The Agentic Data Layer for GTM

02:07:37.810 –> 02:07:38.950
Julia Nimchinski: Thank you so much.

02:07:39.460 –> 02:07:50.979
Julia Nimchinski: And with that, coming to the Hsc. Stage next Patrick Spitchellski. Welcome to the show. Our favorite Gtm. Engineer. And speaking of N. 8 N. And clay.

02:07:51.300 –> 02:07:52.260
Julia Nimchinski: Welcome.

02:07:52.780 –> 02:08:00.740
Patrick Spychalski: Appreciated. Yeah, probably going to be a little bit more appraising of anything on this one as well. So maybe a bit of redundancy. But hopefully some new stuff, too.

02:08:01.130 –> 02:08:03.759
Julia Nimchinski: What a wild card! Let’s get into it!

02:08:04.180 –> 02:08:18.070
Patrick Spychalski: Awesome. Well, sounds good. Well again. Hello, everybody. I am Patrick Spitchowski, co-founder of the kiln. We are an agency that implements. AI. Tools like Clay and N. 8 N. To go to market teams to create efficiencies ideally drive more pipeline.

02:08:18.070 –> 02:08:38.209
Patrick Spychalski: And we’ve been using a lot of AI agents as of recent. And so I figured this would be a good stage to talk about what we’ve been using AI agents for what tech we’ve been using to set up. And actually, you know, deploy our AI agents. And I figured I could show a few kind of like real life use cases that we’ve been using to, you know, actually

02:08:38.290 –> 02:08:53.310
Patrick Spychalski: put AI to use. There’s a lot of feel like there are a lot of Linkedin posts every day that I see of like crazy workflows. People build AI agents, but I often wonder how many of the like, like, how often those are actually used and so I I just wanted to show a few use cases of

02:08:53.360 –> 02:09:12.220
Patrick Spychalski: kind of real life situations in which we’ve been using AI agents. So I will share my screen. The 1st thing I’m going to do is walk through a few couple like just a few tools that I think are super useful and have been really valuable for us and our clients, and I recommend everybody. Check out, and then I’ll show how we’ve been using those tools. And some more like granular use cases.

02:09:12.820 –> 02:09:14.899
Patrick Spychalski: Okay, you can see my screen, Julia.

02:09:15.570 –> 02:09:19.609
Patrick Spychalski: awesome. Okay, so gonna quickly walk through these pieces of tech

02:09:19.880 –> 02:09:39.659
Patrick Spychalski: and then going to walk through a few ways in which we actually use them. So this 1st piece of tech 1st tool that we use quite a bit internally is called Exa AI. For people who haven’t heard about it. It is an incredible AI agent platform. We use it specifically for lead sourcing for a lot of our clients. It’s really good at list building, and it uses AI agents to do so. So

02:09:40.060 –> 02:09:56.009
Patrick Spychalski: you’ll see kind of shortly how we use it. But for those who have used things like Apollo or zoom info. Think of it as almost the next iteration of that platform, but powered by AI agents. The next tool that we’re planning on using is clay. We use clay a ton

02:09:56.180 –> 02:10:16.999
Patrick Spychalski: for those who don’t know what it is. It’s probably best described as a go to market enrichment and workflow and sequencing platform. So it’s can find really niche data at scale for people or companies can find essentially any publicly available data on a person or a company at scale. And you can use it for a variety of awesome workflows which we’ll dive into shortly.

02:10:17.180 –> 02:10:28.840
Patrick Spychalski: The next one to Julia’s point earlier is N. 8 N. We are huge. N, 8 end users over here. We think it’s an incredible tool, and it’s a really great way for non-technical people and technical people alike to use AI agents easily.

02:10:29.090 –> 02:10:37.620
Patrick Spychalski: And then the last one we’ll show is lovable. Which I think was also mentioned in the the last chat. But again, we cannot say enough good things. It’s honestly an incredible tool.

02:10:37.870 –> 02:10:38.650
Patrick Spychalski: so

02:10:38.880 –> 02:10:46.270
Patrick Spychalski: wanted to quickly. Just walk through in my mind some things that AI agents can do right now. Really? Well, I think

02:10:46.280 –> 02:11:00.609
Patrick Spychalski: what at least what I see on Linkedin are people kind of pushing the boundaries of what AI agents can or cannot do, but oftentimes pushing the boundaries is not what like mid market and enterprise companies optimize, for they optimize for something like guaranteeing the ability to work.

