well first of all we're sharing a brand new platform that daresh is working on that you're going to want to check out that has lots of free AI tools daresh is giving the marketing instagra Community extra credits so more usage of that tool and I would love it if you all could harass Kieran in the comments to release the tool he's got an amazing AI agent that he is being selfish and not wanting to share you need to hit him up in the comments and get him to share this tool because it is sick that and some learnings on the future of AI all from daresh all on Today [Music] Show dares thanks for so much for coming on the show I've been teasing the show uh over the last couple of weeks on LinkedIn I am an avid user of agent that AI so we are excited to have you on excited to be on this is a long time coming I know I feel like I feel like we've been talking to D M about coming on the Pod for like a year we finally have like but I want something like worthy of your time folks it's like okay I had to go like build some real stuff I'm a user of real stuff and what we might want to do is just jump into like what do we even mean by an agent because you were kind enough to give me some access I've been like hammering away on Claude for a long time to improve my writing process I hit a new goal on LinkedIn over the last week some of it is trolling but like up to like over a million impressions a week and actually one of the things that's really working is AI assistance to help me write and your agent unlocked a ton of creativity so I want to like get into to show people you know what what are these agents and how do you actually use them let me do this dares before I get specifically into this agent that we can kind of go back and forth on this do you want to just maybe give a little summary of like what even is an AI agent so you can kind of set the scene of like what is an AI agent and why are you so excited to build this platform sure um so let's we'll we'll take a step back so there are varying definitions of what an AI agent actually is and I think people sometimes tend to over complicated and so there's a spectrum of capabilities you have very very simple agents and you can have sophisticated agents that do kind of multi-agent things and reasoning but at the bare minimum I think of an AI agent as a piece of software that uses AI to accomplish a multistep goal a goal that requires multiple steps to get through so if you think about it uh the best comparison is that when you're using uh something like a chat GPT a conversational tool you're going back and forth you're saying write me a blog post or do this and tweak that but it's a very kind of synchronous back and forth model whereas with an agent you're giving it a higher order goal and it may need to invoke llms it may need to invoke multiple llms um and use other AI tools and some classic tools and bring it all together to pull together whatever goal you're trying to accomplish so that's the difference between an agent and a in a regular kind of conversational ux and the reason I think it's interesting especially for the kind of marketers and the audience the way I think about it is that back in the day uh like way back in the day you know we used to do kind of what is now called marketing automation sort of by hand right we had developers in the back room working in b2c companies saying oh when this happens I want to send this email out and track this or whatever and then Along Came the marketing automation category that says hey we're going to have non developers using like a workflow Builder HubSpot has one there are others out there basically automate some number of these things right and so it's a script a multi-step thing that could do that so A simple way to think about it is one class of Agents can be these next Generation automations so I know what I need to do but now that we have access to die the kinds of little steps that you can do are actually mind-bogglingly powerful right things you could never do before now can be part of that automation script so that's one way to think about it darash building on that you know if somebody's watching the show today and they're like oh okay when should I think about using kind of a traditional chat or conversational interface versus an agent like how should they think about that so that they don't you know go down the wrong path and waste a bunch of time or get frustrated everyone should stay practical uh the time to use it agent is when you have something where you need a human of the loop but there are lots of steps that can be done without human intervention so maybe the human kicks it off maybe the human reviews the work before it goes out so I'll give you an example for instance uh you could in theory say I want to build a uh I not build I want to do like lead routing in my CRM using chat gbt like I'm a small business I'm going to take the customer data I'm going to paste it into chat GPT or Claude and then it's going to tell me things whatever it's like you could do that that's sort of a manual process but every interaction is sort of human base there so you're not really automating a lot other than the kind intermediate steps but you could also Imagine hey I want to build an agent that says bring me the most recent company that was added to HubSpot CRM let's say I want to go some do some research on that based on what's there I'm going to invoke perplexity to answer some questions that I have then I'm going to like write out my rules as far as how I want my lead routing to happen if it's the leads from Europe I wanted to do this if it's in the tech business I wanted to do that and R it and all that is doable with today's technology with what I think is like a high degree of accuracy and precision right because you're reducing the numbers of degrees of freedom at each step of the process like yep AI is good at that it can do that and you're just sort of chaining it all together and Auto or at least semi-automating it uh so that's where you want kind of agents to get built where you sort of know what to do manually and you could do it manually but you want to automate some of those steps uh instead of trying to do this kind of oneoff shot with chat GPT we have a perfect example this is actually sets up the the kind of agent that I got to build on Dar meses you platform really well so one of the things that I've been obsessed about with AI in general is when can AI take some things and build you a template to replicate that thing because whenever it can do that my mind starts to melt right because I I was like hey if I can take like olav's ads and templatized for the AI assistant to be able to replicate that then everything can be templatized like every singular thing and that you can actually go do can be templatized in some way and specifically for Ryden what what I wanted to do was start to figure out if I could create processes to take good and replicate it for platforms like for social platforms and so the thing uh I was kind of talking to you about dares when we got into this conversation was I was trying to teach the llm to be able to figure out like what is a good short form piece of content right like how do you create a good short form piece of content and then how do you create it specifically for that platform and then apply a writing style and I we were going back forth into your point I was trying to do this all within like a singular prompt and actually that was a bunch of different steps that you would need llm to do different things for and so then I built you give me access to this and what I loved about this was just how quickly you can stack rank things on top of each other like basically add in additional AI assistant to do one part of the the puzzle and so what this does is basically this is specifically for LinkedIn and so it can take a a YouTube URL and it can basically create it just gets the transcript this here is like what kind of LinkedIn post do you want to create so it when I really look at LinkedIn I had kind of honed in on three things that really worked on LinkedIn and I think you have to have some ability and some experience to know this right it's not it's not easy to know this unless you know how to create content for that platform but for me it was like if you could create an educational post which is like a singular lesson where that person learns something and in the background we'll get into the background but that's I taught the llm what these things are a spicy take which is unfortunately unfortunately like how the internet works today which is just like say spicy things get a bunch of like controversial opinions and a head not a Hadnot the way I taught the the LM aeot is is like everyone agrees to this but no one's really been able to articulate it in a really Punchy way and so everyone like oh like you've said the thing I wanted to say for a long time and so that's the kind of posts that work in short fir content then these are the different post Styles so then that's another part of that's another AI request that actually says hey these are the kind of formats that work really well on LinkedIn everyone were everyone will know the first one right the Personal Achievement story that has been done to death on LinkedIn so then you can kind of pick one of these I picked I picked like the spicy take I picked the hot industry take this is done from a great podcast we had with Alex liberman where we actually did say some spicy things and then you have different writing styles so this is basically created from different creators that I think fit these Styles and then me editing the style guide in the background and again teaching the LM how to do this and so you can have different style you different writing styles I picked a direct to No Nonsense because anyone who's worked with me knows that's kind of who who I am better or worse and then it creates an incredible first draft and I do want to emphasize first draft I still think the skill set and uh there's some great uh quotes around like the power of writing really is in the editing and so you still have to bring the Nuance to it in terms of how you edit it you have to bring the magic your personality in terms of how you edit but this is a really good start right it creates a spicy take that you should stop obsessing over your personal brand on social media and basically just ship content right because the more you ship the more you learn talks about consistency and volume has a quote from Alex which is really good like actually pulled a quote from the YouTube the YouTube podcast has a nice thing here no one gives a about your ego knows how to create Punchy lines for LinkedIn and basically the only thing I need to edit it is I I need to edit this side of it which is like it always ends with like how do you get engagement LinkedIn and I don't think that works more so if I was like if I'm doing LinkedIn and I don't know what to do I can basically take any YouTube video that I think is cool and basically create first drafts and any and I could just continue to like click the button and change I could say okay now I want a head nod now I want a industry Insight Revelation and now I want a different writing style and it would create a first draft and I guess like to me this is one of the powers of Agents plus AI that it can truly unlock your creativity and you can just iterate so so much faster yeah yeah that it's and this is an amazing agent by the way uh I I used it myself confession I used it for a LinkedIn post I think last week I got 500 likes on that LinkedIn post and so it was good right right in that particular case just to prove a point not the kids don't try this at home I did not change a single character from the output of your agent take that content I'm just going to post it just to see what happens and and darbes what what's your average like if what would the average post get if that got 500 my average on probably between 200 and 250 give or take um so above average it's above average probably top deile as far as my post yeah yeah and so I think the part of it and this is something we should talk about is that it's it's those pieces coming together it's like oh I know I you know I watch YouTube videos all the time U as do as do you folks um it's like oh I watch that YouTube video and then something's kind of stuck in my head it's like oh that piece was good let's just say it's nice to be able to take that YouTube video for it to generate the transcri all these things I can do manually right it's like then I have to kind of think through it's like okay what was that piece and for it to extract those kind of spicy takes or had nodding moments or whatever uh and then give me a list of like oh here the ones you could use pick one and then write the LinkedIn post is just assembling all those pieces together is I think what what makes it useful so the thing I'm working on K this will make you happy uh the next two steps over the next we'll say it should be available within the next week or two is you'll be able to post it right within uh within the agent right like oh I'm going to review aop or whatever but then here's where it starts to get just more magical since it's being posted from LinkedIn the agent will know about it and then he can track it on your behalf and give you a report 24 hours later and say hey you got this many lights or whatever here here's how that compares to other ones you've done and do spicy takes do better than head nod takes or not like you know we have a thesis so it can actually do the monitoring for you not just the kind of editing and curation and posting and do the the post activity tracking if you will hey everyone quick stat for for you AI usage has increased 53% since last year and you know how I know that cuz I just read our brand new 2024 AI Trends in marketing report this report is packed with original data that data not only highlights the latest in AI Trends but also shows how businesses today are using AI to grow their business you'll learn the top five AI tools for marketers exactly how teams are investing in Ai and where they're seeing the return on