Back to the boards: Jason Akatiff: AI, Media Buying & Lead Gen

Transcript

Click any time to play from there. Speaker labels come from an automatic transcript and may contain errors.

1. Revenue is not the whole business

2. 0:00share
   John Casto All right, guys, got a special guest for Q&A today. We'll backstory on who Mr. Akertif is, what he's been doing the last 20 plus years in the affiliate legion paper call on my marketing advertising world, what he's doing today with AI and a cool software. He, I don't even want to say five code at this point. He just built and what he predicts.
3. 0:25share
   Jason Akatiff going to happen in the next few years based on everything he's doing the last few months jason take it away man yeah um you know just really briefly for those that you you that don't know been in the online marketing space since started in 2003 as an affiliate I was a full-time affiliate doing Black Hat SEO and media buying from'03 to'08, at which point I purchased Ads4Do, which was a tiny network doing like $20,000 a year from someone, and that turned into A4D. We got heavy into Nutra.
4. 1:03share
   Jason Akatiff settled with the federal trade commission in 2012 for being involved in free trial nutra and then moved into auto insurance we're doing somewhere between 10 and 12 000 leads a day in auto insurance then moved into the vsl space for financial newsletters early early vsl days not many people were doing it back then we were one of the largest partners for uh golden hippo on all their stuff in the early days and then uh ultimately kind of the split pass i uh a4d uh which is now a4d nib kind of put somebody in charge of that business and they went into mass tort so about seven years or so in mass tort now do everything in to end from calls to leads and
5. 1:51share
   Jason Akatiff then pushing into third-party intake centers that we manage uh and then uh you know sell signed retainers off to the attorneys and then i went down the path of building an e-commerce brand we had a brand called boundary um it was uh it was a light bulb brand uh we were selling light bulbs we got that up to about 40 million dollars a year in revenue and then uh there was a bunch of issues with shipping containers going to $2,500 to $27,000. Chip shortage increased our cost of goods by 30% and Facebook hit all at the same time with iOS 14.5 our CPAs went up 30% and every sale we were making we were losing 50% on.
6. 2:40share
   Jason Akatiff I really had to ask myself do i want to be in the inventory business because for those of you that are in e-commerce you know in order to scale you have to have product and so at that point in time i you know you know every time you scale you have to order more product and dump more money back into that business so
7. 3:04share
   Jason Akatiff you know we i just said hey i don't ever want to be in an inventory business again uh the only way i would do it is like a single consumable on subscription i think those businesses are super interesting your rice coffees everyday dose etc um but this business was a hard goods business with basically a one-time purchase so you had to expand your catalog in order to sell them more stuff which ultimately meant more inventory um so that's been that and then you know where i think things have gotten really interesting um in november of last year well actually let me go back a little bit uh maybe a year ago uh one of the people on my team had told me

8. Build from a problem you know

9. 3:49share
   Jason Akatiff about cursor um if anybody knows about cursor was like one of the early they took a visual studio or vs code and then they put a layer on top and they started building agentic coding models inside of it they were early to this game And so, you know, I started playing with that maybe a year, year and a half ago, but it was just so much, you could get good, you could get good products or good, good stuff out of it, but it just took so much work. And I was just like, I don't have the energy to do this. Um, so I, I built some quiz based funnels and things like that for that.
10. 4:26share
    Jason Akatiff And then I just kind of put it down for six months, nine months, whatever. whatever it was. And then Gemini came out right around Thanksgiving of last year. I should say anti-gravity, which is the Google IDE, which is another spinoff of VS Code or Visual Code. And they also came out with their Gemini 3 model. And that's what got me hooked. I picked that up and I was just like... started to say, Hey, can you build me a landing page? Can you build me? I needed to revamp one of my websites for one of our brands. Could you build me a website?
11. 5:03share
    Jason Akatiff And you know I was getting really really good output on the first pass and I was like wow this is workable I wonder I wonder what's possible here that led me from doing a website to then I built my own back-end CMS for that website that was kind of my first step one step two Then it was like, hey, could I build a landing page system? So that was step three. We still run on that landing system today that, you know, basically you build a templating system inside of it. That links up to lead router. etc. So that all was built together.
12. 5:37share
    Jason Akatiff I'll tell you more about that later on, because I think we're ultimately headed for a world where landing page builders are not relevant anymore. But then around February, I really, you know, It was like, oh, you know, I hate Leedspedia. I was on Leedspedia for 13 years. And I was like, I hate Leedspedia. In our business, we do everything based on the backend, right? So if you think about lead routing and call routing, it's like comes in. So you got raw, maybe in calls, you've got connected.
13. 6:11share
    Jason Akatiff and then you've got duration right like those are your basic events that happen and then in leads you have you know just comes in and is accepted uh and then you've got is it sold you know and so you've got like kind of a parallel path for both of these and then but on like the uh on the business for in tort which john's very familiar with uh the everything is back end right so everything is cpa and none of these platforms really support cpa so that's kind of what led me to start to build you know stuff to solve that problem uh was you know there's all disjointed stuff and so now i've got probably you know i moved from anti-gravity
14. 6:58share
    Jason Akatiff i think that's a great place to start anybody you want to get into this stuff i would highly recommend you suggest anti-gravity because it has a bunch of free models in it it's very easy to use um but i quickly transitioned over to terminals and i use orca now and um and cloud code and now i've got probably 2 000 hours of development in cloud code from everything from landing pages to video creation to software development platforms harness building and all all that kind of stuff and really that's where i've been super passionate about uh i stood up uh builtwithai.
15. 7:36share
    Jason Akatiff com which is our uh you know just kind of an ai focused everything's free uh ties into some telegram groups but you know my passion is around ai and my passion is around what does a future look like when it comes you know AI and what and we can't tell right like I'm not here some true sayer that I know all the answers but I do use it every day and I could see what I'm doing that other people are not and you know maybe I can share that and that'll be helpful what are some predictions you would make in let's say the near next 12 months based on what you're seeing today in comparison to the last 12

16. Give the model a working structure

17. 8:17share
    Jason Akatiff I mean, the biggest part for me is I don't want to manually do anything anymore, right? I want my, what we'll call an agent, and everybody talks about an agent differently. uh but an agent to me is essentially a harness and a model combined together so cloud code is a harness and a harness is skills and configuration and rules and gotchas this is a you basically take this engine if you think about it if you think about an llm like opus or codex 5.
18. 8:56share
    Jason Akatiff 6 that's an engine well the harness is the rest of the car it's all the rest of the car that that engine goes into power and you know the next you know over the next i would say two to three years i believe there will be harnesses for all types of work right um you know everything has started out as harnesses for coding and the reason that is is coding's deterministic uh so you've got probabilistic and you have deterministic You know, these engines, they're probabilistic. They're not deterministic. Like fixed code, do A plus B gives you C. That's very deterministic. Probabilistic says, ask the same question, get the different answers 20 different ways, right?
19. 9:51share
    Jason Akatiff So if you're doing work in data or analytics, probabilistic is not good. so you know how do you build harnesses around that that is give the ai tools they call them tools right how do you give the ai tools what maybe that's scripts maybe that's access to software So that the deterministic part stays deterministic, and the probabilistic part, that is, how would you optimize these campaigns? Probabilistic. What does all that data say? Deterministic. so when you couple this harness together with a model an underlying engine into that car that's how you really get something that can you know manage and whatnot and i think a great example of this is uh like in in lead router like say you're gonna go
20. 10:49share
    Jason Akatiff well there's no framework there's no nothing so it's you know you've got a probabilistic model that is going to come up with all that stuff and hallucinate a lot right and so you're going to you're going to get this mess and but then
21. 11:07share
    Jason Akatiff once you take once you know some software engineering principles system design data engineering whatever now you can kind of coerce that thing like how are we doing this how are we doing this how are we doing this and by the way i'm no not the greatest engineer in the world but i talk to claude code all day and it teaches me like how do we do these things better um but you know after a while you you get a framework built and so now when claude goes in it's like ah this is how we do things and this is a deterministic rails that it runs on and by the way we have a whole uh script-based testing methodology so it really puts
22. 11:46share
    Jason Akatiff it inside of a structure so we're we're to abstract this into building I would think of, I'm sure some guys here have used NADN in the past, right? You've got a workflow, that's a deterministic workflow. And then you've got the probabilistic, like, write me copy. So in your workflow, I want you to use these tools. That is, I want you to use... Cashvertising, Breakthrough Advertising, Cialdini's Influence, and I want you to use Munger's bias video from YouTube. And I want you to use these four things and I want you to use them as your framework to write copy. And I find those four work really, really well.
23. 12:32share
    Jason Akatiff and then when you're done writing that copy i want you to have two different models i want you to have claude fable and i want you to have codex let's say review that copy that was written and i want you to rate it on a scale of one to ten based on the methodologies of the life force eight you know uh cialdini's influence etc so now it's going to rate itself and if you if any of those come in under seven and i like to run two models so you get uh and then run a combined pair uh now if that any of those come in under a seven or under an eight go back and write the copy again so that that's
24. 13:09share
    Jason Akatiff a deterministic flow with a probabilistic engine inside of it

25. Talk the idea into a clear brief

26. 13:15share
    John Casto So if you were having Claude or whatever, pick your model to build this, walk us through like the general steps of setting this up correctly, and then how you would refine it, because I'm sure there'll be some mistakes it'll make in its own code.
27. 13:31share
    Jason Akatiff Of course,
28. 13:32share
    John Casto yeah. And
29. 13:35share
    Jason Akatiff when it comes to writing code so i'm going to take everybody through my general process say i'm going to start a new project uh let's call it the john casto creative generator i'm going to start this new project my favorite thing to do is i've got some ideas in my head of like what that should do how it should work what it should look like my favorite way to do this process is if you go into chat gpt app uh let's see if i have my phone here um if you go into the chat gpt app here it is there is this uh i don't know if you guys can see that see that little black button there that's the voice button this
30. 14:20share
    Jason Akatiff is the greatest greatest button that exists in my opinion of all ai um because and it's much better than claude's voice but what i do is i go on a walk and i hit that button and i brain dump i want to build a creative system we're working in facebook and tick tock blah blah blah blah blah right and i might talk about that for 15 minutes and just brain dump into it And then I'll tell it, can you take what I just told you and sort it into something that's intelligent that somebody else could understand versus what I just gave you because that was messy.
31. 14:58share
    Jason Akatiff And so then it'll structure that what I just gave it into buckets and all this kind of stuff. So that helps start building this framework out. Next thing I will say, so if you say to it, all right, ask me questions about this product, it's going to give you 20 questions at a time. So my favorite thing to say is ask one question at a time until you're 98% confident in building a product requirements doc or a PRD, because we're going to build a software platform that's going to use AI to do this. This is the plan.
32. 15:32share
    Jason Akatiff and so i need you to ask me one question at a time for all the things that i didn't think about and then i go on a walk and that conversation might be an hour it might be longer right and but through that conversation i'm learning right it's asking me what about this and i'll go oh i didn't think about that where is it going to be hosted what programming language do you want to you know don't know programming languages which one would you recommend and why would you recommend it what are your top three and then why would we choose one versus the other right and i just have a conversation with it and then at the end that's going to spit
33. 16:08share
    Jason Akatiff out what's called a product requirements doc most of mine for complex products are somewhere between 35 and 50 pages long very now that in in the all you know the old school software development world right you have a products requirement doc that was built by a product manager you know and in all the world a product manager talks to stakeholders stakeholders like who are all the people that are going to use this product okay that is a media buyer that is a data analyst that is a creative team member that is the management leadership layer that it right and the list goes on right these are all your stakeholders and they all have different needs for this tool or this platform
34. 16:58share
    Jason Akatiff So then I, you know, I think about all that. I put my product manager on hat. What does John need as an affiliate of ours? What does John need to see? What does he need access to? What post facts does he need? What systems does he need to be integrated? So I think about all of that stuff as I'm having this conversation or doing the brain dump. Here is my, like part of that brain dump. I'm like, I've got to fill.

35. A working build starts the real work

36. 17:18share
    Jason Akatiff that are my stakeholders I got media buyers that are my stakeholders I've got creative strategists I've got actual creative productivity and I just go down that list and i i say this is the experience this person's going to want this is now if you haven't sat in all of those seats you probably want to go have those conversations with those people beforehand i've sat in all those seats over over 25 years so i'm fairly familiar with them um so then that part next goes into uh Next goes into technical requirements doc, and this is basically where I tell the AI to put on its CTO hat and its systems engineer and data engineering hat, and now it's going to go build a whole
37. 18:04share
    Jason Akatiff technical spec. And what I like to do is anytime I'm building this, so like building a PRD, building a technical spec. I just tell Claude, like, run through a loop and, you know, or run these three models and then check that loop. And so then it'll go through. So we go technical doc. then we do what's called test driven development that essentially says we build all these scripts to test everything that comes out to know that it's working with evaluations then we send that into an agent team then that comes back then we run more testing then that starts the user testing and what i find following this methodology or following this approach is uh
38. 18:46share
    Jason Akatiff you know essentially what comes back works but you're going to look at it and go and not quite right change this feature change that feature etc um so it's uh you know then that's when the real work begins to make the product amazing because you as a product person are going to have your product app now this could be as simple of a product as i want to have a ad builder creative builder that's just producing ads. So if we took that example, you know, I like to copy ads that are working in the marketplace, right? So there's some systems out there. You can go, you can see what the top performing ads are. You can pull that data in.
39. 19:26share
    Jason Akatiff You can say part of my process is make templates out of all those. So, you know, turn all those ads into prompts.
40. 19:34share
    Jason Akatiff and so now i have a bunch of templates and then i have my copywriting skill and then uh i might have a library of images that i've known performed really well in the past because a an image ad template and an image at image are going to be different and so i'll stack all those things together in a deterministic way and then i'll have the ai take that with with some prompting and turn that into actual ads but that's again a really light workflow product a lot of it is handled by software you know i i watch a lot of people uh i watch a lot of people try and do this stuff for hermes but you just can't get predictable outcome you've
41. 20:16share
    Jason Akatiff got to build harnesses either harnesses or software
42. 20:21share
    John Casto in order to give the ai rails to run on you know there's so many questions i have off of this so i think just what you said right there about how to how to set it up and structure properly that's like probably its own master class yeah right just from all the trial and error you've been through the last 12 months different models too i'm sure there's a lot of consistency