02:11:00.610 –> 02:11:20.869
Patrick Spychalski: And I think, you know, when you’re building out a workflow with AI agents, it’s usually best to build something that works consistently. And so I wanted to show off a few things that we have found work consistently. So the the 1st thing that I wanted to show was just how exit works from a high level. It’s probably the most basic of the tech that I’m going to show it’s very intuitive and easy to use.

02:11:21.211 –> 02:11:40.030
Patrick Spychalski: You literally just give it natural language, and exo will deploy a series of AI agents to go find a list of companies that match that criteria. So I forget if I’ve logged into the right account, but regardless you’ll be able to see the functionality of how this works. I literally just said, I want to find Saas companies based in New York City that recently made an acquisition

02:11:40.030 –> 02:12:07.719
Patrick Spychalski: and exit will go. And the way it works from a high level is it’ll source that initial list of companies that based on basic firma graphic data. So for example, Sas companies based in New York City like you could go figure that out using Apollo, and then it will deploy AI agents to go crawl every single one of the company’s websites, Linkedin profiles, news articles about the business, and then figure out whether they made an acquisition or not, and then programmatically delete the ones that don’t match the criteria you’re looking for and leave you with a larger list of the companies. So

02:12:07.720 –> 02:12:35.390
Patrick Spychalski: it’s been super useful for us specifically for running kind of like micro campaigns to target audiences. So if we’re reaching out to thousands and thousands of people or companies, X is usually not the best way to do it so far, just because it’s we found the AI is a really good at sourcing kind of these smaller lists of businesses. But if you’re trying to find a very specific niche group of companies, I would highly recommend giving it a go. It’s been pretty awesome, so that in the in the cases that we’ve used it for so far.

02:12:35.820 –> 02:12:38.780
Patrick Spychalski: The second thing we’ve used AI agents for a ton

02:12:38.940 –> 02:12:43.920
Patrick Spychalski: are doing automated research on prospects. Like accounts and contacts.

02:12:44.050 –> 02:13:10.550
Patrick Spychalski: So I’m in clay right now. You can see clay has a very spreadsheet esque format, and clay has a feature called Claygen in the platform, and Claygen is their clever, you know, naming for an AI agent that can go do research for you on a specific company. So you can see here, I have a clay table. I have a list of companies. They’re all fortune, 500 businesses, and I’m looking to do research on whether they have made a carbon commitment.

02:13:10.915 –> 02:13:29.119
Patrick Spychalski: You know, explicitly to the public that we can, you know we can find. So I went into Claagent. I was very easily able to prompt Llagent to just give it context, you’re tasked to figure out if a company has a public commitment to low and carbon initiative initiatives, you give it an objective, you give it some instructions, and Clay even has an AI

02:13:29.220 –> 02:13:57.309
Patrick Spychalski: AI that will allow you to generate this prompt for you. So you don’t have to type it out, and then when you press, run, it will go and do research for you, and it’ll even tell you the steps it takes to go figure out how it found this information. So you can research very niche nuanced data on companies that previously were not available to do research on before. This can be super valuable. I mean, you can imagine. If you had to do this for a list of 500 companies manually, it would have taken a ton of time. And now agents can do it for you at scale, really quickly.

02:13:57.710 –> 02:14:15.710
Patrick Spychalski: So that was a really cool use case. I wanted to demonstrate a second use case that we found for AI agents are not only doing research on companies and returning it to you, but actually making high level assumptions and predictions based on that research. So in this table, what I was attempting to do was create a custom software at scale

02:14:15.710 –> 02:14:33.710
Patrick Spychalski: connecting lovable, which we’ll show you later to a clay table, allowing you to create essentially like thousands of pieces of software at scale at one time. But I had to figure out what I wanted to create for each one of these companies, like the idea was to import a list of companies, figure out what piece of software to build for them, and then send a prompt for that software to lovable

02:14:34.089 –> 02:14:55.070
Patrick Spychalski: so I use Claysian once again to go do research on this company, and then actually generate a like a solid software idea on something that would add value to their company and then create a prompt on it. So not only is it doing research on the company, but it’s also making these like high level, abstractive decisions and strategies based on the information found on these businesses. And

02:14:55.070 –> 02:15:04.004
Patrick Spychalski: it ended up being a really cool project. And you know, I was able to, you know, using levels. Api create like dozens of pieces of software at scale. Maybe I can even pull one up if I

02:15:04.450 –> 02:15:07.500
Patrick Spychalski: if they haven’t expired yet. Let’s see here.