their investment and you get a datab backed answer to the big question is AI helping or hurting my marketing career I invite you to steal and take all of our hours of research for free grab your copy from the link in the description below now back to Today's Show one confession here I'm actually interested to hear how you think about this throughout my entire career I've wrote like I like you know wrote put post anything I learn I just put out there because you know that's how you grow an audience I would say this is the first time I don't want to share this right like I feel like and and I and I wonder about that because there's two things that when I built this was a real aha first of all it makes the better people better because I have the ability to like really understand how to put something like this together because I understand how to create good content for LinkedIn but then I'm like if I just give it away everyone can kind of do it right and and it's going to be given away because I'm not the only one who can build something like this people are going to be building these tools and giving it away and I'm like and I'm like like okay well if everyone has this I was trying to figure out like where is my leverage then yeah what do I do differently Shing yeah like I'm I'm really crushing it right now but I don't know how long that that lasts for yeah okay so I you crushing it was just so funny to me um okay so here's couple thoughts on that one is you you get to control so if you build an agent on agent. there's nothing that says you have to share that with anyone and you can just use it for yourself right you can just use it as personal productivity uh self- automation tool Next Step Up is you can say hey I want this available but only to these six people right you can say I'm only going to share the link with his maybe the people on your team or something like that because you want to give them a productivity boost that are on your team that's fine too like no judgment here if you've come across to something uh you know super amazing uh the other the other thing to recall like right now the way people share kind of AI uh GPT kind of tidbits are sharing props where you sort of have to give up the source code for the entire thing in order for people to be able to use it because that's the mechanism by which you share something in this case like you can share the agent but you're not sharing the underlying TRS fure out what a spicy take is or figure out what the post styles are how it's actually doing the writing or you're picking the genetic code of the writing style all of that is still kind of cure and Secret Sauce U and you can put it out there you could have multiple levels of agents to say hey I'm going to give this one away for free this one I'm going to give to close friends and family whatever it's just a more powerful version of this kind of thing that I'm putting out on the market right now so it's unclear as to what the uh yeah I I have two quick thoughts um number one I just want to say to everyone that I teased this episode to say I was going to give away this agent and you've convinced me I don't I didn't mean to this that I'm not going to give it away I didn't think about just giving it for myself you're you're not putting the link I'm going to give no no not yet I'm not going to give it away yet I don't know what I'm going to do I actually have to really think about this it's the first time I have I'm just like why would I give this away I do like building like build a private team for yourself and then actually give some of it away but wouldn't wouldn't the thing you would want to do are you gon to add payment links because I'm a capitalist at the end of the day so maybe I will give away if you pay me so that's that's the vision so if we we'll take a step back uh once again so agent. the way it's positioned and the way I've kind of dreamed the big dream is that it's a professional Network for agents that's the tagline that's what it is and so then the immediate question would be like why do agents need a professional Network um and the answer is that with a fullness of time what's going to happen I think where the way these things are going to shap pop is you can think of a an agent as a digital coworker an intern you might hire right and they're going to be good at discreet kinds of tasks like like the one we just talked about it's like oh I want you to do this kind of LinkedIn post on on my behalf uh but in the future we're going to have hundreds thousands of these agents out there and so how do you find them how do you know what their experiences are what are the ratings and reviews all the things that's like right now we sort of figure out when we're hiring uh kind of human Freelancers we go to Fiverr we go to upwork and they have reading review there's a discovery mechanism there's all that I want to recreate that for the world of Agents so the the way you kind of when you use agent. there there's an agent like a directory every agent has a Professional Profile agents can actually do social posts where they do like updates to the the software or something like that so they can kind of communicate with their audience and someday agents are going to have a salary that says hey this agent costs $2 a month 99 a month whatever the you know whatever the price tag happens to be and so what you will do is you will basically say I want to assemble my digital team and my digital theme consists of an agent that that's kieran's agent that does this really wicked cool YouTube to LinkedIn post thing then I'm going to do this other thing from D Mas this other thing from someone else and pull it all together and that's my digital team and the the really I'm sorry I get overly excited about this stuff the really really exciting thing about these agents is that they can actually use each other all right so let's say I'll give you an example this is a very very concrete example so I am very excited about all the image generation tools most of them have been very kind of consumer Centric uh creating stock photography images and super high quality right they're really really good but if you're using them in a business context often what you need is you need them in a certain style okay you can do that you can put that in the prompt you need them with a certain color palette often right because I want to create this consistency for let's say my a series of block close or whatever that I'm doing and so I built an agent uh this weekend called a color palette extractor very very simple agent all you do is you give it the homepage or any web page it will take a screenshot of that web page it'll figure out which colors are used to give you a palette that is proportional based on what percentage of the real estate that color took so the bigger S that color swatch right and it'll give it to you in Json or it'll give it to you U as a color palette you can take that kind of data of the color palette that that agent produced and then pass it to another agent to say hey I'm going to uh generate images