We’re trying to create automation that replies to 5 star reviews on TrustPilot so our agents don’t have to.
It’s a rather tumultuous adventure so far as the automation worked at first and then it stopped and when it came back to life it gives only one answer (although our prompt gives creative examples of answers)
Would it be possible to check, as we’re running out of ideas.
In the example below, I’ve used this action to generate both a post caption and hashtags. Perhaps you could predetermine that ChatGPT should provide 5 different types of answers/output. So in your structured data, you would specify Answer 1, Answer 2, Answer 3…and so on and your prompt can determine how each answer should be crafted.
Hello and thank you for your answer,
What we’re trying to achieve is to feed OpenAI with e.g. 10 sample answers and make it choose one and modify it according to the customer’s review.
However when we swapped the module to the “transform text to structured data” we are getting the error below: