Hi! Would love some guidance on how to best assemble a complex AI-personalised email paragraph within Clay. Specifically: assembling a range of scraped data-points about contact + company into Clay (e.g. Linkedin posts, recent company press, etc), to combine that with problem statements we're testing, to create that para in Clay. Would then export that field into Hubspot, as a custom field per contact. Thanks!!
Hey Philip, thanks for reaching out! Happy to help with some info. Usually, most people wouldn't ask AI to write an entire draft or majority of it because it could be inconsistent at scale since the inputs/data-points will likely vary and modify AI's reasoning.
Most often than not, people use AI to personalize very specific sections of their drafts based on the inputs, so if that's what you have in mind then it work out.
Not so recent, but here's an example of what that may look like if you happen to have several different items to talk about but want to prioritize one vs others under conditions. Example of chatGPT personalization waterfall + conditions: https://www.loom.com/share/f73d46385002402fb8e4f88016329ef8?sid=7e3eebf5-4f29-45d0-9201-65de05598cf6
Thanks Arturo O.. Follow-on questions:
Re LinkedIn Post scraper: How do I ensure the longest post is selected, if multiple are found? I have one example of a user with 8 posts, but the one selected was essentially a "great, well done" repost, so wasn't useful
Re News article scraping: can I use a waterfall to ask a 2nd provider (not sure which to pick?) to search for news articles if the first fails - or perhaps scrape a company's linkedin posts
Hey Philip, thanks for the additional q's! There are different ways to go about this for question 1:
You can use a formula column and ask it to return the longest post from the list, and then that's what you give chatGPT instead. Else, there's an integration called Filter List of Objects that allows you to filter by keywords found if that's important for you. Filter list of objects: https://www.loom.com/share/5a648ca690c645c8a19261af0630aeaf?sid=b8bf7bfb-a861-4745-ae9e-7a9f74bd4aef Filtering with formulas example but you would prompt it to find and return the longest "post" -https://www.loom.com/share/c81034190c954983b4d303110bd92339
For question 2:
You can also use the LI post integration on company LI URLs. We do have a News integration waterfall if you'd like to use that option, I personally find that PredictLeads is the one that usually has the most results, Google is okay but it's also like a standard google search that could return similar posts that may not have anything to do with the company (unfortunately, that's how their endpoint works).