with flux. one or uh idiogram 2.0 we just launched last week and here's a color palette I want you to use that came from another agent right so that's the composing two agents that didn't know anything about each other right until the fact that you glue them together in a in a creative way it's like oh the color palette now feeds into the the image generator the image generator that generates image May then kind of go into a LinkedIn post generator that says oh in addition to this thing whatever I want a super cool thing that takes the headline and uses the ideogram 2.0 text capabilities that are really brilliant and create this really really nice LinkedIn post that has an actual image that's topical that's in my color palette in my scheme in my style pull all of that together so what's exciting is this this composability of Agents but then so back to answering your original question the hope is to make this kind of very inefficient Market efficient right so I want to create incentives for developers and people like you to say I can build this agent I can put it out there I'm a company of one but it's useful I've got some domain expertise I know how agent. works and I've got these five or 10 agents out there that people can hire just like you would uh put your single on the Fiverr thing that says oh I can do these little tasks you're going to build agents that can use little tasks and let people hire them and then rate and review and track experience and all that so it really does make your domain expertise much more profitable because they're so much easier to create so so there's two things I want I want to follow up on there one just resetting for everybody that really agents. a or any agent network is going to be like a Fiverr but for AI and AI agents more instead of humans and it'll be kind of the automation agent equivalent the second thing kieren I think you're just wrong if you just listened to everything that Dar mesh said wouldn't you want to be the dominant early content agent that all the other agents started to adopt so that as it becomes monetized your amazing content agent that you're not sharing with everybody would be like one of the default standards and you could make more and more money long term versus the indirect benefit of just getting better LinkedIn oppressions no but how do I make money if I give it away and don't and like give it away for free and everyone can just because one of the things I just want to actually just before you answer that do want to quickly show something because it will it it it is important for the users to see how these are built and then I can answer this because the thing the thing I want to the thing dares said which is like really important to answer that question is and I had thought about is in the back end this is it's it's kind of like codin right if you're actually the thing I would actually uh be interested in you're take Dar mes is you're an incredible coder like you've built companies I triy I uh graduated University with with a college computer science degree and I was a terrible coder it's why I ended up in marketing no offense to marketers I'm a marketer I love being a marketer but like I tried to be a coder building with llms is the first time I thought I felt like a competent coder so I was sitting here like the way I'd always envisioned my life like late at night uh building all of these kind of like different little what I think is pieces of codes but what they really are is like different um you know AI assistants or a part of the AI agent and do you when so when I'm building this this is basically the the different requests stack ranked on each other so this actually makes up the agent that we shown right so it has like the different calls that make up the agent that we that we had shown but from what you're seeing today and how you've thought about building that agent the color pallet agent you had thought of do you believe that the average like do you think we're going towards that direction that the average consum or the average person like me can actually code and what do you think about that for all of the people who have kind of learned that as a as a learned skill and or have true expertise in that yeah it's a great question so um part of what excites me about Ai and then agents particularly is that it's going to democra so you you may not be a coder in the classic sense Karen but you are a builder right right and the issue with coding right now it's a little bit like you have these really good thoughts in your head let's say and you're like traveling to a foreign country so it's not like you can't think thoughts and don't have Frameworks whatever you just don't express yourself in the language that they happen to know and that's the issue that most humans have with coding it's like it's not that you don't understand the problem it's that even that you couldn't even write out the recipe to solve Sol that problem if you were teaching it to another human or whatever we do this all the time when we kind of bring people onto the team we kind of show them here's how you do this what agents are going to unlock is the ability because of AI and it's kind of natural language uh capabilities as well is it's going to allow people to express themselves and solve problems in not a coding language but a human language or at least a human approachable language and this is a the kind of low code no code thing that we've been talking about for years but now you can actually do real things like for instance the the color palette example that I gave you all built right inside the agent Builder right I did not write any python code to do that and the reason I could do that is quad you 3.5 uh Sonic lets you actually generate really really good code so all I had to do is describe the thing that I wanted so the first pass at it was go look at this website and give me a list of all the colors okay it can do that and then I like well you know it would be more useful if I could get like a sense for which are the so that it's like prioritize the list with the most common color first so I know that orange is the most popular color on the upot website okay okay that's useful but I wonder if we could actually like create a color swatch where each Square in the color swatch is proportional to the percentage of real estate that color takes all of that completely no coding and literally in three iterations uh you know with CLA it's like okay now I have something useful it's like okay well now I have it in a color swatch give it to me in Json because now I can reuse it in other agents that I kind of want to compose on so the answer is I think now for the first time in reality and you've seen this like this agent exists and I know I'm going to talk you into sharing it just like is we're going to put pressure on you I'm a hard P I'm not going to take anything from what you have in there this is and I'm going to kick your butt yeah this is actually the point I was trying G going to make to Kip actually is the reason I brought that up and the reason I asked