43. Plan the destination, phase the build

44. 20:45share
    Jason Akatiff one thing one thing i think is i i will recommend in that sense is be a dreamer right you know in the old world of software what is the minimum viable product we're going to go to market with that's great in the old world of software and the new world of software be a dream as big as Because what that's going to help you do is it's going to help the AI build a better data architecture and system architecture to think of all the things you're thinking about in the future. So go into that PRD, dream big. If I wanted to be able to do this and this and this and this and this, right?
45. 21:25share
    John Casto And then go,
46. 21:27share
    Jason Akatiff okay, I need you to build me a system and a data architecture that can support that.
47. 21:32share
    Jason Akatiff but we're going to do a phased build so i want you to be always thinking about those things in advance but we're only going to build this one chunk that's an mvp because what ends up happening oftentimes and i've been on a bunch of calls for this somebody's like oh i'm going to build my mvp i'm going to go to market and then they on another piece and another piece and another piece and then they they basically engineer themselves into a corner and everything's breaking and all their tech is breaking and they you know they touch one thing something else breaks that's bad systems and data engineering and by the way i'm not a systems and data engineer right i just know then
48. 22:10share
    Jason Akatiff that stuff is important so i talked to claude about it i'm like i'm not a system and data engineer i have a little experience in analytics and and data from all my years of buying media
49. 22:20share
    John Casto so you know i need you to put on your data and systems engineer hat and here's my big dream how do we architect this so that we don't engineer our way into a corner I think that's really really important so I see what you mean there and I've like from personal experience I'm not a dev you know this but obviously like you I understand what needs to be done just from the business of media buying and we've been able to build a lot of really cool internal tools but I see how they break just from our internal usage, but we're also not building with the bigger picture, the big dream, because it's just, we're building it for a simple use case.
50. 22:58share
    John Casto But then when we try to tie tools together, that's, you see how like complex this is, even with the amazing capabilities we have now. But I've got a million questions I could ask you, but this isn't really about me today. So I wanted to jump into... We had 250 registrants for this. I think about 30% of them asked multiple questions. I had Claude consolidate the questions into subsections and then questions within. So I'm just going to rapid fire these, try to get me consolidated like quick hit answers, and then we'll see what comes from the live chat for everyone that's on here live.
51. 23:39share
    John Casto and what questions they have as well and guys uh just as a reminder i am recording this i know everyone's trying to join with note takers i keep accepting the keeper re-requesting this will be put on youtube um so yeah so let's jump into the questions and oh uh before we do i do want to say i said jason how do i put together a quick registration, tie in the Google calendar, send out notifications, everything. He's like, and then he answered something because I use his software for my actual business. And then he said, just use Lee router, do this. I was like, no, no, dude, I'm talking about the webinar. He's like, yeah.
52. 24:19share
    John Casto So we actually built this in like an hour. just working with claude tied it into the api and the mcp server with json software and we're able to put this whole thing together i didn't finish routing the email and sms notifications but like obviously you guys all got the google invitation so i did something right so shout out to jason okay so if you were starting out today jason um we have four questions within the section the first one is if you were starting a business from scratch today what exactly would you build what vertical which paid channel it's a great question

53. Choose a route that fits your skill

54. 24:58share
    Jason Akatiff you know i i think of that in uh i think of that in multiple tiers right and then kind of first question I always ask is what's your core skill set? You know, are you, do you lean more towards the product person? Are you more a salesperson? Are you a media buyer? Because that would, that would vary, right? If somebody is like, I don't really know how to use any of these tools, but I'm a great sales guy, I would give a different answer than, you know, it would building up demand in something, right?
55. 25:35share
    Jason Akatiff Whether that's home services or whatever, like building that demand up and making your margin, doing that thing that you're great at and then understanding how to tap into the supply side of the industry. Hey, I'm a really a great media buyer. You know, where should I go from here? Well, like, you know, it really depends i think that's just so so hard to say right uh let me go media buyer just because i think that's an easy one you know media buying uh be an affiliate I would tell you first, you know, first and foremost, it sucks to be an affiliate. You know, it's hard. You'll figure something out. Your shit will get stolen. You'll be successful.
56. 26:21share
    Jason Akatiff There was an advertiser we worked with. We were doing $25 million a year, and they ripped all our ads and cut us off as an affiliate. Like, it's a brutal place to live.
57. 26:32share
    Jason Akatiff um and and i wouldn't say stay there forever but i've had affiliates i i've got a single affiliate that ran with us he's out of the space now but he scaled up to 35 million a year in revenue as a single affiliate um so like there's a lot of scale there you know if you if you stick at it um i would all i would always recommend also whatever you choose stay focused like just become the have the goal of becoming the best in the world at that you know i think entrepreneurs which i think is a lot of people on this call
58. 27:07share
    Jason Akatiff know it's like shiny bobble syndrome right like oh i want to be a media buyer oh maybe it makes sense to be an advertiser now maybe right it's like my question is like have you built the systems around that skill that you've already created that those things will continue to run without you there and whether you do that with ai which is totally possible now uh ai tooling and systems or whether you do that through people and teams it's it's uh totally possible but if we go back to hey i don't you know i'm brand new to all this i always say to start with legion i think legion is the lowest barrier to entry and it really comes down to data

59. Buy a feedback loop you can afford

60. 27:47share
    Jason Akatiff When you're running Legion, your conversion rate is somewhere between 2% and 50%, right? And that feedback loop is so important because let's take Mass Tortoise as an example. If your feedback loop is... Hey, I need to generate a lead that leads going to cost me somewhere between 150 and $800 to generate just for the lead and that lead is going into intake and intake is, you know, they're going to convert at 5% to 10%, maybe as high as 12.
61. 28:26share
    Jason Akatiff so now i need twenty thousand to fifty thousand dollars to for just a couple of ads to get a very basic signal and all this all media buying is is it's a combination of creative and looking at data signal and analytics right so if you enter i don't know debt or uh some home services stuff like your conversion rate might be 10 or 20 percent and so your feedback loop to success and money is much tighter now i only have to spend 200 and i'm gonna see some leads show up well in mass tort you might spend 10 000 to start to see some retainers
62. 29:12share
    Jason Akatiff it's a very different game so think about as a media buyer and as affiliate you know what is that high converting thing where you can start to build the muscle nobody starts in mass torque right you start in auto insurance or you start it because there's a good signal right and think really thinking about uh thinking about that signal back and back and forth i think it's really

