You Won't Believe What AI Is Doing to Your Mind

This content was created by Eric Siu. This content was uploaded by Eric Siu. This is a YouTube discussion episode titled "You Won't Believe What AI Is Doing to Your Mind" (published 2026-03-17). It is a conversational interview-style talk between Eric Siu and Neil Patel covering AI’s effects on work habits (making people either learning machines or lazy), deploying AI agents in Slack for sales/recruiting/SEO and coaching/performance management, practical marketing/SEO implications (including ClickFlow promotion and ChatGPT ads performance vs. Meta/Google), the importance of APIs and documentation for agent integration, and agency/relationship dynamics — with some operational tangents (eye strain, devices) and a brief product promo/close.
Chapters
I realized recently that AI either makes you more of a learning machine or more of a lazy machine. So the people that use AI, for example, I'm going to guess for Neil. Neil probably asks a lot of tax questions to his AI, okay? So more or less as a tax strategist, right? You're learning more. I'm sure there's a lot that you've learned about things that you just didn't know before, right? But my point is you're asking a lot more questions that you usually would've asked before. You're not just kind of, you know, like taking it and say, okay, AI, do the work for me. So When I say AI makes you a learning machine, you ask more questions and you actually expand your mind more. That's for smart people, okay? Now, unfortunately, many people, many people, like, you want to be lazy and smart. You don't want to just be lazy, okay? So, because smart and lazy people try to figure out ways around things, right?
They figure out creative ways.
But if you're just someone that just throws it over the fence and you're just like, ChatGPT is going to solve the problems, ChatGPT will write my prompts for me, ChatGPT is going to write all the content for me. Then you will get slop, right? So that's the first point I wanted to make. But the second point to what you're saying is, I've had my AI agents, they've started to invade my Slack channels, right?
Mm-hmm.
And so they're working in my sales room, they're working in my recruiting room, they're working in the SEO room too, and it's working with the team. Now here's the thing, I see it, their team's asking for data pulls, and it's giving the team strategies, outside-the-box strategies, and it's rating our strategies, and it's also offering to do the work too. Now here's the other thing that you just mentioned that gave me an idea. So Neil and I just made an idea baby. Let us let everyone know that we just made a baby. So the idea baby that we made here, Neil, is if you have it coaching your team, okay, like imagine your kids are going to be working with personalized AI tutors, right? Everyone on your team now has a personalized coach, okay, that has all the data that understands what that team's supposed to hit, and it's going to help.
Okay.
coach the team, but it's also good at what? It's also good at performance manage the team too, okay?
Yes.
Whether you like it or not, that's what's going to happen, and it's going to show like a leaderboard that way, and I think that's a good and a bad thing, right? I think we're always just trying to do better for our companies, and I think this is great. Like, I didn't have that idea for the coaching piece, but I think I'm going to inject that a little bit into my little bots that have invaded Slack, which it's been tremendous. My team's like, oh my God. I'm like, you say you like it, how useful has it been? Oh my God, 10 out of 10, 10 out of 10. I'm like, okay, all right, we'll do more.
Yeah, dude, I was telling Ethan about this concept because he has his AI lab where he helps people set up—
Is it really a lab?
Uh-huh.
Okay.
And he helps people set up, you know, all this stuff internally, whether it's OpenClaw or whatever you want to call it.
Does he charge $6,000 for OpenClaw setups?
Uh, I think it's, I think it was either $5,000 or $10,000.
I think that's fair.
Right. So, and then including the PC or something like that.
That's a really good deal.
That's why I think it's closer to 10 grand.
If you include the Mac Mini, he's probably make— that's like the Mac Mini will cost like $1,000 or so. And then the rest, you know, he figures out.
Yeah.
Yeah.
I think it's 10 grand or somewhere around there.
I think that's a good deal.
So, you know, I was telling him, I was just like, dude, I'm like the opportunity. And I don't know if he's, if he's doing this or not, but I think the opportunity in AI right now If you're trying to set up a big business outside of creating an LLM, which is very difficult and expensive, and creating those lovelables is very hard to do, you know, for every lovable that we see at $400 million in ARR, there's, I don't even know how many that are struggling, right?
Yeah.
There's too many to count.
Yeah.
Is do consulting where you help train people's AI to produce better outputs when their employees use them. So instead of training the employees, train the AI to adapt to the humans and do this at mass scale for everything from marketing to engineering, et cetera. It's pretty much another version of Bain or McKinsey or Accenture, but specifically focused on just training AI to deal with people wherever they are versus trying to train the people. Because I think it's just going to be too hard to train the people.
Exactly. So that's why I think the sooner you get it working for yourself as in the organization, like your leaders, whoever's running AI, Then you get these bots to invade your Slack channel or your team's channel. Everyone's going to learn because your team's going to feed it context. It's also going to feed— like, it's just going to be like this self-reinforcing loop. And that's what we've been doing. And we've getting— I will say we're getting more and more people reaching out to us saying, hey, we're hearing what you're talking about. Like, we want these enterprise agents set up. Like, we want these revenue agents, we want these recruiting agents, we want all these things. So the market's kind of pulling us in that direction. So we'll see where it goes. But I do think it's like what Hiten's doing for the setup is huge. I think You and I, like what we're talking about right now, even what's happening on Twitter is probably like the top 1% of the 1%. The vast majority of the world doesn't know about this stuff. And that's where I think the opportunity is. Like if you're listening to this and you're like 18 years old right now, I would just do this.
Yeah.
Yeah. Quick break. One of our ClickFlow users just told me that they're saving 90+ hours a month on content alone. ClickFlow is our AI SEO product that actually does the work, not just here's what you should do. It does it. It creates content in your voice. It helps you with other areas such as finding the right internal links. and creating FAQs. We spent years making it, perfecting it with our clients so it doesn't just sound like AI slop. You can get a 14-day free trial at clickflow.com. If you're serious about AI SEO, you should check it out. Back to the show.
On a side note, speaking of AI, did I ever share with you the results that we ended up finding? Because, because of our size and scale, and we're not the biggest agency, but we have enough data on certain industries where we know if you're recommended number 1 by ChatGPT versus number 2 versus number 3.
You should do that and then talk about the ads.
No, that's already out there. We put that out there.
Oh, you did? I, I, I, you know, you didn't say it here. No.
Yeah, yeah. I don't, I don't know if I said it here. Uh, I'm trying to end up finding the chart, but it was interesting because I was just like, huh, everyone knows that being position 1 on Google, it's worth so much more than being position 2, but AI is way more of a winner-take-all than traditional search. So if, if you count the most revenue you can make is position number 1, let's call that 100%. If you're position number 2, you do roughly 32% of position number 1.
I was just going to say that.
And guess what position number 3 is compared to position number 1?
5%?
Close, 8%.
Yeah.
And position number 4?
1.5%?
Very close, 1%.
Yeah.
So pretty much if they're giving, if ChatGPT or Gemini or Perplexity is giving you a list, it's number 1 takes almost all the revenue, then number 2, and then it just pretty much doesn't matter.
This is the old way of SEO. Back in the day when there weren't any ads or anything, like number 1 got the lion's share and then Like they started pushing it down more and more as the click-through rate started going down more and more. So we'll see what happens, but that's good to know.
Dude, I don't have a— I don't have it posted. I'll end up showing you on this. Check this out, how ChatGPT ads, because we were talking about this over text.
So what Neil has pulled up, so it says how ChatGPT ads perform against Meta and Google Ads. So it shows ChatGPT versus Meta, Plus 256%. Okay.
This is for lead quality, I believe, right? Yeah, I can't see, but let me see. Let's see. Yeah, lead quality.
Okay, so 256% up on ChatGPT versus Meta, but ChatGPT versus Google, 50% down.
Yeah, it's down. Yeah, exactly. 49%. Close enough.
What's the bottom one?
The bottom one is cost per acquisition. So CPA.
Yeah. So CPA, ChatGPT versus Meta, 46% cheaper, and then ChatGPT versus Google, 38% cheaper. So. It's clearly a winner versus Meta. And then, you know, you're still going to get, well, even Google if you want. Yeah, because that's CPA basis.
Yeah, because as a marketer, we only care about CPA, right? Lead quality from Google is much higher than ChatGPT, but you also have to look at cost. Google Ads are some of the most expensive because lead quality doesn't assume cost. You can end up spending 3, 4 times for a click because Someone may end up typing in like, I'm looking for a global digital marketing agency, or I'm looking for the best ABM account, ABM software solution that works really well with LinkedIn. And then Carrot could end up showing up, right? You can end up bidding on that, but it's going to be a very expensive cost per click. Lead quality, I totally agree. Google is extremely high on lead quality. But the ads are just so much cheaper right now on ChatGPT. It's the Wild Wild West.
This just goes to show you again, every time with marketing, whenever a new ad channel comes out, you want to hit it really hard before the costs start to go up.
Yes, and a lot of people are struggling to get into ChatGPT ads, but they no longer need to struggle because ChatGPT just announced a partnership with Criteo.
Yep.
So now people should be able to get access to ChatGPT ads free.
Good for Criteo's business.
Yes. I was shocked that they partnered with Criteo.
Nobody talks about Criteo anymore.
Yeah, because they're known for like more of like a spammy-ish affiliate-ish type of—
Must be a relationship somebody had at OpenAI with them. Yeah.
I would assume that Trade Desk would've got it.
Who knows? Yeah. A lot of this stuff is relationship-based. Yeah, it totally is. So, okay. But let me ask you this kind of separately, because we're talking about just marketing in general right now. Have you added, do you have APIs for both Ubersuggest and AnswerThePublic?
I don't know.
So I think APIs as marketing is going to become like very big because you have the agents crawling the web, right? So I think your technical documentation is going to be important and then your APIs, because the way like I'm starting to work more now is within the LLMs, I'm just asking it to do stuff. So yesterday I was messing around with one of our agents and we use this thing to send cold emails called Instantly. I'm like, I don't want to make 5— I don't— first, I don't want to write 5 different campaigns. Okay, that's number one. And then second, I don't want to load all the copy into these individual campaigns. And I'm just like, hey, can you like create 5 campaigns for me and create like an expert panel of like legendary copywriters and email copywriters and then recursively score them until they're like, you know, 90+ out of 100. And it does 5 versions for me. I'm like, these look pretty good. And I'm like, well, I don't want to load them. I'm like, anything you ask me to do, I just don't want to do it, right? So I'm like, can you just like do it for me? And like, it loads everything. I check it. I'm like, oh my God, it loads up, right? So I'm like, with the software you have, with the software I have, we need to have APIs for it. So then the agents can discover it first because I don't think other people are doing it that quickly right now. And then you have to have good documentation and Mintlify has really good SEO, like not SEO, but software documentation. What is it called again? Mintlify.
Mid? M-I-D?
M-I-N-T-ly-fi.
Oh, mint. Like mint, like ice cream.
Or yeah, that's true.
Toothpaste.
So Mintlify, they've been growing faster and all they focus on was documentation. And I think they're going to grow even faster now because all these agents, that's all they read is documentation. They don't care about like, you know, show me like a, you know, bikini sports or something like that, you know?
Yeah. And before you got into that topic, you mentioned relationships and how people close business through relationships. Do you want to take a guess on what portion, or actually I'll ask you first, what portion of your agency revenue comes from relationships?
Well, what do you define as a relationship? Like they know you from your content? Like what?
You know the person.
Okay.
So not that they know of your brand or they know of the Single Grain brand. They may also know that, but you also know that individual or employee on your organization knows the other person who's interested in working with you.
I say 20%.
Oh, wow. Yeah, that's a good chunk.
Yeah, but we're also smaller, so.
Correct. On our SMB division, I don't know what the numbers are, but maybe somewhere similar, I'm guessing. I don't, I honestly don't know. I do know the majority does not come from relationships in our SMB division that I'm 100% on. Yeah. You'd be surprised over the last 3 months of our sales. I don't have margins numbers yet, but—
Like 2%?
Relationships?
Yeah.
It's almost all relationships. Oh, okay.
So that's you like, but you know these people is what you're saying. Are you reaching out to them?
Are you specifically taking, are you talking about me and my relationships or relationships within my organization?
Okay. No, I was talking about you specifically.
I don't know my percentage. I know the percentage of All of us combined.
Yeah, all of us, like that makes sense. That makes complete sense.
Yeah, so when I look at like my managing director of Australia and his relationships, or Javi who leads LATAM for us, or Mike G, our CEO, majority, when I say majority, I'm talking about 90+% is starting to become relationships. Now some may be classified as like, oh, you got an RFP from this company. Oh, we got an RFP from that company because I have a relationship with someone who works on there, or my team has a relationship with someone who works there. And we may end up winning that RFP. And if we did, you could be like, oh, you got it because of RFP. Well, if we didn't have the relationship, we probably would've never been included in the RFP.
You know, I think so relationships aren't going away anytime soon, but when you look at a lot of maybe these agencies or service companies that have become really bloated over time, they still have the relationships. So like the way is, can you acquire the relationships and AI-ify what you can? And it's not like it's a new thesis or anything, but it's like, but it's clear to me that a lot of these are moving like dinosaurs, but the relationships are still incredibly valuable. But once they start to get hit more and more, they can't take it anymore. They're going to have to tap out and sell.
I think what people forget about relationships when you're doing deals, relationships create trust. And what we found is when people, when you do good work and people move from one company to another and they know you, they, yes, there's change when they're going to a new organization, but they also like a feeling of comfort and they know that if they hit you up, And they work with you again, they can rely on you. They already know what to expect and you should be performing. And I think that trust part is something that people overlook a lot in marketing in which they're just like, yeah, we want to put the message out there. What can I end up doing to maximize my conversion rate? But they forget to really emphasize on building trust. And Google's talked about this for ages with the double E-A-T, right?
Mm-hmm.
The last one being trust in E. If you end up building that trust, I just think it works so well for you in the long run.
That's why throwing nice events are good for building relationships at scale sometimes.
Yeah, and let's go back to dinners. Do you still do your dinners here?
Every 2 months or so, 3 months. I was doing it every month for a little bit, but I'm so heads down with this. You know, my eyes hurt like hell every night when I go to bed because I'm on this until I burn my eyes out.
Have you tried drops or?
I do drops, but you have the LASIK issue I have, right? Where like your lipids don't, the oils don't come back as quickly, so your eyes dry out quickly. So I do use drops, but then because I'm just, every time my eyes come back, I just get on it again, I burn it out. But my point is, yes, I've been doing the dinners, but less. And then the thing is, all those dinners have been AI dinners, and then all those founders are also doing the same thing. They're burning their eyes out every day.
Yeah, I have the same issue. I didn't know you had laser eye surgery.
Yeah.
But yeah, I did have the same issue. Yeah, my eyes are dry. I don't really feel it like you feel it. I'm also not on a screen as much. I'm more on a phone call. So I'm not looking at my screen.
You're talking.
Yeah, I'm talking. But every time I go to optometrist, they always tell me like, oh, you need to do this. Try putting this like Vaseline type thing in your eye. I'm like, why? My eyes are dry.
It didn't help. I did it for 6 months. It didn't help. So I went to a doctor right next to the doctor I have in Beverly Hills. But so The big thing I found, this is a little tangent, but like if you look at your phone too much and it's a dark room, that it like, it screws it up way faster. So don't do that. Yeah. Like I workshopped those with ChatGPT. I was like, why is that? It's like explaining all the science. I'm like, okay, makes sense.
So. Dude, the time I look at my phone more than anything when it's dark is on an airplane.
Yeah. Which is really bad.
Yeah. To watch like movies and then I fall asleep.
What do you watch on your phone? Your phone on your—
Phone. Dude, I copied your phone. You know how I used to have the MacBook, not MacBook, iPhone Air? I ended up buying your exact same phone. Oh, nice. Because when you're on airplanes—
When you watch movies on this, on the airplane, on the iPad. I got rid of the iPad. Why?
It was just harder to deal with work.
It's so good for movies. Yeah.
My problem with the iPad, I've been using, you kind of nailed it earlier. You mentioned I'd probably use AI a lot for tax-related questions. I use AI a lot for structuring-related questions and spreadsheets. So still finance. It's not always tax, but even structuring indirectly, it is related to taxes, finances. Is this related to taxes one way or another? I've always found that, and because the AI is just so much easier for me to crank out reports and slice and dice the data, it's just harder to get the data in a format I want on an iPad. And it's just easier for me to do crap with a computer.
So you're using Claude Code within the Claude app, I'm assuming?
Yes.
Right. Okay. That's how you're using it.
Okay.
Yeah. There's so many things we can talk about there, but we also export a lot of data to Excel. Yeah.
And then when you export to Excel, 'Cause I'm really good at using Excel. It's just easier for me to use Excel on my laptop than the iPad, 'cause the iPad versions of a lot of these softwares, yes, it's pretty much the same, but it's not identically the same. And it's just for some reason I just do better using it on my laptop.
Imagine if you have a— if you had a secure version of a Claw. So like NVIDIA is coming out with Nemo Claw, which is like their version of Open Claw, but it's secure, right? So imagine they come out with that and it's sitting on your phone, it's ingesting all of the text that you're getting on the updates and all that. And because you You work with it as your copilot and the charts you make, it understands your behavior over time and it builds on that and it just starts to make the charts for you over time and it makes better charts for you too.
We've tried using AI for the charts so much. We try every single month because if we did that—
It's okay, just keep trying.
Yeah, we would save on the design budget. You want to know the biggest problem when we actually put the inputs into AI for the charts? This is going to sound like the stupidest problem. I wouldn't say it's not 100% of the time, but like And it's not even 50% of the time, maybe like, uh, 15 to 20% of the time it messes up the design.
Yeah.
So it doesn't keep the exact style guide no matter what you tell it. And then I would say 25 to 30% of the time it messes up the data.
Yeah, that makes sense.
So then, and then to fix it, because we look at the cost, right, to fix it and find the errors, because a lot of the people, the designers don't find the errors, that's more costly to us, the reputational hit, than just having it done right. So—
Oh, you can't show that to clients, obviously.
No, no, we can't. But what we find is people will say they check it and they don't check it. So then when we do it manually, there's less errors and the cost isn't that high. Like my design cost for the charts is call it around $3,000 a month. At $3,000 a month, yeah, I don't care. And I don't want the reputational hit of just AI messing up, but one day it will get it right 100% of the time. Will it be 6 months or a year? The technology is there, they just got to fine-tune this stuff.
Anyway, that's it for today, guys. Uh, please don't forget to rate, review, and subscribe, and we'll see you later.