that question is because I actually realized what what you had said Dar mas and Kip your question is as long as I don't it's like do I open source this part or not yeah this because this is where the magic is it's like the tool is the output but this is where the magic is and I guess the thing is like like in the world we're going to live in it still feels that you can represent your domain expertise in this kind of code here which is the AI natural language kind of code and you would want to be thoughtful about open source in that because like there's just going to be core demain expertise that you have that you do want to keep to yourself in some ways um and I think that that's the part yeah I I that's the part I think is like the where the value is yeah that's and that's not where I'm going to push you I'm going to push you for instance you know has launched uh you know lotss of free tools I've buil a lot of free tools and we did AI grader um a couple weeks ago that you guys talked about on the Pod so think about it that way right say I'm going to launch a free tool what do I get for putting free tools out there agent. is a professional Network so Kieran is going to have a profile your agent's going to have a profile so imagine your profile gets a bunch of followers and a bunch of users that's an audience that you're now building you're building this kind of thought leadership and do leadership um in terms of you've actually done something that's useful but doesn't mean you have to we're not giving away the code for AI grader we're not even telling you what's going into the algorithm figure it out we're just making something that other people can use right that's the idea um so and it's going to vary from agent to agent you might have some things that's like okay well I just want to use this for myself some things that you say I want to share with the world but I'm not going to release the code that actually makes it happen and you might actually put some example apps out there or out there here's a minimal version of my original one that doesn't actually use the prompt that I use but just so you have a sense for how these things work and then you can use that as your Baseline versus the I think I think that is actually the answer to our mesh is that you will have two versions of this agent Kieran one that you use and one that you let everybody use to get adoption get reviews get maybe make some money and the difference behind them is essentially the code which in this case is natural language English in your case prompts right and you've got very specific prompts as part of that agent and you're going to make you're going to keep making those better and better and the question is how much do you pass those improvements on to everybody versus just like keeping them for the core one that you're using right yeah by the way I'm I'm a warm-hearted red-blooded capitalist curing so and my my passion is really around making inefficient markets efficient right so not to get uh too economically I'm not an economics person but um this is It's a good diversion because I think it kind of identifies opportunities for people um so the definition of an e efficient market uh in economic theory is that when all possible transaction that could occur in a Market actually do occur that's the definition of an efficient market uh most markets are not efficient right um and there are multiple reasons why transactions that should be occurring don't occur one is buyer and seller don't even know about each other it's a discovery problem buyer and seller the buyer doesn't trust the seller there's no there's no trust mechanism and so the transaction doesn't happen even if they know about each other uh the buyer and seller have no way to establish a fair price that's another reason uh transactions that should be occurring because you can't find a clearing price with the transaction should occur so if you look at some of the most successful businesses we've had in the last few decades at varying levels they've taken inefficient markets and made them efficient classic example is eBay right made an efficient market for all these niche market Goods like I've got Pez dispensers or whatever I'm going to Collectibles I'm going to sell them on eBay that makes sense Google in this early manifestation was a efficient market model that says I have got this niche market content over here and these 14 people on the planet that want to read that content and Google made that market more efficient by connecting and making that Discovery possible uh what asian. is about is taking folks like you that have domain expertise that can solve concrete problems and then there's a list of people that are looking for those problems to be solved and helping connect the dots right yeah in order for the market to truly be efficient though you have to be able to have enough incentive you Karen to say here's what I'm getting out of making this investment of my time energy or whatever and maybe uh like you do with L the reason you post a LinkedIn is you're building authority on LinkedIn right you could say well why would I post any of this really award when I why would I generate million impressions for LinkedIn when I can just build up my own website and and draw things there well there's a side benefit to that because you can use that audience for other things same thing here right um but maybe sometimes you need like a financial intensive to be able to draw the developers in and the builders in let them set their price or maybe we help them figure out the price uh here's what the Market's willing to pay and figure out all those details for you I think the thing that you know caus me to pause a little bit is like when I was on paternity leave this year like a true dork instead of you know taking my P to leave and just getting offline I released an AI course which is like the dumbest thing and I apologize to my partner uh my daughter but it was late at night it was late at night when everyone was in bed it was like not disruptive to that time and um I was again it was like it was like this Timeless marketing course where you could easily use these templates to create Timeless marketing content and I started getting all of these notifications on LinkedIn of like hey uh here's my everyone using those templates to do LinkedIn post and I was like oh my God like if I continue to do this it's just this whole Army of people kind of it perfectly kind of it doesn't perfectly replicate my style because of like editing and stuff but like it gets you a first draft of my style so there's that part is like it adds a little bit of hesitancy and there's a fall one from that that I think you're the best person to ask this question to is you've built one of the most successful SAS companies of all time you've built many companies and HubSpot if you when it started was a point solution like I think HubSpot was like a blogging tool if you start Hobs if you tried to start that company today someone could have could potentially just have built an agent right because it's