63. Look where you understand the customer

64. 29:37share
    Jason Akatiff um i think there's a tremendous opportunity in this this this new era you know if you're like hey i really i i kind of like the media stuff but really i'm a product person i'd love to build a product business you know there's people on this call there's people in the bulletproof chat there's people in other chats you know maybe partner up with somebody that's going to do the media buying and be like i'm going to build the product and i think I think the biggest opportunity for the next five to ten years is two-fold. Number one, all software is going to be rewritten and rebuilt. Like the software of olden days is going away.
65. 30:21share
    Jason Akatiff you you don't have to be a software engineer or anything like that anymore to build software so we're gonna see a lot of fragmentation in the market that's industry specific so if you're like hey i know uh i was a gardener for one point in my life i'm gonna build the crm for gardeners
66. 30:37share
    Jason Akatiff that opportunity is out there right and i'm gonna charge whatever now you got a digital product and that opportunity i think that opportunity is tremendous whether it's building harnesses for ai to solve problems uh people call them agents it's building software like to me if i was a product person this is where i would be focus, as you know, I already am on lead router, the listicle builder, stuff like that. That opportunity is there, the costs and all that kind of stuff that you used to have to do back in the day, right? Like, if I wanted to build lead router, I need to go higher, I need to raise capital, go higher 50 to 100 developers.
67. 31:16share
    Jason Akatiff And it would take probably two years, because the development cycle product person to project manager and CTO and then that turned into scrums and then that turned into goes to elite dev devops whatever that goes into developers oh there's a two-week sprint that's run for that development work at the end it deploys back sometimes there's real-time cicd but that now the product person looks at it goes that's not right and we run through all that again that's how enterprise software is built i can i can basically do all that by myself with ai and none of that i could be the product person i could be a really lightweight technical person that i've learned
68. 32:05share
    Jason Akatiff um and i do have some coding background like back back in 2003 i wrote a lot of php stuff for like three years but never enterprise grade development So you can just have it teach you this stuff. So I think that the software and harnesses is a huge, huge opportunity for anybody. If you have experience and skills in a space, you understand a customer really, really well. You know, I think the default for any affiliate is like, I'm going to build affiliate tools. I would recommend against that, right? Like every affiliate I know is building their own affiliate tools.
69. 32:41share
    Jason Akatiff think it's really a hard place to live um but if you're like hey i i really know restaurants and i know the biggest pain point of restaurants is this as it relates to marketing i have an idea for a concept or i'm going to talk to claude about a concept or chat gpt and then i'm going to go try and build that and i'm going to put it in front of a couple of restaurant owners and then if that works then we start buying media and start getting subscriptions and people on the platform right so that's i think one of the very very largest opportunities i think the second largest opportunity is ai is a transformational shift
70. 33:19share
    Jason Akatiff so much like we had the transformation from the industrial age or into the industrial age then we then had a transformation into the information age that was the internet and computers we have another opportunity which is another transformational shift into the ai age what i'm doing with ai today i'm not saying special i've just put 2000 hours into all the workforce will be doing this in somewhere between three and five years so the question is who's going to train these people to use these tools because it isn't as simple as you just talk to the ai and it does everything for you that's not how it works AI is very much like employees. It's like dealing with the employees, right?
71. 34:05share
    Jason Akatiff So if you've had employees, you're really used to managing them, keeping them on task. People are like, they hallucinate. I'm like, well, people hallucinate, right? Oh, they're hard to manage. Well, people are hard to manage, right? I see AI very much in parallel to humans. and granted it will it get better over time yes will it produce something that's of quality yes if you keep it narrow enough scoped and you understand what it's going to deliver so i think our whole workforce of the whole world needs to be trained to use these tools whether that's a these tools for prospecting, whether that's a developer needs to use these tools for deployment and production, whether that's a product person, whatever that is.
72. 34:53share
    Jason Akatiff People need to learn to use these tools. And I think retraining that workforce is going to be a trillion plus dollar opportunity. coaching and then consulting and implementation, I think are the other really, really big opportunities. So if you're a product person, that's the direction I would go. And then if you're on the sales side, I think you know building out sales teams and capturing demand for whatever space that is software sales uh final expense agent buyers whatever that might be is is kind of that third pillar depending on the core skill set two-part question going off of this so what is something you

73. Learn through disposable experiments

74. 35:36share
    John Casto believed AI was capable of doing the way you predicted it that turned out not to be true the last 12 months? And then second, are you more optimistic or more pessimistic from your position six to 12 months ago about people being put out of work and bad things coming from this?
75. 36:00share
    Jason Akatiff both great questions um let's talk about the first one first i never i think the worst thing that you can do in the world period is make assumptions i don't make assumptions like i assume i don't know anything in every conversation i go into sure i come with 25 years of experience but john and i
76. 36:23share
    Jason Akatiff i listen to where he is what his pain is what's going on how to help him right and i don't make any assumptions that i know the answer i tell him what to do right so and i i think you have to go into ai the same way and i think it's really important uh much again much like employees somebody comes to me they're like i'm a media buyer i did xyz blah blah blah blah blah and i'm okay well that's that's cool you know my so my assumption based on what you're telling me is is you're a badass so we'll bring you in here and let me see you can do right and i do the same with ai so i spin
77. 37:02share
    Jason Akatiff up a lot of where i think people get scared is they're like here's my baby here's my one project i don't want to like do anything outside of scope or that's going to break it you know and i'm going to encourage everybody to be starting new projects all the time that are what i consider throwaway projects And this is how you learn the boundaries of what an AI or a human employee is capable of. You go in, you know, I dump that, create that PRD, I give it to Cloud Code, and it's 50 pages long, and I just see what it comes up with, right?
78. 37:44share
    Jason Akatiff you know a lot of times i'm way over impressed and it's able to do something that i didn't think was possible and i think this is what begins to expand so like anytime new models go out a lot of times i'll go through this process i'll take some old spec that i have and i'll just say hey build me this and i'll see what it what it comes up with right and that's uh i think it's you know it's it's it's really important to constantly assume you know nothing and then understand how you how you learn within the scope of this human or this ai on your second question uh ai is not going to replace humans the way i think about this

79. Expertise still judges the output

80. 38:29share
    Jason Akatiff is i went to school for computer engineering for a very short amount of time until they kicked me out because i surfed twice a day and i didn't go to class uh but it takes me back to that so if we look at a microchip i'm sure some of you know about binary code which is essentially ones And in the early days, the way people built software was kind of switching those gates, and then they built it into cards, and then they built a language on top of it called Assembler. So Assembler was a way to switch those gates at a higher level. So I say this one thing, and then seven gates switch rather than me manually having to switch each gate.
81. 39:21share
    Jason Akatiff Then you had Assembler, then you had NCC come out, then you had C++ come out, then you started having PHP and all the web-based models that we know today, or languages like Python, stuff like this, right? And each layer of that is an abstraction away from the gates on the chip. Well, guess what?
82. 39:44share
    Jason Akatiff ai is just yet another abstraction layer away from that gates on the chip and so the way i think about it in business and i'm a huge follower and i recommend all in podcast to everybody um all the time it's my favorite it's for venture capitalists that are chatting about technology and politics and all uh capital and all that all the things that go into that a huge fan um but you know they're typically a bleeding edge in the market and what's going on because they're all silicon valley guys they've all moved out of silicon valley because of california's craziness but uh they're all silicon valley originally guys and uh with that they you know they're have a pulse on what's going on
83. 40:30share
    Jason Akatiff and so at first everybody was like laying off like oh we're not going to need these employees we're not going to hire but now software engineers are starting to get more and more hired and and specifically software engineers and i'm going to use that because they're typically on the leading edge of this um so as people learn to use these tools their productivity is 10x 50x what it would have been so if you take somebody that's smart already really good at a job they understand the scope they understand how it works
84. 41:04share
    Jason Akatiff and then you arm them with these tools or they go and learn how to use these tools now they can be 10 to 50 or maybe as much as 100 times as productive as they were before and i think where humans really play a really important role here is great i am a media buyer i know media buying really really well well ai knows media buying based on all the things that it's And if you ask it to build you a media buying system, you're going to get a shit product. But that person that really knows media buying well is going to be able to say, no, don't do that. Do this.
85. 41:40share
    Jason Akatiff Another example here, my good friend Julian Reyes, you know, he writes VSLs, right? Forty five minute VSLs. And, you know, he was kind of shied away from from using Claude for copy.
86. 41:53share
    Jason Akatiff and you know i really pushed him and pushed him and he finally got into it he's like you know it's giving me good stuff but it's not you know it's it's a six out of ten it's a seven out of ten and then i work through it and we go through and it gives me ideas and we go through and we change the copy we change the wording whatever we go through this vsl and i can produce as high a quality or higher quality vsls than handwriting and rather than doing that in three months i can now do that in 16 days and so so i think ai and the tool in the hands of very smart intelligent people that know their discipline
87. 42:34share
    Jason Akatiff is where it where it's going to win it's going to it's going to take time it's going to probably be some disruption like there was in software engineers right there was disruption but now the software engineers are learning to use these tools so if we think about i run a company like you john casto you run a company you're hiring and your market competition all the employees know how to use ai tools so their cost is this much to produce this much outcome and i'm not using ai enabled employees and my cost is this much
88. 43:17share
    Jason Akatiff that's why i think retraining the workforce to use these tools is going to be so important because in business i if i've got if i can get more and more people and i can grow the business and i can get ai enabled employees to do that the productivity out of that and what that is going to allow me to do as a business leader is next level right i had all these dreams and i think a lot of people I talk to in business that are deep in this AI stuff. They're like, dude, there's like a thousand things I wanted to do over the last 10 years, but like human capital is just expensive. And, you know, we had to prioritize something else.
89. 43:55share
    Jason Akatiff Now I can do all those things that I thought would make my business great. And I could do it with smart people wielding these tools.
90. 44:04share
    John Casto So finding A players and empowering them, supercharging them with tools is the answer. I mean,
91. 44:12share
    Jason Akatiff I had some devs that, you know, they wouldn't embrace the tools. And I'm like, sorry, guys, this is not the place for you. If and when you go learn these tools, you know, we'll consider hiring you back. And they did. And we did.
92. 44:27share
    John Casto So where do the junior roles play? Media buyers, devs, where do they fit into this equation if everyone's hunting for seniors?
93. 44:35share
    Jason Akatiff I think that's the hardest part. You know, I think the future of education is a big question mark, right? Like, what does that look like? There's something called Alpha School. They talk about it on All In podcast a lot.
94. 44:51share
    Jason Akatiff and it you know alpha school is two hours a day but it's all ai enabled learning so it's like you have a teacher and a tutor all all combined together and the scores that kids are getting out of two hours a day of education in alpha school are way better than the scores they're getting out of an eight hour day sitting in a boring ass classroom which i hate school personally um so i think the future of education is going to look different i think how do we bring up people and i you know we think about the information age there was a lot of people that worked in factories didn't know how to use computers but then you wind up with a
95. 45:28share
    Jason Akatiff generation of people like me i i started playing i was lucky enough i grew up in cupertino where apple computer is and my dad was running a business we had a TRS-80, which was an old school Tandy computer. I started playing video games and using BBS modems to kind of work on the pre-internet, play video games with people. So I think just kids as they grow up, they have natural dexterity with this stuff. And as they enter into the workforce, they already know how to use these tools'cause they're second nature, right? It's like Oh, how are we going to retrain people to use a phone, a mobile phone, right? Well, my kids are way better.
96. 46:08share
    Jason Akatiff At one and a half years old, they had iPads. They could unlock the iPad, go to YouTube, find their stuff before they could talk, play games, whatever. They couldn't speak, but they could control that iPad. you know i think there's a a period of time here change is hard for people uh learning new skills is hard for people so i think people have to decide you know am i
97. 46:35share
    Jason Akatiff you know i think there's a place for teachers with expertise um again on that retraining the workforce where it's like oh hey i'm an expert at this i don't know how to use these ai tools but you know i can guide you guys on how to uh somebody that knows how to use these ai tools to use my brilliance or my excellence all my years of wisdom and experience and implement them in these

