Demo • Clay — Claygentic Distribution

946
02:44:40.150 –> 02:44:48.240
Julia Nimchinski: Amazing, thanks again. And welcome to the show, Patrick Spiccholski, our community favorite, Collegentic Distribution.

947
02:44:48.700 –> 02:44:50.119
Julia Nimchinski: How are you doing?

948
02:44:50.570 –> 02:44:53.980
Patrick Spychalski: I’m doing well, cannot complain, and pumped to be here, thanks for having me.

949
02:44:54.480 –> 02:44:56.079
Julia Nimchinski: We’re pumped to host you.

950
02:44:56.540 –> 02:44:58.079
Julia Nimchinski: What are we seeing today?

951
02:44:59.190 –> 02:45:08.249
Patrick Spychalski: Yeah, so today I’m planning on showing off a system that we’ve actually built for ourselves internally that helps automate, I’d say, around, like, 90%, probably, of the

952
02:45:08.490 –> 02:45:18.599
Patrick Spychalski: the pre-call prep that is required for us to get on a demo call with a potential prospect. So, it’s something I’m pretty pumped about, and yeah, pump the show.

953
02:45:18.980 –> 02:45:19.649
Julia Nimchinski: Love it.

954
02:45:19.750 –> 02:45:21.010
Julia Nimchinski: Let’s 17.

955
02:45:21.810 –> 02:45:24.180
Patrick Spychalski: Cool. I will share my screen.

956
02:45:26.000 –> 02:45:27.400
Patrick Spychalski: Can you see my screen, Julia?

957
02:45:29.100 –> 02:45:39.280
Patrick Spychalski: Amazing. Okay, so, the system I built today is one that prepares my agency internally for, getting ready for calls when a prospect books one.

958
02:45:39.320 –> 02:45:50.310
Patrick Spychalski: As a quick high level about what my company does, which I think is relevant to the table, we are an agency that helps build go-to-market systems using AI tools like Clay and N8N for mid-market to enterprise companies.

959
02:45:50.310 –> 02:46:00.629
Patrick Spychalski: And you can probably imagine, due to the fact that we’re a relatively high-ticket service, there’s a lot of preparation involved in getting ready for a sales call. Two of the main things we usually do

960
02:46:00.680 –> 02:46:20.339
Patrick Spychalski: is build out a PowerPoint deck or a slide deck outlining some of the things we can do for the company, as well as build a, proof of concept in Clay, so we can generally show them the type of stuff that we can build for the company specifically. And everything we do is very custom, so it’s tough to prepare all this stuff. Like, for every large company we get on a call with, it takes, like, hours to do

961
02:46:20.340 –> 02:46:29.080
Patrick Spychalski: all of this work. And so I was like, how can we do this faster using AI agents? That is our job as a business, so it’d be pretty embarrassing if we couldn’t figure out a way to do it.

962
02:46:29.220 –> 02:46:35.540
Patrick Spychalski: So I wanted to show you this workflow. You can see we have, 5 different companies that have been imported.

963
02:46:35.540 –> 02:46:49.929
Patrick Spychalski: into Clay from HubSpot. For those who don’t know what Clay is, I feel like that list of people is getting smaller and smaller as time goes on. It’s a go-to-market workflow tool, so you can automate a lot of really cool things, and it has a lot of great enrichment capabilities. But, we import these 5 companies into Clay.

964
02:46:50.160 –> 02:46:54.630
Patrick Spychalski: You have, the… a few basic pieces of information from HubSpot, like name and domain.

965
02:46:54.760 –> 02:47:11.289
Patrick Spychalski: Clay has a bunch of enrichments in the platform, so the next step we wanted to take was to find more information on these companies, and they have this enriched company integration. All you have to do is throw in their LinkedIn profile or domain, and you can output all of the data in their LinkedIn. So, we did that.

966
02:47:11.290 –> 02:47:17.400
Patrick Spychalski: I’ll put a few different data points we found valuable, like where the company’s based, their description, their industry, their employee count, and revenue.

