Can I Build This Clay Table Faster Than Sculptor (Clay's AI?)

0:00 I just used Clay's new AI feature to build an entire end-to-end Clay workflow in under 10 minutes with natural language. Now, I'm going to try and beat my own score using just these two hands, so to speak, mouse and keyboard, no natural language to see if Sculptor actually speeds up the process of building clay tables or just gets in the way. You can see here my last table was built with Sculptor and I have the video on LinkedIn and on YouTube if you want to watch it so you can compare against the results from this video. But you can see here we have a company table with domain record ID, company name, company LinkedIn URL enrichment, employee account URL, uh company LinkedIn URL. We have a revenue waterfall, company address and company summary as well as the industry classification. Now the rules for this video are I can only use these two hands. I am going to use chatbt to write some prompts using my voice talking to chatt because I do that anyway when I'm building in clay without sculptor. That was possible before sculptor hit the market which is clay's new AI building assistant. So, I'm gonna do that, but I'm not going to actually use Sculptor to build the clay table.

0:58 Okay. Three, two, one, go. Okay. The first thing we're going to do is we're going to add the company name import. So, I'm going to say company name. We're going to add the company domain import column as well. Domain. And then we're going to add the record ID here as well. Okay. Next up, we're going to populate some dummy data. I'll say Senoso. Senoso.com. And then record ID is one, which is a fake ID. Now, we're going to go ahead and enrich company to get their LinkedIn data right here. Okay. Boom. Um, and we're going to pull employee account, and we're going to pull out their LinkedIn company URL right there. Okay. So, boom. I don't want this. Actually, I already have this. So, we're just going to get their company uh LinkedIn URL. Okay. Great.

1:47 Now, oh, this is messed up. Sorry, bug. And then we want employee account. So, where's employee account? There we go. Okay, great. So, now I have employee account as well. Next up, we're going to go ahead and get the address. So, we're going to do a clay agent prompt for that. We're going to use uh 40 mini. And I'm going to go to GPT to build my uh prompt here. Okay. So, here we go. Please build a prompt for Clay's research agent that can based on a company's domain only research the company and look for their headquarters address using the standard US mail postal code for uh format with a JSON output that has street, city, state with the state abbreviated with twoletter code like MD or CA and the country. the country being fully spelled out like United States not spel not just USA or US but United States and then output the result in JSON format according to the US postal code.

2:50 Okay, so now we have this here going to say turn this into a prompt for collagent. We're going to enter and then we're going to wait for this to go. It's probably going to be uh pretty good. Okay, here we go. Grab all of this. Going to go back to clay and we're going to dump. Oh, whoops. No, I don't want any of this. 40 mini and we're going to dump this in. And then I'll say here is your input and I'll say domain and then I'll say company name and I'll say send uh name.

3:38 Okay, great. Now we're going to use JSON schema generate from prompt to create those columns and then we're going to hit run. While that's going it should be fine. We're going to create the um summary company summary prompt. Actually no, we're going to do a few other things. We're going to knock out revenue. How am I doing on time? Only 3 minutes. Company revenue. Great. And we're going to remove this from the waterfall. And then we're going to go ahead and we're going to run that. I'm also going to go for uh marketing employee headcount. I forgot if I did that in the last video or not, but I was supposed to. So, we're going to go for marketing or say headcount, and we'll see if we can get headcount by title over here. Employee count by criteria.

4:20 And we're going to open this up. And job title keywords. I'll just say marketing uh marketing demand generation and growth. Okay. And then add fields roll count and hit go. Uh how are we doing on the Okay, great. We have street, we have city, we have state, and we have country and we have postal code. So, we have everything we need right here. Fantastic. I'm not ordering this because I'm trying to beat the clock, but I have all the data. There's revenue range.

4:54 There's the headcount and all we need left is the industry classification. So, we're going to go to fort mini and we're going to say um I think I just need something. Okay, one second. Back to JBT. Now, I want you to create a collagent prompt that just takes a company domain, goes to the website, and creates a concise company summary paragraph that explains what they are, who they who they sell to, what kind of company they are, their target customer, their products and services. And we're going to use this summary for use in an industry classification prompt later on. So be sure to be descriptive enough knowing the use case of your output.

5:29 Okay, so now we're holding at five minutes. So here we go. Going to just grab this and [sighs] okay, great. Awesome. Back to clay. So we're going for the 40 mini output here. And then we'll say input and we're going to add the domain here. And then we're going to run this. And then while that's going, we're going to go ahead and and um grab the the prompt that we need.

6:11 Okay. Please create an AI prompt using GPT40 that can essentially take a company summary and then assign the most relevant industry based on this Okay, my video died because my camera overheated. I'm going to try and turn it back on and just push this thing to limit for a second here.

7:23 Yeah, we're back and blurry. Okay, here we go. And we're going to go ahead and run this. And then we're going to be done with this workflow. We're holding at 6 minutes and 40 seconds. Even with the camera malfunction. Oh, I forgot to use Oh, no.

7:53 Here we go. Run one more time. We should be in good shape. and there we go. Industry is other, which I agree with. Okay, so there you have it. We did this all in 7 minutes and 12 seconds. That's the final time here. I'm going to go ahead and show you. So, we did this a few minutes faster than clay. I think building a real clay workflow obviously is much more nuanced than uh what you see here.

8:26 And a real clay workflow that's that's in-depth for a real customer would take probably hours. Would Sculptor actually help? Maybe, maybe not. Uh I think in future iterations it could. And I actually have certain use cases in mind that it's great for. I actually posted a video or will post a video recently. I've recorded a video showing specific cases where creating annoying function columns is very useful with with the sculptor prompt. you can tell it what to do and it will just sort of get you more than Clay's AI formula generator. It's much faster and easier. But building entire clay tables, I think we still have ways to go before Clay's sculptor feature is really ready for that. I built it faster myself. It didn't really speed me up. If anything, it kind of slowed me down, but I think in the future it has a lot of potential. So looking forward to see where this goes.