98. Look for a temporary advantage

99. 46:59share
    John Casto going to jump into the next section of questions future of the industry so i'm going to spit several questions at you and i want you to just give me because there's so many questions to go through yep so first question in the section is will there still be a future for affiliates brokers agencies egen paper call will humans keep any edge or do agents eventually replace us
100. 47:22share
     John Casto the future of pay-per-lead you left nba when it became a race to the bottom how do you see finding success in paper lead now third question when everyone can pump out insane creative volume with ai what's the actual competitive edge for a market yeah however you want to answer those three because i think there's some overlap there too the future
101. 47:46share
     Jason Akatiff We go back historically, like when I came into this space, everybody was emailing and CanSpam came out and it was like, oh my gosh, everything's going to end. Nobody's going to make any money. The industry is going to go away. And then, you know, it's consistently something's always happening, whether that's in the world or whether that's what's happening in technology. But things will... move forward right innovation happens and where i think affiliates really shine is innovation right new platform comes out right i remember the early days of facebook or i just to give you guys some example when i started in this business facebook myspace none of this existed right everything we did was on forums we talked on forums
102. 48:34share
     Jason Akatiff right none of these platforms existed face uh google barely just came out with uh or had they bought double click and you know uh buying display and and on google was like brand new right so if you think about those times um and and i think this is important as well what what oftentimes humans do we as humans we understand the now. That's where we are today. That's what we understand. And so we take the future, technology of the future, and we put it on the today.
103. 49:09share
     Jason Akatiff And we go,"Well, the today, it's going to put everybody out of work." But actually, AI is going to create a lot more business and a lot more opportunities and things you never thought were possible. So if we go back to 1993, when I first got on the Internet, you know, through AOL, you know, it was like you could say, oh, all this is going to happen. And it's like, oh, we're doomed. Right. Like because I'm projecting the future on the now.
104. 49:40share
     Jason Akatiff that's never how it works as things evolve things grow opportunity gets created the pie gets expanded there's more people on the internet there's more people using tool like innovation creates opportunity and I think affiliates are the best people in the world at figuring out this traffic source doesn't work everybody says it doesn't work I've got it cracked right and what's going to happen is in one or two years from now that traffic source is going to become industry standard everybody's going to have it cracked right and so i think understanding innovation and understanding where how do you stay if you just are doing the same thing everybody else does you're not going to win as an affiliate but if you go find those
105. 50:29share
     Jason Akatiff pockets i'm going to go be the best guy in the world advertising on reddit you know uh it's hard there's this but like i'm the expert at that or i'm the expert at snap or i'm the right like carve out your space that's leading edge that all the other people haven't figured out yet and that's where you find your alpha in the market now if you're like i'm gonna run payday loans and i'm gonna put it on uh google search and i'm gonna build on payday keywords That's not possible. There's no alpha there. All the margins have been sucked out of that market. The big players own the payday companies. You can't win.
106. 51:06share
     Jason Akatiff So always be thinking about where can I find opportunity, whether that's a traffic source, whether that's a vertical, whatever that is on the affiliate side, but there will always be opportunity. I think we've seen pay-per-call over the last five years or whatever it's been come up, right? Like if you went 10 years before that, you know, nobody was talking about pay-per-call. Oh, pay-per-call is going to eat the lead gen business. No, it's not. It's just going to create more market share.
107. 51:36share
     John Casto What about brokers and agencies? Are you seeing them get pushed out and squeezed or compared to affiliates?
108. 51:45share
     Jason Akatiff yeah i mean i i think there's a certain healthy amount of that um when i came into this space i first uh i first you know bought ads for dough most of our offers were somewhere brokered between somewhere between three and five times jumps right you could see the redirect links and all of those people taking a piece of the pie or scrubbing or whatever they were doing right And there was enough margin in the market that all those people could take a piece. But as the margin squeezes out of the market and it gets more competitive, well, those people can't take a piece anymore and that is going to consolidate down.
109. 52:26share
     Jason Akatiff So you're always going to see this ebb and flow of a market. like if there's a ton of margin in the beginning you have a ton of brokers but as that margin consolidates over time you have less brokers and less brokers but guess what the the thing about brokers is they're like cockroaches they just go find something else to broker because there's always inefficiencies in the market so creative at scale this is our next most popular section

