challenge.