02:15:09.080 –> 02:15:13.089
Patrick Spychalski: So yeah, using the prompts generated by Claygen, I was able to create

02:15:13.880 –> 02:15:18.200
Patrick Spychalski: individual pieces of software for each one of the companies. So this is a

02:15:18.300 –> 02:15:28.880
Patrick Spychalski: company called Emitter. We created an AI dashboard for their company. Specifically, I have no idea what Emitter does. I have no idea what this dashboard does. Can’t describe it to you. AI made it. But it’s pretty cool.

02:15:29.450 –> 02:15:50.009
Patrick Spychalski: Okay? So we’ve covered. Now that AI agents can source leads for you. They can do research for you. They can make high level, abstractive decisions based on research for you. So what I next wanted to show is that you can both create your AI own AI agents. Train your own AI agents, and they can actually take actions within tools for you.

02:15:50.140 –> 02:15:50.990
Patrick Spychalski: So

02:15:51.110 –> 02:16:13.440
Patrick Spychalski: this is another thing that we’ve built for a client of ours where we’ve built an Mcp client on top of Exa. So you saw Exa and what it does. But we wanted to figure out a way for our client to communicate with Exa in slack, and have an AI agent essentially prompt exa to do things for us. So in theory and actually in practice, now that this workflow is active. What this workflow will do is you can send a slack message saying, Hey.

02:16:13.460 –> 02:16:26.140
Patrick Spychalski: slack, bot! I want you to go find me a lit like a list of 50 companies that match this criteria, and an AI agent that we’ve prompted will essentially process the in the Json prompt that we’ve parsed out from back here.

02:16:26.670 –> 02:16:30.960
Patrick Spychalski: and it’s trained on an Mcp client that we’ve built specifically on top of Exa.

02:16:31.270 –> 02:16:46.799
Patrick Spychalski: And it has a memory. So it’s able to figure out. You know what other prompts that we’ve sent in in the past, and we’ve connected an open AI Chat model to it, and then it will send back via slack a list of all these companies. So you now have a lead sourcing model built into slack that can find you contacts and accounts.

02:16:46.799 –> 02:17:11.319
Patrick Spychalski: and you don’t ever have to leave the slack platform. So the the main thing I wanted to highlight here is the fact that AI agents, if you build an Mcp like a model context, protocol can actually interact with tools themselves. So instead of just doing research and scraping the Internet for you, you can actually have it go actually like, act upon tools like you can make a an AI agent that acts in notion or acts in slack, or even sends out emails using an email sequencing tool. And so

02:17:11.573 –> 02:17:23.399
Patrick Spychalski: super powerful if you’re able to create these. And and this is by the way, in Nan which I would highly recommend. You can see some of these workflows can get. This is even like a even a crazy workflow for any then, but it can get pretty wild the stuff you can build in the tool.

02:17:24.129 –> 02:17:32.799
Patrick Spychalski: So the very last thing I wanted to show that AI just can do. Actually, I think the second to last thing is that they can also

02:17:33.080 –> 02:17:45.189
Patrick Spychalski: pretty much replace entire aspects of a company or a job that you would otherwise have to hire for. And so you can see, this is a very basic N. 8 n. Workflow. But what what is actually underlying this workflow is somewhat complex. So

02:17:45.190 –> 02:18:06.680
Patrick Spychalski: we used to have copywriters on our team and a problem with having a copywriter on the team was mostly the fact that when a client wanted copy written for a campaign that was urgent, we would have to wait a really long time to get that copy back, and the copywriter. Oftentimes, in the beginning days of our agency was me, and I didn’t have the bandwidth to be writing copy for these clients. I had other things to do. And so it would end up getting done a lot slower.

02:18:06.709 –> 02:18:09.760
Patrick Spychalski: So I was able to train this AI agent

02:18:09.790 –> 02:18:32.760
Patrick Spychalski: on the copy examples that I created, and it actually continues to train itself by sending all this, all the copy that it writes to an air table, and then retrieving the memories aggregating it, merging it, and then just constantly training itself, using an air table database that we’ve created with copy. So if you’re able to prompt these AI correctly, you can replace entire tasks or entire sections of a business

02:18:32.830 –> 02:18:41.890
Patrick Spychalski: that otherwise would have cost money. And you’ve had to hire people for so this has been a ridiculously valuable workflow for us, and has replaced a lot of our general copywriting functions.