110. Turn creative feedback into a process

111. 52:57share
     John Casto sub headline that AI gave us was AI slap came up verbatim for multiple people. The room wants volume but with quality control. So a question in this section is mass video creation with a creative director's QA. quality assurance baked in. How do you keep it from becoming AI slop and wasted credits? And I'll add the second question is, how do you make great static ads with AI concretely? What's the current stack?
112. 53:29share
     Jason Akatiff I mean, this goes all back to the first conversation, right? Like that's the most important thing about thinking about business systems and workflows, whether it's with people or whether it's human and AI, they call it human in the loop. Whether it's people or whether it's human, the better the systems are, the better the workflows, the better structure there is around it, the more probabilistic your outcome is going to be. So yeah, if you're just like, hey, I've got this cool prompt, I'm going to put it into cloud code.
113. 53:59share
     Jason Akatiff uh and you know it's gonna make me cool at yeah maybe you know you're probably gonna get a certain you know in a good ad you know whatever percentage of the time but if you put a little harness around that some skills uh you know you're gonna get better output and like you know you can go deep into the harness or you can or you can go oh i'm going to go more into the workflow side of things again visually imagine innate in here steps in a process in the olden days we call them standard operating procedures or sops is what a employee would follow the steps right do this then do this then do this well ai is the same so if
114. 54:38share
     Jason Akatiff you give it a rails to run on or a process to run on the probability that you're going to get a better probabilistic output from the deterministic path that this probabilistic thing runs on is better right that's that's how you stay away from the spot I think I would think about like I would track percentage wise. So like if I'm to build a system for this, you know, here's my rails. I've got human in the loop at the end. Right. That's your creative strategist reviewing it. Now, the way I'm probably going to run that is I'm going to on a scale of one to five, one to ten or a thumbs up, thumbs down.
115. 55:19share
     Jason Akatiff rate the quality of this ad right from that human and so that they're going to go in oh this is that now that's going to feed that back into your harness model that's train then and now it can start to build a loop where it trains it you know trains itself based on that human feedback and over
116. 55:42share
     John Casto you know that human feedback that that harness and that's the skills around that harness will get better and better over time so would you say there there are fully automated copy plus static generation where a human absolutely through his drafts yeah i i was at uh geek out mykonos and also uh i spoke there for james and uh nick shackleford and then i was at um
117. 56:06share
     Jason Akatiff barcelona affiliate takeover which was a great show and uh i'm speaking for alex there you know the how am i thinking here uh i saw speakers talking about they already have systems where they're automating image ads and or video ads right andre who you might find on twitter like they're automating both image and video putting it into big caps and you know you know they're doing quite a bit of spend every month they're only two guys uh big systems automated the nice part about workflows and systems is you can improve them over time well i don't like that output let me make some tweaks to the workflow so now that
118. 56:51share
     Jason Akatiff is going to be better right so you've got this workflow with all these nodes or steps to it and so like once you have that structure that system now you can work on improving that oh i need to add another step in here oh i need to add a human in the loop in here and over time you find a workflow that that really performs well i think jordan uh
119. 57:13share
     John Casto Jordan did a really good demo in Mykonos at Geek Out.
120. 57:16share
     Jason Akatiff Yes, yes. He was like,
121. 57:18share
     John Casto he had a feedback loop of winners like based on the data of the performers. That guy,
122. 57:24share
     Jason Akatiff he's brilliant. Yeah, that guy's super smart. I would definitely follow him. I don't even know if he's on socials or what, but I met him there. I spent a lot of time with him. But yeah, you know,

123. Give design a structure and a vocabulary

124. 57:35share
     Jason Akatiff high you know number one recommendation for ai the more you can give it uh something that's working with structure and a framework to work within the better so we talked about a you know a system to produce this stuff but then there's also okay i want to build a landing page cool well how do you build a good landing page well it's a page composure like you know page layout it's copy it's imagery and then there's a design system right like these are your four pillars in a great landing page um you know uh and then you know there's skills on top of that that is like ui skills or you know some stuff like that it makes things convert better or worse
125. 58:23share
     Jason Akatiff that's some skills that lays on top but it's really those four pieces so like if i'm to go build a landing page i'll use that but this could be an image ad or whatever So I go find a page I like and I say,"Hey, go pull me the wireframe." And one thing about AI that's really important is use words it knows. If you can learn the words it knows, you'll get better outcomes. So one of the things I've started to put together and I've probably published on Built with AI is just vocabulary, right?
126. 58:56share
     Jason Akatiff uh what does uh you know ai knows the word triage it knows the word frameworks it knows the word wireframe right you can say hey could you pull the structure out of this thing and da da da da or you could say just can you pull the wireframe out of this page and it knows exactly what wire frame means so as you're talking to ai over time and you see it use the same word three or four That's a memory peg to say, I need to learn this word. Let me write it down on a piece of paper. Because when I talk to my AI, now it knows exactly what I mean when I say this specific word.
127. 59:31share
     Jason Akatiff So if I'm building a landing page, I ask it to extract the wireframe out of that. The next piece, I'll have it go in and say, pull out what's called the design system. and that's your colors and fonts and you know page spacing and all this kind of stuff that's called a design system uh next i'll say uh from a content perspective what's in it by section right okay there's the hero up top headline subhead then there's a section about problem agitation then there's a section about whatever And then I'll have it go into each section and and kind of analyze why that copy is there and what potentially is making that copy perform there. Right.
128. 1:00:18share
     Jason Akatiff So now I have, you know, this kind of three pieces and then. you know from there my favorite thing to do is like i go to dribble uh you could go to behance as well these are design sites where people you know they post up their designs their designers doing stuff so i love to use those as resources to go and find a design i really really like and so i go find a design i like so now i have this framework i go to find a design i like
129. 1:00:53share
     Jason Akatiff and i bring that those screenshots in i just drop them in and i say okay i want you to build me a landing page maybe that's for debt i want you to build it for auto insurance so now it has a framework of how the whole thing lays out it's got a design system it's got example designs now you're going to get a much more probable much more deterministic outcome from the AI and then you can continually layer in more steps to that process through your harness or through your workflow.

130. Keep tactics provisional

131. 1:01:28share
     John Casto next section media buying and offers i'm going to give you two questions i want you to pick one first question more technical you find a potential winner what happens next scaling cbo it's an avio cost controls variations walk through the sequence second question to choose from which legion vertical has the most potential and highest margins right
132. 1:01:53share
     Jason Akatiff So the first one, I'm not in the trenches of media buying at all. You know, if you guys haven't been to Built with AI, there's an open source section. There is something on there I call Breadwinner, or I think it's called, I think on there it's called the iScale. uh facebook ad platform so if you're into ai and you want a a tool to build ads that's already got a framework for you um and also uh can hook into your facebook uh and deploy ads through it there's a full open source tool there for you to that you can build on top of. You're like, I don't know how to build a framework or software.
133. 1:02:35share
     Jason Akatiff Cool, go download that, grab the GitHub, put it into your cloud code, and say, hey, there's this repository. I want to host it locally or host it on a cloud player or something. Can you pull this down and set it up for me? And it'll pull it down and set it up. And then now you've got a structure or a framework to work within to build and deploy ads.
134. 1:02:56share
     Jason Akatiff so i work on the problem from that level i think of everything in terms of systems so a lot of it's you know the thing about the media buying is it's changing each and every day right like i can tell you right now oh this is working for us and i i talk to the biggest and the best media buyers in the space all the time they're like oh cbo's are working and then it's a week later and they're like oh single ad and single ad set with abos are working now it changes every day so back to my very first statement on this call don't assume anything even if you it was working yesterday assume it's going to be working different tomorrow
135. 1:03:33share
     Jason Akatiff and you've got to keep Myoplasticity, which means keeping your brain flexible. And you've gotta always assume you don't have the best answer and you've always gotta be looking for the new one. Two ways to do that. Build yourself a strong network of people. Bulletproof Scaling's got a great network of people. Build your tribe. You know have four or five other people in there make your own little telegram group and share tips Right like this is what affiliates do. This is how we stay ahead of everybody else This is all where all the cutting-edge stuff on the strategy side Exists, so, you know or go to trade shows, you know, definitely go meet people in person
136. 1:04:16share
     Jason Akatiff And so that's that's a big big piece, and I attribute that to 80% of my success, and then the other 20% is just hard work. Just be testing, assume you don't know anything, assume the market's changing, assume the ad networks are changing. be the first to test anything new that they roll out you're like oh facebook rolled out whatever i hear it's garbage that's my trigger to go yeah i'm going in you know that that says there's opportunity here right so nobody has cracked this yet that's where i'm going to find the outcome and which legion vertical in your opinion has the most potential and highest margins Depends on the affiliate. Depends on their structure.

