The Autonomous Marketing System

01:32:21.120 –> 01:32:32.150
Julia Nimchinski: Next up. We are proud to present Metadata’s Gil Alouch, founder, and CEO and Lisa Sharapada, Vp of AI and Gtm. Strategy.

01:32:32.710 –> 01:32:39.060
Julia Nimchinski: and we are going to dive deep into the autonomous marketing system. Welcome to the show! How are you doing.

01:32:40.270 –> 01:32:44.500
Gil Allouche: Hi, Julia! Great to be here. Thank you for for the stage. Hi! Everyone.

01:32:45.380 –> 01:32:50.239
Julia Nimchinski: Our pleasure. What’s what’s new in the world? World of agentic? AI.

01:32:52.530 –> 01:32:54.300
Gil Allouche: Juliet at Lisa. Do you want to say Hi.

01:32:55.920 –> 01:33:07.499
Lisa Sharapata: Yeah, hey, everyone. Thanks for having us, Julia and yeah, lots to show. So we can definitely jump right in. Is it? Okay, if I just pull up my slides.

01:33:07.500 –> 01:33:08.619
Julia Nimchinski: Yeah, definitely.

01:33:09.190 –> 01:33:10.140
Lisa Sharapata: Okay, great.

01:33:11.360 –> 01:33:14.420
Lisa Sharapata: Let me share my screen here.

01:33:17.860 –> 01:33:20.280
Lisa Sharapata: let’s see if I can just make it a slideshow.

01:33:23.100 –> 01:33:24.640
Lisa Sharapata: Alright, how’s that?

01:33:25.760 –> 01:33:26.660
Julia Nimchinski: Looks great.

01:33:26.880 –> 01:33:40.290
Lisa Sharapata: Alright! Alright! So real quick. We can just give a a little intro here on Gil and I Gil, do you wanna just give a little overview of your experience.

01:33:40.500 –> 01:34:07.039
Gil Allouche: Yeah, totally. Hello, everyone. My name is Gil Alush. I’m the founder and CEO of Metadata. I’m originally a software engineer, robotics and AI. And then I went into marketing. I did 3 stints as a Vp of marketing in b 2 b companies, and then I started a company that at least tries in theory to replace myself so that you don’t have to hire a technical marketing head or a technical marketing agency. And that’s what we do. We do advertising on autopilot.

01:34:09.540 –> 01:34:20.679
Lisa Sharapata: Alright. Well, and I’m Lisa been in the b 2 b tech space for quite some time now, and a 2 time. Cmo, and I’m excited to be here with Gil and metadata and

01:34:20.770 –> 01:34:50.360
Lisa Sharapata: helping to kind of lead the charge, and also figure out how all this works, so that I’m the 1% of marketers who has a job. And all this is done. So yeah, with that, I want to pass it over back to Gil. And we’re gonna just kind of start going through some of the use cases that we’ve been working on and building and just really talking about kind of this, this real shift that I’m seeing isn’t just about using AI. It’s

01:34:50.360 –> 01:34:55.489
Lisa Sharapata: how do we let AI do the work? So from there I’ll pass it off to Gil.

01:34:56.830 –> 01:35:01.740
Gil Allouche: Awesome so for everyone’s context, I gave this presentation for the 1st time, maybe a month back

01:35:02.040 –> 01:35:08.179
Gil Allouche: on a yacht with a group of maybe 15 Cros and Cmos and

01:35:08.450 –> 01:35:14.220
Gil Allouche: the feedback was so good that we decided to start propagating this message out there.

01:35:14.600 –> 01:35:30.210
Gil Allouche: You know, I started by showing a funny how it started, how it’s going, meme, because I think everyone can relate to this. Maybe a few years ago, when we were thinking about AI and all the innovation and singularity and all that good stuff. We’re thinking about flying cars.

01:35:30.630 –> 01:35:40.490
Gil Allouche: And and, you know, sitting in a beach, and most of the work is being done for us. Well, in reality we’re hooked to Linkedin with insane Fomo.

01:35:40.600 –> 01:36:02.190
Gil Allouche: And all these prompts and cheat sheets and all the things that are just making us more and more nervous. And I can tell you that as founders and Cmos, we’re asked to do a lot more with a lot less. And it’s really we’re at the Fomo stage. So if we move to the next stage, I’ll give a next slide. I’ll show a little bit of an example. 1st of all.

01:36:02.470 –> 01:36:05.530
Gil Allouche: why, why should we care about this?

01:36:05.640 –> 01:36:15.319
Gil Allouche: And the main reason that you should care about this is because we all of our lives are going to change. And I think sometimes, Lisa, if you want me to take over the slides, let me know I’m happy to

01:36:15.440 –> 01:36:17.840
Gil Allouche: 2 30. You don’t don’t have to. Yeah.

01:36:19.351 –> 01:36:22.750
Lisa Sharapata: No, we’re good. Let’s keep going. I’ll try to keep up with you.