967
02:47:17.740 –> 02:47:29.560
Patrick Spychalski: Clay also has enrichment waterfalls, and so another thing we wanted to figure out and research prior to talking to them is their funding. Clay waterfalled a bunch of different enrichment data providers here to find the company’s funding. You can see its output here.

968
02:47:29.790 –> 02:47:48.420
Patrick Spychalski: We also did another search to see how many sales team members the company has. So you’re able to actually throw in some job title keywords specific to that company. We put in SDR, BER, a bunch of, you know, like, sales team member keywords, and we figured out how many sales team members they have based on the LinkedIn search that we ran.

969
02:47:48.870 –> 02:48:00.619
Patrick Spychalski: We then scraped their 10K for some base-level financial data, like profitability, fundraising, revenue, all that good stuff to, to validate the revenue metrics we got from the previous enrichment.

970
02:48:00.860 –> 02:48:03.820
Patrick Spychalski: And this is where it gets pretty cool. So the first thing we did.

971
02:48:03.910 –> 02:48:15.209
Patrick Spychalski: was we wanted to ICP score this company, and so I prompted an AI agent to go scrape the web about this company, and I gave it a ton of different criteria about which we qualify by.

972
02:48:15.210 –> 02:48:34.240
Patrick Spychalski: Scores, weights, I don’t want to go through the entire prompt, it’s pretty boring, but you can imagine, everyone has their own scoring system. We fed our own scoring system into this prompt, and based on that scoring system, we were able to output a final score from 1 to 10. You can see most of these companies, given that they’re large enterprise software businesses, people that we target pretty, pretty heavily, they’re all great fits.

973
02:48:34.490 –> 02:48:43.369
Patrick Spychalski: Next, we went and found their tech stack, so we found the entire company’s tech stack using this integration. We threw in the pieces of tech we care about the most, so these… this is a…

974
02:48:43.590 –> 02:48:55.800
Patrick Spychalski: Limited, but, you know, still pretty long list of companies that we integrate with or work with, pieces of tech specifically, and so we figured out whether these companies work with these pieces of tech, and we output 3 examples.

975
02:48:56.130 –> 02:48:59.620
Patrick Spychalski: So now we get to the part where we do really deep research.

976
02:48:59.900 –> 02:49:12.389
Patrick Spychalski: And we create both the POC and then the PowerPoint deck. And we actually generate a clay table, you’ll see in a second, automatically, and we generate a PowerPoint deck automatically using this clay table.

977
02:49:12.630 –> 02:49:19.630
Patrick Spychalski: So, or not PowerPoint, but a Gamma AI, like another, you know, I’m sure people have heard of it, but like an AI presentation software.

978
02:49:19.760 –> 02:49:43.010
Patrick Spychalski: So the first thing I did, in this part of the process is I prompted Klagent to do a… give me a high-level summary of the sales team for each one of these companies. So you can see it gave me a really long description of, like, everything involving this company’s sales wing. So, how much they pay their sales team members, how their sales team is structured, how they’re incentivized to close deals.

979
02:49:43.010 –> 02:49:47.219
Patrick Spychalski: We also did the same thing with their revenue model, so to figure out how they charge people.

980
02:49:47.220 –> 02:49:56.189
Patrick Spychalski: And we used all of this data that I found… I’ve shown you in the previous few integrations to recommend some specific use cases as to how we could help this company.

981
02:49:56.260 –> 02:50:14.949
Patrick Spychalski: So, I provided another AI column, a list of 17 use cases that we’ve done for companies pretty consistently, and I said, based on all of this information that I just found, tell me the 3 use cases that are the most valuable for this company, and then describe them in detail. And you can see it outputs a bulleted list of 3 really

982
02:50:15.170 –> 02:50:17.479
Patrick Spychalski: Great use cases in detail.