137. Design the buyer side of the business

138. 1:05:00share
     Jason Akatiff You know, I think calls is, you know, we did a lot in leads for life insurance. I think calls for life insurance is super interesting. I think
139. 1:05:16share
     Jason Akatiff you know a lot of the uh as i see it a lot of the future of calls and and lead gen stuff is going to be just selling direct to broker direct to agent not into these big floors and then having software to be able to do that and support that just again as that as that market continues to uh continues to consolidate date gets more competitive how do you find your alpha your alpha is going to be by hey maybe that's immediate buyers uh connecting up with a sales guy sales guy you go sell to insurance brokers or you go sell to a gardener commercial gardeners right bring in the direct buyers
140. 1:06:00share
     Jason Akatiff You know, make sure your systems are automated and efficient in order to bring in those buyers. But that's what's going to really create opportunity for you, you know, longer term. I think this is a lot of what the future looks like. Different verticals are very different when it comes to calls and leads. You know, somebody putting in a a thing because they want luxury furniture or they want a quote on a backyard, make their backyard look beautiful. the probability that they're going to answer that call or respond to that text message is much higher, in my opinion, than if you're running, say, auto insurance. You know, auto insurance just gets tougher and tougher. You know, people get bombarded with stuff.
141. 1:06:50share
     Jason Akatiff So I think it depends on what the vertical is and then really, how do you want to run your business? Do you want to build a large business with infrastructure that is salespeople, technical people, right? Do you want to go that route? I would look for a larger TAM market, like total addressable market to build that in. If I'm a single affiliate. You know, I'm probably looking for some kind of a vertical where I can kind of carve out my niche. You know, oh, I'm the guy that does palm.
142. 1:07:22share
     Jason Akatiff tree installs in sunny you know sunny areas like we really have that on lock we know those ads work and we know the commercial gardeners that put those palm trees in and we've you know you can build a 20 million dollar a year business and calls or leads just doing that or uh you probably know these guys have gone and built a tremendous business about just leasing porta potties right and they've taken that whole value chain consolidated it squeezed the margin out of it they're still they buy a lot of affiliate traffic on the calls and lead side but they're really they're the buck stops they're not brokering anything there's no steps in that
143. 1:08:10share
     Jason Akatiff There's a lot of, there's just a lot of ways to think about, there's no one answer to any of this stuff. It's how does somebody want to run their business?

144. Make operations keep working

145. 1:08:18share
     John Casto Next section, agents and AI workflows. One question is, what are the top things AI runs autonomously in your business today? And which part of the workflow of those was the hardest to get actually working? Oh,
146. 1:08:36share
     Jason Akatiff you know, because we're in such a data heavy business, to me, if I look at lead generation, call generation, to me, 95% of the problems are deterministic problems. That is call routing, lead routing. data analysis media buying effectively is is data analysis and analytics right like yeah there's a there there's a huge lever is creative but there's creative and then there's everything else and creative is very important but It's not all that important to scale. It's a piece of scaling a business. And don't get me wrong, I'm a huge advocate of great creatives. And, you know, that's a huge lever inside of the business. But also another huge lever is just the operational efficiency of how the business runs.
147. 1:09:34share
     Jason Akatiff And then, you know, I think if we go back to the beginning. know harness workflows you know you should con like whatever you're doing that's where your focus and your attention should be right there was a big mad rush of like open claw yeah everybody jumped on the open climb for me i was like nah i'm gonna stay what i'm doing and i just stayed the course right and a lot of those people that were open claw and hermes people have now come back to cloud code and said okay i need to build harnesses and frameworks and workflows in order for this stuff to run on because I'm just not getting the deterministic outcome I need.
148. 1:10:16share
     Jason Akatiff That doesn't mean it doesn't have probabilistic pieces in that workflow. But at the end, I want a system that I can feel confident about. I can go on vacation and it continues to produce results for me. So if you're going to put your focus, put it on those.
149. 1:10:33share
     Jason Akatiff you know workflows and harness and and again i'm you don't need to be an expert at this stuff cloud code's already an expert right so you can just talk to cloud code and say hey i'm stupid which i do all the time or i'm not smart enough to understand what you just said could you explain it to like like i'm a third grader because sometimes i feel like it right like whatever right and it'll break it down into to very simple terms and and if you put enough hours in and over time and you you just assume you don't know anything and anytime you don't know anything definitely ask what the hell do you mean by that can you explain it to me
150. 1:11:10share
     Jason Akatiff in simple terms or can you walk me through what those steps would look like or you know just get really in the habit of working with because where you'll be uh i worked for tony robbins as a speaker for a number of years and you know he talks about just one percent improvement today on an angle you know over a very long period of time puts you at a very different destination
151. 1:11:35share
     Jason Akatiff So if every day you're just adding a little bit more to your AI vocabulary, a little more to your software engineering understanding, a little bit more to how to build harnesses, a little bit more to workflows, where you'll wind up in a year to two years from now is just going to be such a different place than you are today.
152. 1:11:57share
     Jason Akatiff what was your biggest win the last six months with ai uh you know i've been i've been locked in you know building building lead router um you know lead routers eight hundred thousand of code and tests and you know up until a few weeks ago I was the only developer on it and so it was very much you know what was my biggest win I mean we've got a lot of great customers that have signed up for lead router and you know what makes me happy is watching people grow their business watching people learn watching people be successful being able to help them with that teaching them stuff i love to teach i love being a teacher
153. 1:12:49share
     Jason Akatiff um so you know that my biggest win is people getting on that platform and again i'm not trying to sell it i'm just that's what i do every day i talk on telegram as a lot of you already know And I work on lead router, right? And then I onboard customers and show people the power of what the future looks like. And so, you know, hearing feedback, there's a few people that have said, like, this is unbelievable. Like, I've been on here two days and I've already figured out how to do everything via cloud code. That's another thing for me is, like, getting people to use Orca or anti-gravity and cloud code.
154. 1:13:26share
     Jason Akatiff some apis and really start to understand the power of that i mean john you and i did it together right i walked you through that process i don't know how long ago it was um but you went all in right like that is what i love to see give those people those aha moments where they're like holy this just opened my mind to something i didn't even think was possible um And I'll be publishing on our YouTube, the Lead Router YouTube channel, I'll be publishing a lot of that training and materials. But I really, you know, if you can, if you can have a place in people's mind, it's like, hey, I started this with Jason.
155. 1:14:05share
     Jason Akatiff That's a special place for me to live in people's minds. I'm long past the point of needing money. I live a very fairly simple life. you know i i love to work on the computer i love to build stuff i love to build business i love to work with great people um so you know i just think about all those all those things next section wrapping up here favorite call and lead ops here's a fun one for ai