01:36:22.750 –> 01:36:24.919
Gil Allouche: No worries, so I’m still the second slide.

01:36:26.555 –> 01:36:27.100
Lisa Sharapata: Okay.

01:36:27.610 –> 01:36:29.420
Gil Allouche: Hi so I think one of the

01:36:30.030 –> 01:36:32.540
Gil Allouche: no, the the one the one right after.

01:36:33.010 –> 01:36:33.434
Lisa Sharapata: Okay.

01:36:34.090 –> 01:37:01.330
Gil Allouche: So you know, if you ask yourself why you should care. I think many people talk about AI in future tense, and that’s because of comfort for psychological comfort. We’re trying to push this into future, maybe procrastinate a little bit. But the truth is that the vast majority of the things that you thought are going to be possible in the future are already possible. 2, 3 months ago I remember myself in July last year, when Claude released cloud desktop or about to release the browser control.

01:37:01.380 –> 01:37:22.359
Gil Allouche: I realize that things are going to be not a little different, not like 20%, 30%, 50% difference. They’re going to be like 99% different meaning that the vast majority of the things that we know are going to cease to exist. And we’re going to have to reinvent ourselves if we want to not only keep a job, but be ahead of the curve in terms of enjoy our job not be like archaic

01:37:22.370 –> 01:37:47.860
Gil Allouche: and maybe even get paid 10 times the amount of money we get paid today because we’re going to be indisposable in the market. And so I think that’s the main reason you should care about this, because this is happening right now. And it’s really a decision. It’s a binary decision of whether you’re part of it, and you know exactly what’s going on and how to use it to your advantage, or you’re behind stressed because you don’t even know how to start. And getting started is usually a commitment. You do to yourself.

01:37:47.970 –> 01:38:15.630
Gil Allouche: 5 min a day, 10 min a day doing something has compound interest. Right? It’s just like working out or doing anything else in life. Instead of thinking about being Arnold Schwarzenegger. In 3 months you can do a 5 min workout, 10 min workout every day with AI and be smarter and start seeing how it changes your life every time you have a question, instead of going to Google, going to Chatgpt every time you have a piece of data or something about yourself that you want to. Ingest into an AI model you go and you upload

01:38:15.630 –> 01:38:28.890
Gil Allouche: a context or some data set or something of that. So that’s the main reason you should care about this so that you can be really ahead of the time and be part of the change versus part of the resistance, the inevitable

01:38:28.890 –> 01:38:50.680
Gil Allouche: change that is going to happen, and of course, as a founder and CEO, I have a personal interest in making my business profitable. The days of building a business, and we raised a good amount of money from metadata. The days of growing at all costs are gone. My investors, the Board, are asking me every meeting about profitability, about the rule of 40 about the

01:38:50.870 –> 01:39:11.130
Gil Allouche: revenue per employee. Right? You look at the companies like lovable right? There are still a super small team with double digit 1 million dollar arr. This is the new, the new way of doing things, and and there is no way of doing it without really adopting your AI in in the highest, highest percentage.

01:39:11.310 –> 01:39:15.969
Gil Allouche: And then, last, but not least, you want to scale without scaling resources. Right? You want to scale from.

01:39:16.130 –> 01:39:38.560
Gil Allouche: I don’t know. Depending where you are in your business 1 million to 10 million without hiring 15 people more, which is the usual formula of a Vc. Backed company. You want to be able to have a 20 people company with 50 million dollars revenue. That is possible today. If you are AI 1st meaning every time everything that you do, you build it with a workflow, an agent or a system that is AI. First, st

01:39:39.700 –> 01:39:47.230
Gil Allouche: let’s move on to the next to the next slide. This is just a small example of those prompts I was referring to. I think you all know this. If you’re in Twitter.

01:39:47.250 –> 01:40:16.460
Gil Allouche: I guess X or Linkedin, this is all you see? Right. Put cheat sheet on the comment to get my best prompt for Chatgpt. If you’re a vibe marketer, put here and I’ll share with you my N. 8 n. Workflow, or my make.com workflow, and what it really does, at least, if you have adh like me, you will save it. You’ll bookmark it somewhere. You’ll take a screenshot of this nice cheat sheet, and you’ll save it in some folder that you’ll never open again.

01:40:16.460 –> 01:40:29.960
Gil Allouche: You just do it for a psychological comfort that you’re doing something that you’re taking some step. But the truth is, you’re not taking any step. You’re just procrastinating. The way to get ahead is to spend X amount of minutes or hours every day

01:40:29.970 –> 01:40:45.980
Gil Allouche: implementing and adopting this new technology that is called artificial intelligence. It is not super futuristic. It’s not something you can’t grasp. It is absolutely graspable. It will become a household tool it already has in many ways. You just haven’t realized it.