983
02:50:17.650 –> 02:50:19.000
Patrick Spychalski: For what we can build out.

984
02:50:19.280 –> 02:50:29.359
Patrick Spychalski: Then I told another AI column, pick one of these use cases and give me a step-by-step spec on how I could build this use case in Clay. And so, it…

985
02:50:29.370 –> 02:50:39.709
Patrick Spychalski: use a very high level of, like, an OpenAI API call to do this, but it was able to actually figure out how to build this table in clay based on the use case generated in the previous thing.

986
02:50:39.890 –> 02:50:41.700
Patrick Spychalski: We wrote this to a Google Doc.

987
02:50:42.190 –> 02:51:00.130
Patrick Spychalski: And then, we were actually able to use, it’s essentially Clay’s answer to OpenAI Operator, but a tool that can actually take actions on the web to create this clay table. Not just… not the clay table itself, I wish we could do that, hopefully we can do that soon, but at least generate the table to, a relatively accurate

988
02:51:00.130 –> 02:51:07.469
Patrick Spychalski: Percentage for our sales team member to start with. As well as now they have a document, outlining step-by-step how to build the table itself.

989
02:51:08.150 –> 02:51:21.640
Patrick Spychalski: Finally, the last thing we needed to do was generate a slide deck. Generating slide decks for anyone who’s done it, which I’m sure is everybody on this call, is annoying, it’s arduous, it takes a ton of design time, and I didn’t feel like doing it anymore as the person running founder-led sales in my company.

990
02:51:21.640 –> 02:51:32.109
Patrick Spychalski: And so, I used this really expensive 150 credit per row clay integration to create this slide deck for me based on all of the data that I found in the previous steps of the table.

991
02:51:32.110 –> 02:51:36.590
Patrick Spychalski: You can see it did a great job. It output 10 slides and the text associated with each slide.

992
02:51:37.130 –> 02:51:55.610
Patrick Spychalski: And then, again, I mentioned this earlier, but there’s this tool called Gamma. If you haven’t checked it out, I would really recommend doing so. It can create slide decks for you using natural language prompting. And so, we sent all of this to their API, which is currently in beta, they just came out with it recently. If you have a pro plan for Gamma, you can use their API.

993
02:51:55.680 –> 02:52:11.679
Patrick Spychalski: to generate slide decks for every single one of these companies programmatically. And so here is an example of what the slide deck looks like. You can see it has a high level on the kiln, what we do, and then it goes through each one of the use cases in depth. So, really cool.

994
02:52:11.680 –> 02:52:20.639
Patrick Spychalski: And, of course, this is not, like, something we’d immediately present to a client. I wouldn’t recommend that, but it’s a great start, and it’s significantly better, in my opinion, than doing all of this from scratch.

995
02:52:20.640 –> 02:52:30.589
Patrick Spychalski: saving us a ton of time in our sales process. And then we update all of this data back into our HubSpot to keep the loop closed, and to have a central source of truth for everything here. So…

996
02:52:30.600 –> 02:52:44.290
Patrick Spychalski: I know I just talked a lot, but hopefully that was, somewhat valuable, and of course, happy to answer any questions. And I feel like this just generally shows both the power of clay and how powerful AI agents are in being able to solve high-level tasks and take actions for you today.

997
02:52:45.620 –> 02:52:54.270
Julia Nimchinski: Always phenomenal featuring you. Patrick, thank you so much. Aren’t a ton, and people want to know, what can you share about Clay’s AI roadmap?

998
02:52:55.430 –> 02:53:06.469
Patrick Spychalski: Yeah, so, Play initially started just with LLM text generation calls, and so, like, the ChatGPTs of the world, it was, like, what was their integration? You could use ChatGPT at scale?

999
02:53:06.640 –> 02:53:10.400
Patrick Spychalski: They’ve now developed two AI agents that can do research for you, as was shown here.