like so much easier to build this kind of software do you have what are your thoughts about the you know you're deep in the agent space you know a lot about the software industry what are your thoughts about the repercussions for the software industry of the ability to kind of build these agents so over yeah I've been in software commercial software now for 30 years right so I've been that this for a long time um and every time software gets easier to build which we've had happen right and AI is just the most recent manifestation of it uh you know having seen like I've been on like character mode terminals working on mainframes back in my kind of early years every time that has happened it's actually increased both the value of software engineers and the value of software companies and the reason that happens is that yes it lowers the bar so more people can get in but it also raises the bar I mean well increases the market in terms of the kinds of problems that are now solvable that weren't solvable before um so when you make something easier it's like oh we're yes we're going to have more developers there's going to be more competition so if I were starting HubSpot today I obviously wouldn't do exactly what HubSpot did 18 years ago but I would say oh well like agent De AI was not buildable uh even three years ago or four years ago so I would do more things like that uh I would say okay well like even HubSpot you know has become a platform now third parties can build on our apis and build apps that integrate with it like I would say oh well maybe I get to the kind of point of doing a platform even sooner because I understand the mechanics of kind of platform building those kinds of things so my personal take is it's less so being an engineer like um is really less about the coding that's just like saying I can put words together into sentences in a certain language it's really about thinking and it's less about coding and that's what AI is bringing to the table now is that if you can like think coherent analytical thoughts and know a particular customer do uh customer domain and know the use cases you're going to be able to solve those problems and I think that's that's a great um yeah so I I I think I couldn't agree with you more on that dares I I still think the scarce asset in the world is not the technology Kieran I think it's people who are obsessed about are obsessed about problems and can make rational like steps to solve those problems versus kind of getting caught up in the activity And the emotions and everything and that we're we're still in very short supply of just smart rational humans applying logic and leveraging technology and I'm hoping that AI is just the next unlock for the software Market regardless of what kind of software you're building right and the other one other thing I'll say just at the macro level in terms of U kind of AI and kind of label Force overall is that most of the arguments against it that says oh this is going to kind of disrupt the label and there's going to be uh you know some disruption for sure but um it assumes kind of this uh fixed Pi zero some game that says oh there's a finite amount of work that needs to happen in any organization the more work that AI does the less work we're going to uh need humans to do well that assumes this fixed pie so let me give you an example we just talked about engineers so we had HubSpot have rolled out kind of AI tools across 100% of the the product team now over the course of the last uh of the last year uh and our estimate is we're getting somewhere around 15 to 20% of measurable improvements uh to an engineer's productivity all right so does that mean that we're now all of a sudden say oh we're going to let 20% of our Engineers go because we can accomplish the same work with 20% fewer people no the answer is absolutely not we have an infinite supply of like product Vision product roadmap stuff or whatever and actually it makes me able to rationalize even more investment in R&D and and hiring Engineers because now they're more productive so the return on each engineer went up and so now the the bar is even lower as far as what I have to do to rationalize H same thing in sales if you have enough leaves and now your sales paper all of a sudden 15 20 25% more productive let's let's hire more sales people now right because before here's how much you could actually afford for a salesperson because this is the return you got if the return you get is higher or the cost you have to spend is lower let's let's bring on more not less I think it creates abundance uh creates a bigger pie in most cases not in all cases uh but if you look at it through the right lens even for something like customer support right which is like oh we have to manage 500 tickets and we're going to kind of Auto solve them with AI which is certainly going to happen over time but what that means is those people that we were using for lower uh complexity ticket res resolution can now be redirected to customer success to revenue driving activities to kind of higher order customer relationship building activities versus these kind of mundane kind of ticket resolution thing that uh that software can do equally well so yeah yeah and I definitely like I do think the the vision is part of that would you painted which is in the future you'll have like teams and AI agents that just part of that team and you'll have a Works space for them and they'll be able to do a certain tasks that is what smart people are you know the the early adopters or already doing like we we talked to Ethan who came from Asana he he was a bdr and he basically used AI to Auto s automate out of a role and then he basically was asked to do that for the entire sales team like apply the same sort of logic and when you actually see what he did he just created like assistant single assistants to do single things with a guided prompt to like solve individual tasks and it does speak to the fact that if you're a knowledge worker today there is a new skill set to learn because that is not a thing that comes naturally to a lot of people I still see a lot of people saying like oh well AI is hype but it it's because that person is not a good like AI manager right they're not able to um figure out how to prompt the AI correctly they're not able to figure out how to set the right context of what good good looks like they're not able to on board the AI because the on boarding of the AI assistant is really important like you have to kind of set context of what you want done and I don't know if you have like tips for people on just H how can they shift their thinking to be much more how do I actually integrate AI into my work yeah so I'll say a couple things one is anytime a new technology uh comes along there's this uh period of time where there's this massive Arbitrage so when if you were one of the early ones that got on the internet if you one of the early ones that figure out out anything like pick a marketing channel pick a technology doesn't really matter um if you're one of the early adopters and