In order to move to the next phase I need to understand how we might be able to efficiently tackle the items for Phase II.
Watch the inspiration video for background:
(Prompt mentioned in email at the bottom of the page
Questions to be Resolved Before Phase II
Speed
Please see the Sana experience with regard to speed. What is the limiting factors that cause our experience to appear so much slower?
What would need to change to have our experience match? Let’s be creative, from pre-data processing of common processes to caching to anything else that makes sense. Could we branch processes for small versus large accounts?
No table Response Presentation Layer
What prevents us from such an elegant presentation layer (Sona) for common questions that don’t require table output? Anything? What would you need from me for the templates/structures?
Questions about Phase II
Scheduling Tasks
Please spend some time reviewing exactly how the new Tasks functionality works for OpenAI (screenshots below from my account). I would envision us being able to give users the option to either run a task of schedule it to run on a cadence and then delivery the results as an email. (In the platform we would have the email. For this phase, we would require users to provide an email address).
Please consider what would be required. Again, this would need to be live only in our connected universe so it would only be a demo to validate the processes.
Imagery
If a url for an image is contained in the API pull, can this be presented in the output? How much flexibility could we have in this output generation? Here’s what ChatGPT can do now (also see how the conversation is set with the user to the right at maybe 60-70% max with with the lighter grey box? This is what we are trying to accomplish).
Now, imagine that we are asking for a location record with a small number of returned locaitons. Instead of a table or just text, we present imagery (yes, the image URL would be part of the API). What do you expect as a challenge for this effort?
Prompt
Reputation Summary
#Context
This GPT is focused on analyzing individual reviews contained in json files to summarize their content and analyzing monthly reviews to compare overall sentiment towards the company.
#Function
This GPT needs to be able to create a summary of reviews and the overall sentiment of inputted files and chat, including presenting analysis over time. This GPT needs to identify names referenced in reviews to highlight employee feedback. This GPT also needs to provide strategic advice for improving reviews based on the feedback provided by reviews.
#Tone
This GPT has a conversational tone, you are crafting a story surrounding the metrics provided.
#Format
For the title reference the name of the business instead of the file name. Start with an overall summary that includes the number of reviews and average review rating, followed by a single sentence summarizing the month.
Then present topic analysis and identify the top topics shared by customers in the month, then share the top 3 with a sentence explaining them.
Then include verbatim review quotes, highlighting employee names when they are mentioned, all without file search citations.