01:40:45.980 –> 01:41:09.040
Gil Allouche: but it is time now. This is like the 19 nineties, right where the Internet just came about. Some people laughed at it right? The blockbuster. CEO laughed at Netflix. But this is becoming a reality in a second. This is going to be the number one productivity tool, the number one data analysis tool, the number one action-based tool for you, and the earlier you adopt it, you’re going to be just that much ahead of the curve.

01:41:09.830 –> 01:41:11.499
Gil Allouche: So let’s get started. Yeah.

01:41:11.500 –> 01:41:28.580
Lisa Sharapata: Gil. I thought it was interesting what Tuba said on a prior session, about how the more you’re prompting honestly like the more you’re you’re probably getting in the way of letting AI be as powerful as it can be. So with that I’ll pass it on.

01:41:29.190 –> 01:41:37.150
Gil Allouche: Yeah. And there’s definitely some, some truth to it. The models are still learning. So, depending on what work you do if you don’t give it enough context.

01:41:37.180 –> 01:41:47.280
Gil Allouche: it may just hallucinate right? And the problem with Llm, the problem with AI, one of the biggest problems I have with it is that it’s really good at the talking part.

01:41:47.310 –> 01:42:11.240
Gil Allouche: It may give you a response that is so convincing. And then you catch an error, and it’s like you’re right on. You’re totally right, my bad. I’ll not repeat that again, and then you ask it again without context, and it’s going to give you the same error again with convincing language. And so it’s up to you to keep AI honest and prompting and biz logic. A set of instructions is still the best way today to use the general AI.

01:42:11.310 –> 01:42:15.590
Gil Allouche: But of course the models will evolve. It will become more and more

01:42:15.810 –> 01:42:24.739
Gil Allouche: promptless. I do think that will happen eventually. So just a little bit in terms of the old way. New way. Right? That’s an easy way to talk about paradigm shifts. So

01:42:25.050 –> 01:42:47.949
Gil Allouche: the old way is having a pyramid of hierarchy where you have a big boss at the top, and he tells the employees the teams what to do, and every person has one role that they’re best at. That is the old way of doing things. You have a human. You ask the human to create a blog post for you, or creative, or a copy, or

01:42:48.460 –> 01:43:00.189
Gil Allouche: or run a campaign for you, or do some data analysis, or, you know, structure the data in your Hubspot, whatever it is, and you wait a day. You wait 2 days. You wait a week or 2 weeks and you get your results back. There is some feedback loop.

01:43:00.800 –> 01:43:10.690
Gil Allouche: That’s the way it’s done today. The way it’s done by new companies today and by all companies tomorrow is, there’s going to be a person who is a supervisor, a person who has good context, flat org.

01:43:10.850 –> 01:43:16.240
Gil Allouche: They have all the context about what’s going on in their business and outside, and they have

01:43:16.240 –> 01:43:42.159
Gil Allouche: a custom Gpt. It can be a custom. Gpt it can be a Claude can be any one of the Ji vendors, and that custom gpt the reason it has the word custom in there is because it’s trained on all of your data and has all of your context and has all of your preferences for communication right like when I communicate with Claude or with Chatgpt. There is a very concise way that I like Chatgpt to talk to me. You know I don’t like

01:43:42.170 –> 01:43:51.750
Gil Allouche: a lot of fluff in the conversation. I like, you know, fact-based conversation. If if it detects errors, or if I’m asking something that is unreasonable. I want it to let me know versus

01:43:51.810 –> 01:44:09.300
Gil Allouche: bullshit me around. And then at the end, I see that I’m getting the wrong results. That’s my personal preference. It also has all of my data. It has my Crm data. It’s my Gmail, it has my calendar. It has my marketing automation. All of the advertising data. It has my website metadata. So it has so much information

01:44:09.300 –> 01:44:22.969
Gil Allouche: that it already processed. So every time I ask something, I ask the Gpt to do something for me. It has all that context, all the historical, all the back and forth communication we had, and creating a custom. Gpt.

01:44:23.110 –> 01:44:34.689
Gil Allouche: Will literally take you a minute in a minute from now you can have a custom Gpt. In 2 min from now. You can have a custom Gpt with your website, metadata in 5 min from now you can have it with all the advertising.

01:44:34.990 –> 01:44:55.129
Gil Allouche: all the ads that you ever ran, so that you can run so that you can ask it to create a new ad creative for you. So it is really not too complicated to do it. And once you have it, you’ll see you become a supervisor of this nice technology that does things for you doesn’t just tell you this is what you can do. Follow these instructions, and you’ll get it done, but it will actually go ahead and do it for you.

01:44:56.810 –> 01:45:00.090
Gil Allouche: We have some examples here. Let’s move to the next to the next slide.

01:45:00.450 –> 01:45:19.380
Gil Allouche: So there are 3 things I’m going to talk about. We’re going to talk about in this. In this presentation. The 1st one are AI workflows. What you see here, the screenshot here is, make.com make dot com is like, imagine Zapier, or if this and that, if you ever knew those technologies before. It was a kind of a automation technology. You connect Hubspot, you connect

01:45:19.380 –> 01:45:35.260
Gil Allouche: Gmail and you can say, Hey, if I’m getting a visitor from here, then send them an email from my Gmail super simple to understand. Right? Here you have the same concept, but with AI 1st integrated in it, meaning you can tell it to do things.

01:45:35.630 –> 01:46:02.490
Gil Allouche: You can tell it to figure out the automation and do the automation. 100 X deployed 100 X without writing any piece of code, not even one line of code, and I’ll go into that in a second. The second one is a custom. Gpt. We talked about it a little bit. I’m also introducing a new term called Mcp and Mcp is essentially a way for you to bring in all the tooling all the capabilities from the software that you have today. Bring it into your

01:46:02.490 –> 01:46:20.420
Gil Allouche: favorite AI meaning I can create. For example, I can connect an Mcp server from my Gmail, and the moment I do that it only takes a few minutes, and the moment I do that I can go into Claude, or I can go into Chatgpt. And I can send an email, not just

01:46:20.710 –> 01:46:38.669
Gil Allouche: review my inbox, not just see. What should I do today but actually take action on it. I can draft emails. I can delete emails. I can add tags. I can remove tags. That’s the power of an Mcp giving Chatgpt the tooling and the ability to take action on your behalf.

01:46:39.110 –> 01:47:00.920
Gil Allouche: and then copilot copilot is a common term that many Saas companies are now introducing, which is, instead of going into your favorite software and clicking, clicking, clicking everywhere. You have a small copilot where you can ask it to do what you want to do at that time. You want to set up a campaign. You’re about to go into a board meeting, and you need some information you’re going to. I don’t know a big event.

01:47:00.920 –> 01:47:14.430
Gil Allouche: and you want to take some sort of action, you can tell it what you want to do, and it will figure out how to use the software on your behalf to get all the actions done and report back just like an employee who is kind of an expert in that software.

01:47:15.260 –> 01:47:17.870
Gil Allouche: So let’s dive into Ir. Workflows first.st

01:47:18.120 –> 01:47:40.980
Gil Allouche: This is what we’ve done here in Metadata. I won’t go into the numbers, but I can just tell you that we have done quite a bit of work in terms of adopting AI and the way we’ve done it is a brutal way, and we had to do it, and that is reducing the manual work. The way we’ve done reducing manual work is we’ve reduced the team. We’ve reduced the team because we want to become profitable, and we need it to become profitable.

01:47:41.000 –> 01:47:54.200
Gil Allouche: And that was, you know. Sometimes scarcity drives creativity right? And so the way we’ve done it is by having a need right from the outside market, from the investment, from the board. To be profitable.

01:47:54.200 –> 01:48:12.169
Gil Allouche: we had to take these actions. And in order for us to figure out, how do we actually automate the manual work that we’ve done? We, together with the team, right? So we have a team that we interviewed want to make sure that they’re down for that challenge. We started a process where

01:48:12.280 –> 01:48:14.910
Gil Allouche: the team members themselves started taking

01:48:15.090 –> 01:48:26.119
Gil Allouche: screenshots, start taking loom videos, loom recordings, start writing sops start and operating procedures of everything that they do.

01:48:26.490 –> 01:48:46.419
Gil Allouche: You can see all that goodness here in the very overwhelming spreadsheet here that I took a screenshot of, and what comes out of that is a step by step, sop for every process. So let’s say, one of our campaign manager wakes up in the morning. She makes her coffee. She goes to the computer and she has 50 accounts

01:48:46.420 –> 01:49:01.850
Gil Allouche: for the accounts that she manages. She goes through the different systems. She figures out okay? Which accounts are doing well, which accounts are not doing well. What are the insights? What are the actions that I need to take? And she starts going through all of these activities. Well, what if you could create an agent

01:49:01.930 –> 01:49:21.230
Gil Allouche: that you can make your coffee? And by the time you finish your coffee you don’t have to get into a 4 h work, Marathon. By the time you finish your coffee you can start sipping it and watch the AI agents open up 100 browser, and go through all of your accounts. The same standard procedure that you do manually is now done for you, and the actions

01:49:21.250 –> 01:49:39.049
Gil Allouche: depending on your comfort level, you can either approve the actions. It will show you the actions that need to be done, and you can confirm them, or if your comfort level is high or your risk tolerance is high, you can just let it take the actions for you right away. That is just one of the agents that we created here. So

01:49:39.050 –> 01:49:50.800
Gil Allouche: we recorded the screen. We recorded the screenshots. We took the videos. We took the Sops, we broke them down into a human in the loop. That’s what AI means, which is a process in which AI is driving.

01:49:50.800 –> 01:50:11.609
Gil Allouche: But there is a human somewhere needed. It can be a small task like a Captcha that is trying to identify that you’re not a bot that you have to join in as a human and bypass it, or it could be a much more complicated task where a human is just more capable in. And it could be many tasks, because humans are more capable in many of the AI tasks. Still, today, especially creative tasks and context tasks.

01:50:12.310 –> 01:50:24.889
Gil Allouche: And then it hands over the control back to AI. So that’s the 1st thing you can do. The 1st thing you can do today is install one of those software or put together a step-by-step

01:50:25.040 –> 01:50:32.650
Gil Allouche: human in the loop process and sop for your most demanding tasks. Think about tasks that are taking the longest

01:50:33.050 –> 01:50:50.559
Gil Allouche: that are producing the most output that are yielding the most productivity for you, and you don’t like the most the things that you know you are not really meant to be like. If you still wake up at 2 Am. In the morning and are batch changing 100 Utm tags. Yes, this is a task. You can automate, for example.

01:50:51.490 –> 01:51:19.610
Gil Allouche: and then once you have the sop you have 2 choices. You can either learn how to use, make.com and start engineering those workflows. It is not complicated. It’s not easy. It’s not just like it won’t take you 5 min. It will take you hours to learn it, but it’s worthwhile. But you also don’t have to. You can go into one of the general AI tools available today. Like, for example, Manusim is one of them. There are many others that can give you recommendations. You can just upload to manus

01:51:19.820 –> 01:51:24.439
Gil Allouche: a video with a with a human in the loop spreadsheet with an sop.

01:51:24.640 –> 01:51:49.169
Gil Allouche: And once you do that it will learn the process by itself. It’ll take over your screen if you’re using the open source, or it will set up its own desktop. If you’re using the cloud version and it will go through the task you can see here. I installed my own mask on my computer. You can see this is the visual. These are the lenses of the AI. It takes the screen from Google. It breaks down everything that is there on the browser, and you can ask it to take action. I remember 5 months ago

01:51:49.400 –> 01:52:05.200
Gil Allouche: I created an automation with Manus open source that logged in into the software that we created and took actions. And it was quite mind-boggling to see it happen. You can do it today. It’s much more evolved even today. So it’s not very complicated. The learning curve for tools like Manus

01:52:05.350 –> 01:52:07.949
Gil Allouche: is days at at Max.

01:52:09.620 –> 01:52:10.479
Gil Allouche: Let’s move on to the.

01:52:11.480 –> 01:52:12.610
Lisa Sharapata: Right.

01:52:14.280 –> 01:52:20.409
Gil Allouche: Second one I wanted to talk about is the custom. Gpt. Custom Gpt, I can’t speak enough about how

01:52:20.910 –> 01:52:41.299
Gil Allouche: how much low hanging fruit are available right? There is getting to the 1% most advanced. AI, okay, it’s there. But there is the 1 h task you can take today that will get you 30%, 40% there, the 2080 rule, right? So Customer Gpt and Mcp server, we talked about it a little bit. This is just an example of something that I’ve done with one of my customers.

01:52:41.610 –> 01:52:47.150
Gil Allouche: we as a software, we have an AI 1st software that takes over all of the advertising

01:52:47.260 –> 01:53:03.000
Gil Allouche: operations right? It creates audiences, it sets up campaigns, it sets up 1,000 experiments. It has a full, full feedback loop, and it can optimize the campaigns to generate pipeline. That’s the software we created. There are still manual tasks in there, for example, generating creative.

01:53:03.180 –> 01:53:05.400
Gil Allouche: We run for just one of our customers

01:53:05.730 –> 01:53:10.780
Gil Allouche: thousands of experiments at any point of time in order to feed

01:53:10.970 –> 01:53:34.360
Gil Allouche: a beast that will run for you thousands of experiments. You have to have hundreds of creatives right? And if you’re going to start handing those tasks to lots of creative professionals, it’s going to create a bottleneck. You’re going to end up with a system that is super fast. But in order for the system to work, you have to wait a week and feed it. Enough creatives. And so what we’ve done. Here is something super simple.

01:53:34.410 –> 01:53:42.870
Gil Allouche: We created a custom Gpt for one of our customers, which is which is included in one of our programs, and we ingested into that custom. Gpt.

01:53:43.210 –> 01:53:47.599
Gil Allouche: The website metadata of that customer for brand guidelines

01:53:48.020 –> 01:54:00.490
Gil Allouche: and all of the creatives we ever ran in this example. It was, I think, 1,500 creatives. You can see it in the next slide, and finally we ingested all the metadata.

01:54:00.740 –> 01:54:06.479
Gil Allouche: all the statistics, all the kpis for these ad. So this ad generated, this many

01:54:07.460 –> 01:54:33.419
Gil Allouche: impressions, these many clicks, these many demo requests, these many mqls. Sqls. This is the pipeline he generated, and all of that goodness was ingested into this custom. Gpt. And after I did it I waited about 5 min, and I asked it to generate 30 new creatives for me, that’s all I did, I told him. Hey, please generate. I think it was 10 more creative or 15 more creatives at the 1st prompt, and he just gave me right off the bat

01:54:33.420 –> 01:54:58.069
Gil Allouche: 15 new creatives that were on the brand guidelines that were different than the historic ones, and those were created, based on the performance of the best historical creatives. So right there on the spot, it took 15 seconds, 20 seconds for Gpt-four point 0 to generate one creative. It took it a few minutes to generate all of the 15, and I had it right in front of me, and I could take those download them.

01:54:58.770 –> 01:55:09.939
Gil Allouche: Now upload them to the co-pilot of Metadata or the Mcp, and ask you to generate 50 more best performing ads. It is as simple as that you can create this custom. Gpt today if you wanted to

01:55:14.610 –> 01:55:24.319
Gil Allouche: copilots. So co-pilots, you can hear this term everywhere. You know, most saas companies. Most, I would say, software companies, you have to log in.

01:55:24.470 –> 01:55:30.920
Gil Allouche: click a bunch of buttons. And then an action is being taken for you. You have to adopt to the software

01:55:31.010 –> 01:55:50.520
Gil Allouche: many times. That’s why you have to hire professional services agencies, consultants to run the software. If you remember the old sap days, they generated a lot of revenue because their software is so complicated. You know, salesforce.com, you know, salesforce.com can be pretty complicated. How much money did you spend

01:55:50.520 –> 01:56:02.660
Gil Allouche: sorting out your salesforce.com the structure, the workflow, the process integrating all of the different systems, setting up Cpq, it’s a nightmare, and so imagine you had a co-pilot

01:56:02.660 –> 01:56:18.039
Gil Allouche: that can go and do things for you. You use simple English or Spanish, or whatever language you use, and just feed it. The context that you want, and it will go and click the buttons for you. And so I’m giving a few examples here. The most exciting example, I think.

01:56:18.040 –> 01:56:42.110
Gil Allouche: is lovable, lovable, is a piece of software where you can type in whatever language I have. My 8 years old build apps with it these days, and you can just feed it or talk to it, and it will create an app for you. It’ll create a game for you. It’ll start building stuff for you. It may not be perfect the 1st round, but the 5, th 6th rounds are going to be pretty amazing. You wouldn’t believe what you can create with it in a matter of hours.

01:56:43.020 –> 01:56:47.200
Gil Allouche: And then the same thing goes for social media posts. You know, you can today

01:56:47.420 –> 01:57:08.910
Gil Allouche: essentially give it some context, give it historical posts tell it to be inspired by other social media influences that you’re excited about and like their style, you can give it, of course, to a researcher for your historical to make sure everything is done with your name, with your style, with your formatting, and it will go and generate.

01:57:09.050 –> 01:57:14.399
Gil Allouche: This is postal, by the way, and I’m right, and I’m reading on the right hand. It will generate 30 posts for you.

01:57:14.840 –> 01:57:22.089
Gil Allouche: and if you build a workflow around it, it’ll set it up and post it automatically for you, too. You don’t have to click even the post button.

01:57:22.380 –> 01:57:49.130
Gil Allouche: And then on the left side, self-promotional. This is an example of our own software, where, instead of setting up 500 campaigns, or even, let’s go in smaller number, even 50 campaigns. If you have to set up 50 campaigns today on Meta or on Linkedin or on X or on Reddit. Today, you can upload this spreadsheet of your own structure to our software, and it will go scan the spreadsheet, generate the creative, fill in the blanks, and then show you a draft campaign with 50 experiments with 50 campaigns

01:57:49.130 –> 01:58:00.269
Gil Allouche: where you can approve it and set it up to go live. This is something that you should demand. And when you look for a software today, if you’re a new company, or if you’re making a change in your existing company.

01:58:00.270 –> 01:58:27.230
Gil Allouche: you should really look for software that already has a co-pilot in there. If not, then you’re going to end up waiting for them to build it for you, because this is inevitable. Ideally, you have a software that doesn’t only have a co-pilot that you still have to log into the software and put a chat. But ideally, you get the software that has an Mcp where you can bring the software into your existing AI stack. So if you’re accustomed to doing things in Chatgpt, imagine you can run everything

01:58:27.230 –> 01:58:35.600
Gil Allouche: from Chatgpt. And that’s actually the demo I want to show you today, Lisa, did you want to go with your bid agent first, st or should I go with

01:58:35.770 –> 01:58:37.540
Gil Allouche: with make and and metadata.

01:58:39.797 –> 01:58:45.040
Lisa Sharapata: You know what? I’ve got my screen up and we’re I know we’re kind of short on time, so why don’t I pull?