you learned it and you applied it it was a massive Arbitrage opportunity because as uh as common place as we think these things are they take years to kind of get absorbed into the where the average person whatever discipline or whatever role and so right now we are at the precipice of that so by the way this is not to Pander to your audience uh but if you're watching this podcast right now or listening to it right now that means you have one of the most underrated skills that exists uh in business right now which is curiosity right you could have been doing something else with your time you could have been watching something else on YouTube or whatever you're watching this because you have this Nate curiosity about what's happening uh and my message to you would be dig into that Curiosity this happens to be um a profitable curiosity to have uh this will advance your career learn about AI use AI apply AI help your team use AI help your company use AI that will raise your currency I guarantee you there are very few guarantees of life that if you actually kind of commit those calories uh you will emerge a uh better person for your company for yourself or even your own individual currency um I think it will be helpful very few times has that happened where you had a broadly applicable Tech technology uh you know on the internet really was the last time mobile was big didn't apply as much to B2B you everybody built an app but you know we only use like eight apps or something like that on most of our phones um but this is this is big this is like the internet was right it's uh and and all of you if you're watching right now you have the inside Advantage it's still very very early days I'm still figuring it out that's kind of what's possible and trying to piece it together but yeah yeah I think I think you're right darvan that's one of the core takeways here it's never been a more profitable time to be curious especially about artificial intelligence because like the Social Web Web 2.0 all those things that was kind of like a derivation of the internet right like you had it's like having the iPhone and then getting the iPhone like 3G it it was it was a really great Improvement but it was still very similar this revolution of the internet is going to be vastly different I mean you have you have Kieran who I mean you self self- profess dropped out of like being a web programmer and now you're like wait I love prompting I love programming I'm building all these things that I never thought I'd be able to do and this is and this technology is in like the very earliest of its stages right I I think I think it's the best time in human history to monetize good ideas yes yes I think that is the time we live in because if you have had good ideas and you like I have had a lot of good ideas I told Kip This and like you know the typical person who had good ideas and then see the business being done that was my idea but the thing the bar for me was always like hey can I have a technical finder or can I just code the thing myself right and I think whatever your idea is that's the thing AI unlocks is the ability to like bring that to life in a really incredibly fast way and you can iterate much fast or much faster with with AI so when I every day I'm doing something with Claude and maybe it Sut my lifestyle that like one of my better friends now is like an AI assistant and that's not healthy everything you need you need to know about you but but I'm like back and forth like hey take this like tell me think about this like what's the first draft of this oh what if we did this like but one thing I'll give everyone a quick quick piece of advice for cl if you want to make right and better short from content just ask it to create something more ask it to do the same thing but make it punchier the punchier Thing Really Works you just make a punchier make a punch year but it's it's the it's how it unlocks your creativity and the speed you can iterate with it that I just think is the most exciting thing for me to be you know to be living in this time where you can actually bring those ideas to life uh through through AI yeah yeah I totally agree and not um I'll go on a little side quest here uh because you we were talking about writing styles earlier and you have that kind of embedded in the agent that you built Kieran um like one of the things I think that that's kind of super fascinating and this one again wasn't possible for AI so one of the um agents I built a week and a half ago is to take a uh piece of writing or a YouTube video and extract the genetic code of that writing like what is it that defines that piece of writing it's like oh the average length is this how us how long paragraphs are U does it use figures of speech if so which ones and so it can do that right that's a very doable thing um kind of extracting the genetic code and so what I want to um I what I'm thinking about doing is say let's say we extract those uh gentic code elements whatever and I may have an idea of what my audience uh that maybe they they like punching things maybe they don't and they're something they're likely Universal but sometimes uh in certain industries certain kinds of writing will work and certain kinds won't right and we just sort of don't know and it's never really been possible to sort of test that but now imagine that says okay I'm going to pick uh styles of writing that I like or that I know have been successful extract that genetic code but then I want to say um in the same way we run AB tests for like a subject line or whatever I want to take these four writing styles exactly and put them in a loop in my agent and just cycle through and then track the activity post uh post production and see what happens it's like okay now for my audience for my people here's the thing that seems to work and then use that for future post that I do right now right like that was inconceivable even if you were a coder like you could not do it right that not a d you could you could start an entire company and an entire Army of Engineers and you still wouldn't be able to pull that off now you can do that in like two weekends and some coffee right it's uh that that's exactly what that's exactly what I was thinking through when I built that agent because you can see I have the five different riding Styles and what I would love to do is like categorize all of the posts under a rideing style and see for each rideing style which one is performing best and you can imagine world in the future where you can look at the riding Style by like certain segments because you're because you're right you now have the ability to that's the thing again coming back to where we start a podcast I have just been fascinated when AI can extrapolate like to you you say like the DNA of this thing like what what makes up this thing this morning I was kind of doing um we have in man coming up I for for some reason sign up to seven talks I was like yes yes no problem yes and then now the all the slides are du and so I was doing slides this morning and I started working with