156. Delegate the build, review the result

157. 1:14:34share
     John Casto Where should the human/AI line sit in the next five years? But let's say the near future. What do you delegate and what stays human right now?
158. 1:14:44share
     Jason Akatiff I think all of it gets delegated to AI in the hands of an expert, right? What's that?
159. 1:14:53share
     John Casto It's got to be a combination.
160. 1:14:55share
     Jason Akatiff yeah so it's who is your add-ops person or who's that great add-ops person that you get trained up on using ai so like in lead router you know i could just take uh posting like well i'll use leads or calls it doesn't matter because it's in our system leads and calls are the same right they all work the same way ping post is the same as rtb it's all homogenous and all works exactly the same way um so it's not like a bolt-on later um so you know you just go into our internal agent and it'll open up the internal agent which is essentially a harness on top of opus where you could just drop your posting specs in or
161. 1:15:36share
     Jason Akatiff you could drop your rtb specs in and it'll just set it all up for you it'll run all all of the tests so i think where the where the what the future looks like is um you know uh if we talk about leads if i get a leads prosper posting spec i drop that into cloud code or codex that's hooked into
162. 1:15:59share
     Jason Akatiff uh lead router via api and then and then cloud code has our api docs and our user docs and i say i want to set this up the advertiser's name the price is going to be y it's a ping post uh and then i want to put attach it to this offer or i need to build a vertical this is a totally new space i drop the posting specs in and in like two minutes it's fully configured
163. 1:16:22share
     Jason Akatiff and then if you're if you're an affiliate or not like you can go into the campaign which is where the posting specs live you can click uh build me a prompt and it'll take you through a walkthrough it's like do you want a single you want just a quiz flow do you want a single page do you want a multi-page how many pages What like page structure do you want? Do you want content on it? Do you want which tracker using using clickflare red track whatever and then what do you want to use for privacy policy terms of service? Trusted form do you need all that?
164. 1:16:57share
     Jason Akatiff Yep Check all those boxes and our system builds you a prompt so rather than a page builder you get a prompt Now I take that prompt, I plug it into cloud code with my API key for lead router. And in five minutes, I've got a fully optimized funnel. So like in that prompt, there's a bunch of stuff like needs to be under one second, you know, needs to compress images, needs all the stuff for best practices for building landing pages. And so then it builds it. And then it uses the API to go into Lead Router. It sets up and configures it all. And then it opens the browser and starts running through that.
165. 1:17:32share
     Jason Akatiff And then it looks in Lead Router and sees if it's working. And then once it's done, it's like, OK, it's set up and configured in like five minutes. This is what the future looks like, right? It's now I'm going to look at it as a human in the loop. I'm going to look at it, is this right?
166. 1:17:47share
     Jason Akatiff Is that-- and what i like about this rather than a drag and drop page builder is like now i can just tell ai hey swap this logo make the make the brand this hey take this uh take this image and build a design system out of it and layer it on this so it looks different right and there's no manual processes and you know in my world i run I think I maxed out at 22 sessions at a time yesterday. I think it was yesterday, the day before yesterday. I had 15 demos and phone calls, and I had 22 cloud code sessions running at once, right? So for me, it's like, oh, you need a page? Cool. Let me stand up a session.
167. 1:18:31share
     Jason Akatiff Let me get it started. And then I just leave it, and it just goes and does its thing."Oh, you need a new feature added to the system?"Cool, let me open a session for that." And now I've got a new co-founder, his name's Vish. He's doing all the internals. He's moving all of our internals to Rust and he's running a bunch of sessions. Now there's both of us running sessions.
168. 1:18:57share
     Jason Akatiff So where you might've had... know 30 developers in the past now you've got you know two guys running 10 or 15 sessions each you effectively have 30 to 50 developers working at the same time through two people makes sense yeah it's not but everybody's good everybody's gonna be here like whether we like it or not whether you like it or not it's because because the market is going to demand it right like what a company can produce what the margins should look like what all these things are it's going to get normalized to ai ai driven employees or ai empowered employees it's going to normalize to that over the next five to ten years
169. 1:19:45share
     John Casto there's such a gap right now but there's a huge opportunity right there's a huge alpha if you're new to you know into these tools now so let's take two more questions so i'm gonna give uh i'll let jason pick but let's give it 60 seconds everyone that has questions put questions in once the question once the 60 seconds is up jason pick two and go for it and while that's happening uh some not questions but just random things that people put in the q a keith said tell jason he's the goat and then bill somerville i'm outing him full last name are you single i love you bill and i love you keith let's give it a few more seconds

170. Share the working context with the team

171. 1:20:34share
     Jason Akatiff I saw a question somebody asked there. They said, do I give my AI agents autonomy 100% of the time? I've always worked that way. This, again, goes back to the original thing. I gave them full autonomy early on, and I understood where the edges cases were and what they did and where they would break things, right? So it's really, you know, just get in the Ferrari. don't get in that you go right get in the ferrari and learn to drive the ferrari right and do it in things that don't matter is is how you want to go about that process so i would always run dangerously skip permissions if you're running claude i'd run yolo if it's codex i'd run
172. 1:21:24share
     John Casto Okay, let's see. I don't see any more questions really coming in, so grab another one. There's a few in there.
173. 1:21:33share
     Jason Akatiff Hold on, I can open the chat here. TechStack, also, do you centralize skills and frameworks and data using tools like Orca? Oh, that's a great question. So, you know, in the beginning, no, right? my my whole thing is like how do i go as fast as humanly possible and it always has been right like how do i drive the ferrari like the best person in the world and you know i didn't have any other people building this product with me and so i didn't need any of that uh now that we brought in other team members you know before i used to run a global uh set of Global set of skills, skills, rules, CloudMD, AgentsMD, whatever.
174. 1:22:23share
     Jason Akatiff Now we've moved to a project level so that when you pull down the repo out of GitHub, you pull down all the skills with it. And then we, on the authentication side, originally I ran.emdlocals, which is basically a text file. in the project to have access to all the keys. Now we've moved to 1PasswordInjects. So 1Password holds everything. AI never gets access to the keys. And then you could just add users or developers to that 1Password. They'll pull down an encrypted.env, a local, actually, that gets decrypted on...
175. 1:23:03share
     Jason Akatiff when it builds or whatever it does i don't exactly know how it works um but yeah we're at one password and then we share inside of uh we share inside the repo for the project now uh some more advanced stuff on that side uh there's a guy named alex he owns tld crm brilliant Really, really smart guy. You know, he's doing some really cool stuff with within his company, basically giving everybody in the business has access to harnesses and agents and stuff like that. That's really separate from. So he's basically made his own repos from my understanding. I'm no expert at this. I just kind of heard us. He's explaining.
176. 1:23:44share
     Jason Akatiff to be made a company-wide repo that anybody in the business has access to that has their own agents and own harnesses uh to do stuff that's really separate from their code base
177. 1:23:55share
     John Casto um a repository for those of you that don't know and github is basically a folder of files that's all it is sounds fancy but it's really just a folder of files jason it's been awesome man i'd love to talk for another hour and a half but i i gotta meet with a an actual human now so for uh for uh for some training i could talk about this stuff for unlimited amounts of time right like it's my it's my passion uh all of you can find me on bulletproof scaling
178. 1:24:29share
     Jason Akatiff You know, if you're into call tracking, lead tracking, you're tired of working on 10 different platforms and consolidating that all together, we've got one unified platform where you can run leads, calls, you can do lead nurturing through SMS, email, and agentic outbound dialing, as well as a new buyer onboarding automation system that's kind of a Shopify of of lead selling and call selling. So it's kind of got those four pillars within it. But yeah, you can find me on pretty much anywhere as@jasonakatif, Instagram or Telegram or whatever. Feel free to message.
179. 1:25:14share
     John Casto - I'll answer, he's a beast. He's definitely helped me. changed my life really in a lot of ways so thank you man i appreciate you appreciate your time we'll link we'll link back we'll probably do this again this one this one great we had 72 people at the at the peak which is pretty badass for the last minute two days before posting it's awesome