02:18:42.405 –> 02:18:49.240
Patrick Spychalski: Last thing I wanted to show is that AI agents can build apps for you. It’s like it truly is

02:18:49.459 –> 02:19:14.960
Patrick Spychalski: like incredible. The stuff you can build in lovable. This is a super basic tool that I’ve built. I mean, it gets way more complex than this. But I built a clay table. I didn’t want the client, I mean. In this case it was just us, but like the I was building out an Mvp. Because essentially a client of ours was like, Hey, this clay table is really cool, like I’m glad you built it for us, but it’s really complex, like we. I don’t want to have to go into this clay table and poke around with it. I feel like I’m going to mess something up.

02:19:15.000 –> 02:19:27.320
Patrick Spychalski: And so I was like, Okay, I’m just gonna build an like a web app and lovable, that allows you to just throw in a Linkedin profile. URL. It’ll send that URL to the clay table, and then it’ll return all the information that the clay table yields you. So this is like a research tool that we built

02:19:27.626 –> 02:19:33.009
Patrick Spychalski: and you can just use natural language to do this, and AI agents can pretty much. Take over and build entire

02:19:33.419 –> 02:19:43.899
Patrick Spychalski: apps and structures of code for you. And we’ve done everything that I’ve shown you. We’ve actually done for clients like they’ve paid us to build it for them, and it’s worked for them and added a lot of value. So

02:19:44.070 –> 02:19:50.939
Patrick Spychalski: hopefully, this is valuable, and I will, you know, leave it to any questions, if people have any, or if we even have time for that, I don’t know.

02:19:51.820 –> 02:20:00.749
Julia Nimchinski: Awesome. Thank you so much, Patrick. And yeah, one of the questions, can we talk about your clients and some cool use cases or more tangible

02:20:00.960 –> 02:20:03.720
Julia Nimchinski: Roi stories, early.

02:20:04.170 –> 02:20:27.169
Patrick Spychalski: Totally so our clients, generally, just for context, are mostly a mid market to enterprise businesses and a lot of like high growth Sas companies. So think of, like the AI businesses that you see in the news a lot. A lot of those type of companies are the ones we generally seem to work with. So one great example of this is we were working with a company that was scaling really quickly, and they had hired 200

02:20:27.170 –> 02:20:38.759
Patrick Spychalski: Bdrs over the past like year and a half, and these Bdrs were all essentially doing the same thing, which was researching prospects for half of their day, and then reaching out to prospects for half of their day.

02:20:38.940 –> 02:20:42.059
Patrick Spychalski: and they came to us because they

02:20:42.210 –> 02:20:59.459
Patrick Spychalski: correctly had the assumption that the research they were doing was super inefficient, and was taking more of their time than it should. These Bdrs were doing like 2 to 3 h of research a day, and if you extrapolate that over 200 Bdrs, you can think of how much time was actually being wasted cumulatively across the sales org.

02:20:59.750 –> 02:21:18.399
Patrick Spychalski: And so what we did is we interviewed everybody, from like the head of sales down to like 5 to 6 of their top Bdrs and asked them, What is the process you’re taking when you’re researching prospects before you reach out to them. And they all gave us kind of disparate answers. But we were able to find commonalities across all of them. And then we use those commonalities to build an AI agent

02:21:18.400 –> 02:21:37.720
Patrick Spychalski: partially in clay, partially in N. 8, n. That allowed us to create research reports for these Bdrs. So, in short, they didn’t have to do any of the research they were previously doing. They could spend all of their time selling. So this essentially doubled their their sales output and sales efficiency immediately, because we were able to replace the research function that these Sdrs or your Bdrs were previously doing

02:21:38.079 –> 02:21:48.149
Patrick Spychalski: and so that was like a really great real world. Use case of how to build them. Another great one is, as I mentioned before. We build a lot of clay tables and nan workflows.

02:21:48.330 –> 02:22:07.739
Patrick Spychalski: Both of those things are very difficult to actually use. And some of these don’t involve AI agents like maybe we’re just building like an inbound enrichment tool, or we’re cleaning up their salesforce or hubspot like those things don’t involve agents all the time, but nonetheless are difficult to use. And so, being able to build things like the Mcp, I just showed you that wraps

02:22:08.023 –> 02:22:22.790
Patrick Spychalski: for example, Exa, to source leads for you or wraps a clay table we’ve built allows the systems we build to actually be usable, which is also super helpful. The last thing I’ll say is, we’ve built some internal AI agents to help with writing outbound copy and running campaigns for us.