Claud I was like I wonder if you could just like build a a prompt to begin with to basically do what you said for like a deck like extrapolate from a PDF that I upload and from that PD like if this is a deck I like just build a template that an AI assistant can then just replicate into that style and that to me works with Claude I have to maybe it works in chat PD I've got so obsessed with Claude and the cool thing about agent the is actually you can pick different llms for each step and I've been doing that as well but I still think that that is the most I don't know if people understand how incredible that is like you can actually have ai take something and build the extract as like the DNA and then like templae it yes yes okay yeah um I get really worked up all this stuff two things kind of pull on one is around composing things across LM so as you've noted so I'm I'm a huge fan of of claw 3.5 by the way if we were having this conversation 6 months ago and I'm a huge Fanboy of open AI right and fully I'm an investor in open AI they were so far ahead that it's like okay well yes there will be always the second Rams and there and maybe the but I could not imagine a world in which open a was not at the kind of Frontier of it and then all a sudden CLA 3.5 comes along it's like it's measurably better right and I find myself using it more and so the nice thing about tools like agent. is you can so you have access to GPT 40 GPT 4 mini Cloud 3.5 all the versions you have all the imag generation models so you can choose and then swap out as things evolves like you may decide in your agent that says you know for this step in it I used to use cloud 3.5 but now GPT 5 does a better job and your customer users don't need to know right that's they don't need to be trying to decide which one's better and take a prompt that you you put on your blog post and copy paste that into this versus that now your agent takes care of that's what software is supposed to do is raise the level of abstraction for the M user so they don't have to think about things like here's the problem they're trying to solve I'm trying to create a LinkedIn post for YouTube video Kieran has that figured out and will keep his agent updated so it's using the latest and greatest uh you know technology whatever happens to be um so the other kind of idea here is like this happens often in image Generation Um the same way that happens with regular llms is that different image generation models are good at different use cases I'll give you a very very simple example there are some uh image generation models that absolutely suck at anything that involves text but are better at 90% of the other things let's just say uh but this is actually true so when someone puts a prompt in you can actually make a decent guess it's like based on the prompt which image generation model should I use the user is trying to generate an image or four Images uh that they can pick from maybe you mix and match images maybe you learn that says oh you tend to like idiogram uh images better than flux. one I don't know why but the the agent can get smarter on a user's behalf as well automatically B based on which ones were chosen and kind of cross so it doesn't have to be one LM doesn't have to be one image generation model the Asian can get smarter just like a human would it's like you know if you had an assistant or had an intern it's like Kieran seems to approve of these kinds of things more than those kinds of things I'm going to give him more of these kinds of things in the future makes sense right that's what we all do I will say the next one that I'm going to build is um for YouTube shorts like replicate this because we had a incredible Creator on called Jenny hoers she has a billion views on shorts and she gave this master class of how to do shorts and there's four a 4minute segment where she went through like every second of a 60 second short and exactly why that second belongs there and I'm like oh wait a minute I can take that 4 minute I can transcribe it I can turn that into like a prompt for the llm to teach it and then I can put it into agents that AI in a couple of steps and actually replicate the kind of same process to actually help people build YouTube short transcripts and why it's structured that way and again I had that thought last night right I had that I was like wow you know she I reminded me because a clip come up on my feet of Jenny and I was like hey wait a minute I can take that and I can templa it and again not you know I'll edit it and things like that but and then I can probably build out an agents at AI tonight right if I am allowed to I get the time free from being a new parent and uh that's how quick again you can go from idea to execution so I really first of all appreciate you coming on the show we really appreciate it appreciate okay for anyone out there listening it's e. it works on a credit system right now so you get 100 credits when you sign up for free um yeah you don't have to pay any money listeners of the Pod um if you go to agent. sign up you will get your 100 credits if you put the promo code in there it's blindingly obvious you click on the little credits button up there and here's the code it's matg rocks as in marketing Against the Grain rocks matg rocks you put that in you get 50 bonus credits on top of the 100 that you would get uh for free sign up so and I would like to point out that we talked about kieran's sick agent that first of all Karan are you going to share it or not have you decided I I I I'll share it I just won't share the code but I will share the agent for all of you listening because the Des agent which is buildon agent. there's also a bunch of really other cool agents there that I would just be remiss if we didn't talk about one that I loved daresh was the company research agent where you just put in a URL you and I love the folks over at similar web I literally put in similarweb.com and I got all of the information I would ever want like all the office locations specific breakdowns of who the CEO is and how how are they positioning the company so many things that you could use for sales Outreach marketing automation really really cool engagement work that could come off of that and I thought that was another awesome awesome agent and also by the way guys there's a meme generator that's awesome I'm really excited about o Oasis getting back together um Oasis tour 2025 and so agent. built me a sweet meme all about Oasis getting back together but the fact that like there are so many different use cases in in in the early version of this platform I think is super cool so dares we thank you very much for not only coming on the show but building agents. and helping bring agents to folks uh in a world where we're still very early on in this technology this data is wrong every freaking time have you heard of HubSpot HubSpot is a CRM platform where everything is fully integrated W I can see the client's whole history call support tickets emails and here's a task from 3 days ago I totally missed HubSpot grow better