1000
02:53:10.400 –> 02:53:27.870
Patrick Spychalski: And then the next step, which they’re currently building, and I was able to test out to some extent, was their operator function. So being able to take actions and tools, being able to take actions on the web, fill out forms for you, and actually complete tasks for you. And so, to me, this is the next iteration of what Clay’s going to do, take actions in tools.

1001
02:53:27.870 –> 02:53:32.730
Patrick Spychalski: And being able to, like, actually fulfill entire tasks instead of just do research.

1002
02:53:33.820 –> 02:53:38.250
Julia Nimchinski: And what was the most effective clay campaign you’ve ever run or seen?

1003
02:53:39.190 –> 02:53:48.350
Patrick Spychalski: Man, good question. We’ve had a lot. Funnily enough, one of the most effective ones wasn’t ridiculously complicated at all. We…

1004
02:53:48.380 –> 02:54:05.199
Patrick Spychalski: We’re running an outbound campaign to Fortune 500 companies on behalf of one of our clients. This one campaign has generated them between, like, $100 and $150 million in pipelines since we started running it for them, and it just reaches out to Fortune 500 leaders, referencing some sort of news that has recently come out about

1005
02:54:05.240 –> 02:54:20.549
Patrick Spychalski: their company, specific to the segment in which they work. So if they’re the head of marketing, it’s not really worth talking about an engineering feat inside of Google. The organization is massive for these companies. And so, we just did a lot of deep research to find specific releases about a given thing that that segment of the company did.

1006
02:54:20.550 –> 02:54:37.829
Patrick Spychalski: And then just ask them to, to chat. This was coming from, like, the CEO of a very large company, so that’s worth noting. It’s not like you could send this as an associate and get the same results, but, it was pretty much just asking for a quick coffee chat to, because they’re usually the services that this person was doing and the company had a lot in common.

1007
02:54:37.840 –> 02:54:46.089
Patrick Spychalski: So yeah, it was a very simple campaign, but it did incredibly well. It’s one of the case studies on our website. It was for, like, a 9,000 person, $2 billion a year company.

1008
02:54:46.860 –> 02:54:53.870
Julia Nimchinski: Curious, what tools are you experimenting with, Patrick? Like, any, anything new and exciting on the AI?

1009
02:54:54.500 –> 02:54:56.829
Patrick Spychalski: Yeah, I’d still say probably, like.

1010
02:54:57.130 –> 02:55:14.549
Patrick Spychalski: 80-90% of the work we do is in N8N clay. We do a lot of clawed code and cursor and lovable work as well. I wouldn’t say it’s what we are specifically, like, the most capable in, but, yeah, those and MCPs are, in my opinion, like, the most cutting-edge tech we can build and go-to-market, and we’ve been experimenting with a lot.

1011
02:55:14.620 –> 02:55:27.750
Patrick Spychalski: I’d say for those who aren’t super technical, though, I would stop at probably Clay, N8N, and maybe a little bit of Claude Code, but at least from my experience and my team’s experience, vibe coding without any technical ability is…

1012
02:55:27.790 –> 02:55:35.509
Patrick Spychalski: A lot harder than if you have any… even, like, baseline technical knowledge. So, we try not to build systems that we… we aren’t confident in.

1013
02:55:36.660 –> 02:55:42.570
Julia Nimchinski: People are asking, what do you wish AI could do on the GDM side of things, and it still can’t?

1014
02:55:43.680 –> 02:55:45.070
Patrick Spychalski: That’s a great question.

1015
02:55:45.340 –> 02:55:50.730
Patrick Spychalski: I mean, honestly, since the operator function launched, it’s really hard to say

1016
02:55:51.070 –> 02:56:03.670
Patrick Spychalski: that there’s not… that there’s anything it can’t do. I would say I wish the operator function was more robust and could do more. As I mentioned, we had Clay build clay tables in this use case, which is cool, but it wasn’t doing it perfectly.