01:58:45.330 –> 01:58:51.859
Lisa Sharapata: I’ll just pull this in here now. So what Gil was just kind of leading off with is.

01:58:52.080 –> 01:59:06.160
Lisa Sharapata: he was saying, okay, so we had the spreadsheet we put together all these different experiments, and within that. We then put this into metadata, ingested it into our campaigns. And we’re saying, we’re running these experiments. And then

01:59:06.520 –> 01:59:33.770
Lisa Sharapata: on top of that, what we’re also doing is we’ve activated bid agent. So what is really sweet about this, and I’ll show how it’s working. But bid agent is coming in and optimizing our bids in real time at a pace that no human could so kind of. Think about it. Like you, you know. Typically the standard problem with Linkedin right now is you’re burning cash, or you’re using their manual

01:59:33.830 –> 01:59:45.390
Lisa Sharapata: bidding. And you’re having a human. Go in and try to keep up with the pace of their algorithm to get your best bids. And there’s just no way that you can do that

01:59:45.470 –> 02:00:10.819
Lisa Sharapata: and and be optimized. So, you know, here we’re seeing how you know how it’s performing. But then, just for the sake of time, I’ve got this T up to show of when we look at our logs and how much optimization is happening. I mean, we’ve have like, last week we had a webinar with Zoom. They were doing about 128

02:00:10.860 –> 02:00:17.340
Lisa Sharapata: optimizations a day with Bid agent. And I mean, here’s an example from our instance where you’re just seeing.

02:00:17.740 –> 02:00:24.659
Lisa Sharapata: you know, it’s automatically coming in here and looking at all of the different campaigns which ads are performing best

02:00:24.730 –> 02:00:34.119
Lisa Sharapata: what the cost per click should be, and making adjustments to it in real time, and you know little tweaks like dropping things down.

02:00:34.180 –> 02:00:54.509
Lisa Sharapata: you know. 84 cents or something is something a human wouldn’t even probably think to do. But when it all adds up and you start seeing, you know some of these savings, even if it’s pennies on the dollar hundreds and thousands of times over and over again in the course of a month like you. The savings are real, but then we’re also seeing

02:00:54.560 –> 02:01:17.930
Lisa Sharapata: really great results from this. From the standpoint of better click through rates higher pipeline and revenue being generated from the campaign. So I’ll just say if you wanna hit me up at the end. DM, me, or whatever I can get you into an early access free trial for it. But it’s really cool. So with that I’ll pass it over to you, Gil.

02:01:21.180 –> 02:01:22.239
Julia Nimchinski: Gil, you’re on mute.

02:01:23.450 –> 02:01:24.919
Gil Allouche: Thank you, Julia.

02:01:25.360 –> 02:01:29.439
Gil Allouche: Okay, let me see if I can quickly show you something.

02:01:29.770 –> 02:01:33.309
Gil Allouche: Okay. 2 things I want to show you. One is our make account.

02:01:34.000 –> 02:01:39.720
Gil Allouche: This is the work of a couple months. So we have something like 17 or 18

02:01:39.840 –> 02:01:55.480
Gil Allouche: agents that we built already from this workflow that I showed you, and some of these agents are multi-agent, so within them there are multiple workflows. For example, we have an agent that every morning goes through all of your competitors, sees all the sem ads they have.

02:01:55.480 –> 02:02:15.029
Gil Allouche: then puts together a spreadsheet, then goes into that spreadsheet, generates copy for all of them, and then goes again into the system into metadata system and sets up a competitive campaign for you. We have another system that ingests all of the competitor ads on all social, and generates a new set of creatives inspired by your competitor ads, especially the ones that perform well. We have so many of those right.

02:02:15.350 –> 02:02:31.509
Gil Allouche: These are, they saved us, I think, something like 30, maybe 35 h. So far, I’m on a mission to get to about 80 90% of the work of the manual work out of the door, so that our team members have more time to breathe, to sleep and to work on strategic

02:02:31.590 –> 02:02:59.819
Gil Allouche: strategic work creative work and then serve more customers have more time to work on things that they do best for our customers. So ability to generate more revenue with existing resources is absolutely possible today. This is one of the ways we’re doing it. If you’re a business owner and need some help doing that, I’m more than happy to show you the blueprint that we’ve gone through and how we’ve done this, using the tools that are available to everyone.

02:03:00.660 –> 02:03:10.320
Gil Allouche: Another thing I wanted to show you is my little Claude here. So I ran this just before our webinar, and I asked it. Something simple. I asked it, hey? And I have.

02:03:10.570 –> 02:03:28.630
Gil Allouche: you know, little cheat sheet here. I have metadata installed here as an Mcp. Right? So, metadata built in Mcp. I connected to this Mcp. You can see the kind of actions I opened here for Claude. I also turned off those 2 because they take forever. They’re pretty big. I don’t want it to think for more than 30 seconds.

