Best Practices for Using AI to Qualify Comments with Claygent

Phil Sergenti · 2025-01-08 13:09:36.433018+00

`need help from an experienced clay operator` I have 1000s of people who commented under my posts, now i want to use AI to qualify them I'm planning to use Claygent to check a profile one by one and tell me if the they are either a competitor agency, a qualified company or else return unqualified. Should I do this with claygent with 1 single passage or maybe do an HTTP call to get all the profile info in HTML/TXT; do the same for company and use straight GPT with no access to internet? What's the best way to do this?

Best Practices for Using AI to Qualify Comments with Claygent | Clay

Phil Sergenti
·
·
Marc Stu spotted lol, love u ❤️
🫶1
Channeled
APP
·
·
Here's how you can qualify your LinkedIn post commenters using Clay: 1. First, use Clay's template for identifying and enriching profiles of people who commented on your LinkedIn posts. 2. Then, you can use Clay's AI capabilities in two ways to categorize these profiles: • Use the ChatGPT integration to categorize the enriched data into your specified buckets (competitor agencies, qualified companies, unqualified). • Use Claygent to research additional information about their companies by analyzing their websites and online presence to make more informed categorizations.The best approach would be to combine both methods: use Claygent for real-time research on companies and websites, and then use AI integrations for the final categorization of the data into your specified segments. Sources: What can Clay do with AI? - FAQ | Clay.com Create company reports | Clay Clay AI Capabilities Claygent vs. AI Integrations
kushagra tiwari
·
·
Hi Phil Sergenti we have built an end to end automation to exactly solve for this use case and more - https://www.loom.com/share/7469e40b992a479ba10b06f864a1c429
Channeled
APP
·
·
Can you tell me more about what specific aspects of the automation you're interested in or having trouble with?
Phil Sergenti
·
·
looks amazing kushagra tiwari, i have a few ideas on how to make it better, would love to chat
Channeled
APP
·
·
Looks like you're still looking for help. We're escalating this over to our support team and someone should be in touch soon. In the meantime, feel free to include any other information that may be helpful in solving your issue!
kushagra tiwari
·
·
happy to chat - moving to dm
owen (Clay)
·
·
Hey Filippo and Kushagra, sounds like you both have this handled! Andreas performs a similar workflow here as shown at about 10 minutes into the video using Claude: YouTube Let us know if you have any questions regarding this process or need any assistance. :)

Phil Sergenti

·

how would you guys improve such a prompt? owen (Clay) & kushagra tiwari?

Your job is to classify the prospect "author/name" as either a COMPETITOR, QUALIFIED or UNQUALIFIED 

here's the linkedin profile: "author/profileUrl"

This is your context:
Our agency sells lead generation services via content and cold email; our ICPs are startups, software companies and agencies (but not those that sell the same services as we do). 

This is the process you should follow:

1. f a person sell "lead gen" "lead generation", "cold email", "cold outreach", "clay", "GTM engineering" and promises leads/booking appointment to their clients, return COMPETITOR

2. If a person works for a company with more than 6 people, and it's not a competitior selling our same services then it's qualified, it's even more likely to be qualified if it's a startup founder, raised money and build a software, that case you return QUALIFIED.

3. if it's not either a COMPETITOR or QUALIFIED, then return UNQUALIFIED

This is very important for me, you're loosing 1000 bucks for every mistake you make, return just one single word and nothing else,  reply only with the classification (COMPETITOR/QUALIFIED/UNQUALIFIED)

classification:

This is your context:
Our agency sells lead generation services via content and cold email; our ICPs are startups, software companies and agencies (but not those that sell the same services as we do). 

This is the process you should follow:

1. f a person sell "lead gen" "lead generation", "cold email", "cold outreach", "clay", "GTM engineering" and promises leads/booking appointment to their clients, return COMPETITOR

2. If a person works for a company with more than 6 people, and it's not a competitior selling our same services then it's qualified, it's even more likely to be qualified if it's a startup founder, raised money and build a software, that case you return QUALIFIED.

3. if it's not either a COMPETITOR or QUALIFIED, then return UNQUALIFIED

This is very important for me, you're loosing 1000 bucks for every mistake you make, return just one single word and nothing else,  reply only with the classification (COMPETITOR/QUALIFIED/UNQUALIFIED)

classification:

Phil Sergenti
·
·
using o4-mini with web access
kushagra tiwari
·
·
improving the inputs from just profile to actually enriching and getting summary+ work ex + company description and then tell the gpt to just classify will work best
kushagra tiwari
·
·
then it doesn't even need to be a web agent you can use the AI content option
Phil Sergenti
·
·
right but i'd spend 1 clay credit per every row, right?
kushagra tiwari
·
·
or a third party api to get the linkedin data
Phil Sergenti
·
·
got any ideas on what that would be?

18 comments