02:22:23.290 –> 02:22:32.049
Patrick Spychalski: Which I mean, I’d say, for, like a Gtm. Agency, the most tangible metric or return would be pipeline generated. And we’ve

02:22:32.390 –> 02:22:33.740
Patrick Spychalski: generated, I’d say.

02:22:33.990 –> 02:22:48.230
Patrick Spychalski: with at least with the AI agent built outbound platforms at least a couple of 100 million in pipeline for our clients cumulatively and and significantly more. Just across all the outbound systems we’ve used that incorporate AI as a whole. So it’s it’s been. Yeah. Truly.

02:22:48.470 –> 02:22:51.040
Patrick Spychalski: a massive value. Add, it seems, for our clients.

02:22:52.580 –> 02:23:04.429
Julia Nimchinski: Super inspiring. There is one question I’d like to address here when you shift budgets in real time. How do you avoid killing a long term play, just because early signals underperform.

02:23:05.220 –> 02:23:07.920
Patrick Spychalski: Sorry. Could you say the 1st part of that when you ship? What in real time.

02:23:08.130 –> 02:23:16.390
Julia Nimchinski: Yeah, when you shift budget in real time. How do you avoid killing a long, long term play just because of, you know, some mismatch with a

02:23:16.680 –> 02:23:17.960
Julia Nimchinski: early signals.

02:23:18.550 –> 02:23:33.709
Patrick Spychalski: Yeah, so I would say, with all the clients we have like again, I’d say most enterprise clients, for the most part, are quite conservative in their decision making relative to a startup that’s fast moving like, move fast, break things sort of startup. And so almost everything we do is like.

02:23:34.078 –> 02:23:55.129
Patrick Spychalski: I’d say subjectively, pretty slow. When we build them out. So like this. The examples I’m showing you in many cases are the result of us working for with a client for like 7 to 8 months, and then slowly weaning themselves off of another long term system. So, for example, that thing I just talked about where we built the research model and applied it to all of those Bdrs.

02:23:55.130 –> 02:24:04.230
Patrick Spychalski: We only had, like one or 2 Bdrs using it@firstst and then we’re like, essentially slowly weaning everybody off of it until they were using the new system that we had built.

02:24:04.609 –> 02:24:25.930
Patrick Spychalski: And and there’s 2 reasons for that. The 1st reason, of course, is just to to prove that the system actually works and is valuable to them. If it’s not. Then there’s no reason to wean everybody off of it. And the second thing is because oftentimes like making a massive shift or a massive change in an org and kind of shake things up short term. And we don’t want that to happen so like slowly having people use the system, I think, is pretty important, especially in larger organizations.

02:24:27.640 –> 02:24:31.099
Julia Nimchinski: Our audience wants to know if there’s any resources.

02:24:31.400 –> 02:24:39.580
Julia Nimchinski: that where they could just take the basically next step to learn more about any and workflows. Some basic workflows.

02:24:40.110 –> 02:24:43.269
Julia Nimchinski: anything related to complex integrations.

02:24:43.610 –> 02:24:44.799
Julia Nimchinski: And yeah.

02:24:44.800 –> 02:24:58.989
Patrick Spychalski: Yeah, so the 1st place I would look, and I still go on here and watch. All of this stuff is this guy Brandon, Charles, and he has a Youtube channel. I don’t know if maybe I can throw it in the chat. I can. I’ll throw it in the

02:24:59.861 –> 02:25:03.720
Patrick Spychalski: I can only throw it in like the host and Panelist section. I’m not sure how.

02:25:03.720 –> 02:25:04.749
Julia Nimchinski: I’ll pass it along.

02:25:04.970 –> 02:25:14.990
Patrick Spychalski: There’s this Guy Brandon, Charles, and he is an N. 8 N. Wizard. He is a cursor wizard. He’s ridiculously good. He’s a Gtm. Engineer as well. And he has an incredible Youtube channel. I would recommend watching all of his videos.

02:25:15.110 –> 02:25:21.289
Patrick Spychalski: If you’re interested in learning more about Clay, specifically, they have a free online university called Clay University, that I would check out

02:25:21.877 –> 02:25:36.820
Patrick Spychalski: and there’s another person if you’re looking more for like outbound related how to use AI properly. There’s a guy named Eric Nooslowski who I recommend checking out as well. I will link his Youtube in the chat below. And you know, Julia feel free to pass it on. So

02:25:37.155 –> 02:25:51.899
Patrick Spychalski: in short, honestly, Youtube is the best way to learn about this stuff. If you have a question about a specific thing in Nan, or maybe and just even want an n 8 n, like high level tutorial. There are so many people on Youtube talking about it, and I found the most value to be there or read it.