1017
02:56:03.720 –> 02:56:12.610
Patrick Spychalski: It was building tables and being able to do a couple basic steps. If it could, like, actually think through how to use the tool properly, which is usually what an MCP would do.

1018
02:56:12.770 –> 02:56:22.790
Patrick Spychalski: That would be awesome, but it’d be great if we didn’t need an MCP to be able to do that, if we could take actions in a tool without having, like, robust API documentation to reference. That’d be… that would be pretty nice.

1019
02:56:24.270 –> 02:56:28.379
Julia Nimchinski: Alright, this… this is a fun one. Where is your agency in 2030?

1020
02:56:29.750 –> 02:56:34.789
Patrick Spychalski: Gosh, great question. I’d say the general direction in which we’re trying to go

1021
02:56:34.960 –> 02:56:38.589
Patrick Spychalski: Is, we’re still pretty confident that

1022
02:56:38.940 –> 02:56:44.340
Patrick Spychalski: 95% of agencies plus in the current space were not built

1023
02:56:44.350 –> 02:57:03.389
Patrick Spychalski: to serve AI implementation, especially with how fast it’s evolving. The McKinsey’s of the world cannot iterate nearly quickly enough to actually serve real value to a lot of large companies when it comes to AI implementation. And so we’ve remained small, for that reason, because I think it’s easy to pivot and for our whole team to learn new things.

1024
02:57:03.430 –> 02:57:05.540
Patrick Spychalski: But the goal would be to scale.

1025
02:57:05.630 –> 02:57:13.680
Patrick Spychalski: maybe not to McKenzie’s side, that’d be absurd, but to a very large agency that can still pivot really quickly with new AI development, I think that’d be… that’d be a nice goal.

1026
02:57:15.140 –> 02:57:19.189
Julia Nimchinski: Another one here is, who do you look up to in B2B?

1027
02:57:21.170 –> 02:57:30.099
Patrick Spychalski: I’d say the two, like, biggest mentors I’ve had were Clay’s co-founders, Varun, Anand, and Kareem. I mean, they are both

1028
02:57:30.390 –> 02:57:40.049
Patrick Spychalski: kind of savants in their space. I think Varun is an operations wizard, Kareem is a product visionary, and seeing them, you know, I started working

1029
02:57:40.130 –> 02:57:51.319
Patrick Spychalski: before I left… before I started the agency, I was working at Clay. I was, like, their 9th or 10th team member, and just seeing them grow this thing into a multi-billion dollar company has been, has been really inspiring. So, yeah.

1030
02:57:52.370 –> 02:57:57.619
Julia Nimchinski: And last one, just for fun, we have one minute. People are asking, would you buy a personal robot?

1031
02:57:58.810 –> 02:58:03.710
Patrick Spychalski: Would I buy a personal robot? I don’t know, I feel like I’d be way too sketched out about that. Like…

1032
02:58:03.880 –> 02:58:08.700
Patrick Spychalski: Like, I don’t know, that’s a little too, dystopian for me. I think I’d be out.

1033
02:58:09.770 –> 02:58:11.109
Julia Nimchinski: Let Elon Musk know.

1034
02:58:11.730 –> 02:58:12.520
Julia Nimchinski: Yeah.

1035
02:58:12.520 –> 02:58:13.200
Patrick Spychalski: For sure.

1036
02:58:13.200 –> 02:58:20.309
Julia Nimchinski: Cool, always great hosting you here, and yeah, the last question is, how can we support you? Where should our community go?

1037
02:58:21.190 –> 02:58:31.689
Patrick Spychalski: Yeah, feel free to just, follow me on LinkedIn, just my name, of course, and then, our website’s thekiln.com, so feel free to check out some of the stuff we’re doing over there, but yeah, that’s all I’ve got.

1038
02:58:32.250 –> 02:58:33.910
Julia Nimchinski: Awesome. Thanks so much.

1039
02:58:33.910 –> 02:58:35.470
Patrick Spychalski: Awesome. Appreciate it, Julia. Bye.