02:03:28.630 –> 02:03:40.900
Gil Allouche: But look, this is the question I asked it. I asked to give me the best performing actor in terms of pipeline created and included creative. And it did some work. You can see the kind of work it went through.

02:03:41.060 –> 02:03:47.690
Gil Allouche: It hides this work for you because you really don’t care, and the end of the day it gave me the 2 results. One of them is the ad that generated 2 and a half

02:03:47.900 –> 02:04:11.989
Gil Allouche: source pipeline, and then some of the influence pipeline. It told me the spend. It’s pretty nice, right? It’s 100 x return on the spend, and then the second one is the lower Cpl. I want to show me the ads that generated the most amount of leads for me. So I click here it opens the link. And right here, hopefully, you can see my screen. It shows you the kind of ad it ran for me.

02:04:12.500 –> 02:04:21.779
Gil Allouche: So those are the really the 2 things I want to show. I can go and ask it. If I wanted to create 3 new ads inspired by those

02:04:21.900 –> 02:04:24.620
Gil Allouche: best performing ads.

02:04:25.920 –> 02:04:30.990
Gil Allouche: I should always expect that everything’s going to crash and burn when you do things in life. But I’m going to try it, anyway.

02:04:32.020 –> 02:04:42.189
Gil Allouche: And now it’s going and calling Metadata, Mcp, and it’s essentially an agent that is going and setting up things for you. It creates the ad.

02:04:42.470 –> 02:04:44.690
Gil Allouche: It sets them up on the Channel.

02:04:44.890 –> 02:04:47.260
Gil Allouche: It it will do the whole thing for you.

02:04:47.540 –> 02:04:49.300
Gil Allouche: You don’t have to lift a finger.

02:04:49.560 –> 02:04:56.390
Gil Allouche: It comes up with a copy. It comes up with a call to action the whole thing for you. If you want it, it will create the content for you.

02:04:56.390 –> 02:05:19.910
Gil Allouche: It will generate all of the components, all of the artifacts, to make sure it happens. So this is just an example. I’m going to stop it here it is working, which makes me very happy. If you want to hear more about it. Of course. Just shoot me an email. I’m also happy to help you on your own specific example doesn’t have to be advertising. This is now available. This technology is now available on all realms, and so I strongly encourage you to adopt it.

02:05:20.710 –> 02:05:34.510
Julia Nimchinski: Gillissa such a fascinating tech and session. And you are definitely way ahead of the market question that will keep coming up in the communities. How do we transition to this reality? What’s the 1st step, Gil?

02:05:34.760 –> 02:05:35.770
Julia Nimchinski: Rationally.

02:05:36.530 –> 02:05:55.990
Gil Allouche: To me. It’s always about Mini wins. So it’s like when you finish your retreat. And you have all these realizations about life, and you have a hundred things to do, you start with the smallest win something that makes your dopamine circulation go. And so I would say, if you open up Chatgpt today.

02:05:56.100 –> 02:05:58.990
Gil Allouche: And you create a custom. Gpt.

02:05:59.860 –> 02:06:09.690
Gil Allouche: you put in whatever data your Crm export, your marketing automation export something that is relevant to a task you want to do.

02:06:10.050 –> 02:06:28.969
Gil Allouche: you should feel great about it, and tomorrow and make a reminder. Put 5 min on your calendar to do 5 min more open cheat without any context, without any agenda, and based on the biggest task that you have, the biggest headache, the biggest stress that you need to solve at that moment. Every day, 5 min, you’ll see great compounding interest from it.

02:06:30.760 –> 02:06:49.339
Lisa Sharapata: Yeah, I’ll actually add to what Bill just said. What I think is really important is, what is it you’re trying to solve for? And then that always helps me to make the time for it, and to make it a priority to say, All right, you know, here is this gap. Here’s this problem I have in my day to day that I need to get.

02:06:49.550 –> 02:06:53.920
Lisa Sharapata: I need improved. I can’t keep spending this much time on. I’m burning out right. And so

02:06:54.230 –> 02:07:05.310
Lisa Sharapata: then you do have to force yourself to sit down and spend a couple hours or a couple of days to figure out how to build out that workflow, but once you do you’ll want to keep going.

02:07:06.240 –> 02:07:12.930
Julia Nimchinski: Thank you so much again. And what’s the best next step? And we unfortunately have to transition to the next session.

02:07:14.680 –> 02:07:15.250
Gil Allouche: Yep.

02:07:15.250 –> 02:07:27.630
Lisa Sharapata: I’ll say, feel free to reach out to either Gil and I directly on Linkedin or also you can go to the Metadata website and and request a demo. If you want to see something more specifically.

02:07:27.630 –> 02:07:37.160
Gil Allouche: Yeah and sign up for make.com sign up for N. 8 N. Sign up for Zapier one of those and start playing with them just out of curiosity and fun, and you’ll get into productivity modes very quickly.