02:25:53.770 –> 02:25:57.079
Julia Nimchinski: Super helpful Patrick, for all the executives watching.

02:25:57.738 –> 02:26:06.290
Julia Nimchinski: The hardest piece that, and we want to address it like literally, every session is the transition to a gentic session scaling?

02:26:06.871 –> 02:26:10.660
Julia Nimchinski: What would be your recommendation? Where do they start?

02:26:11.210 –> 02:26:12.679
Julia Nimchinski: How do they even start.

02:26:13.300 –> 02:26:34.779
Patrick Spychalski: Yeah, so this is how we do it with our clients, and it seems to be the most well received way of doing so is find the most basic possible task you’re doing manually, that you think could be supplemented or completely automated by AI agents and then go build that task out 1st and then continue to do that in sequential order. I think a lot of people get really excited when they see Linkedin posts

02:26:34.780 –> 02:26:47.769
Patrick Spychalski: about like AI. This AI agent replaced my entire marketing department, and people get like, really excited. They’re like this is going to save me so much money. But it’s I, in my opinion, significantly. Better to go to the most basic use case of using AI agents.

02:26:47.770 –> 02:27:03.939
Patrick Spychalski: Figure out how to use those. And you’ll figure out a few things. The 1st thing is the viability of AI agents to replace a certain task. I think people hype it up a lot like I don’t think you replace your entire marketing department with an AI agent right now. I think that’s a little past what it’s capable of doing. So you’re able to find out like kind of the extent as to what these things can do.

02:27:04.357 –> 02:27:27.669
Patrick Spychalski: And then you can also figure out what tools are best for building these systems out. Like, the the worst thing I think you can do is go on Linkedin or Youtube or Reddit, and then just get like the top 10 best tools, I mean. Again, you should go on them. But then get the top 10 best tools and just buy them all like like buy like enterprise plans for all of them. And just be like, Okay, let’s go start automating everything like

02:27:27.830 –> 02:27:48.360
Patrick Spychalski: I would use it for education purposes. But don’t go. And just like make massive brash decisions. And implement AI agents everywhere. I think it’s like use. Make it a slow process. These tools are changing constantly by the time you go from idea to execution. The tool you’re using probably came out with like 20 different updates. So like, you know, I would just say, start start small.

02:27:48.890 –> 02:27:51.894
Julia Nimchinski: Definitely. And one more question here from Steven.

02:27:52.450 –> 02:28:00.429
Julia Nimchinski: that’s it. What tweaks to make for non saas companies like manufacturing that need marketing Bd sales efficiencies.

02:28:01.230 –> 02:28:13.579
Patrick Spychalski: Yeah, totally. I think a lot of the systems we build actually, apply directly to non saas companies. Like, there are some things that remain the same across industries, for example.

02:28:13.730 –> 02:28:20.770
Patrick Spychalski: like every salesperson, is going to be doing research on their prospects. Every salesperson has to do some sort of enrichment, find contact information and do outreach.

02:28:21.201 –> 02:28:41.779
Patrick Spychalski: You know, every salesperson is probably gonna have to continue running an email threads. Like to continue talking to people and prospects day to day. So I would actually argue a lot of the systems we build or can be used. Kind of industry, agnostic. I think Sas companies are the ones using them the most right now, because they’re generally the most like progressive and move the quickest. But

02:28:42.220 –> 02:28:48.309
Patrick Spychalski: I think actually, the biggest impact for AI agents can often be in industries that aren’t in Sas, because.

02:28:48.430 –> 02:29:07.499
Patrick Spychalski: you know, we’ve just talked to these like massive, publicly traded companies before that are still operating kind of in, like the Stone Age from a Gtm. Standpoint, because they haven’t changed anything in 20 years. So you know, even automating one aspect of their sales award will probably yield a you know, millions and and value return. So I’d recommend just following the same path as everybody else in the AI agent space.

02:29:08.290 –> 02:29:15.609
Julia Nimchinski: Thank you so much, Patrick. Always super insightful. And do you have a newsletter? What’s the best next step for our community to support you?

02:29:15.780 –> 02:29:25.900
Patrick Spychalski: Yeah, I would just say, feel free to add me on Linkedin. There’s a link to my newsletter on my Linkedin profile as well, but try to post as much as I can about the stuff we’re building over there. And yeah, that’s all I got.