[00:00:09] Hey everyone, it's Bob. Welcome back to Elevate Your AIQ, your go-to source for insightful conversations on human-centric AI readiness, talent transformation, responsible innovation, and the future of work. In this episode, I sit down with David Arnoux, who is currently the co-founder of the AI Native Venture Studio Humanoids. He's the community leader of the Gen AI Circle and go-to-market strategist at HeyArnoux.
[00:00:33] David formerly co-founded a startup called Growth Tribe that you might be familiar with and in general has been a technology innovator for a long time. David and I explore what it really means to move beyond augmentation and into the agentic era of work. We discuss the four-stage landscape of how AI is reshaping work from co-intelligence and augmentation through full workflow automation and into the uncomfortable reality that some categories of jobs simply won't survive the next few years.
[00:01:00] We also get into responsible AI guardrails for real-world agent deployment, the emerging shift from SaaS to custom-built tools, and how the heaviest AI adopters are running entire businesses with agents doing most of the work in the background. If you want an honest practitioner-level look at where AI and work are actually heading, you won't want to miss this entire episode. Thank you so much for listening. Let's go talk to David. David Arnoux
[00:01:26] Hey, everyone. Welcome to another episode of Elevate Your AIQ. I am your host, Bob Pulver. Today, I have the pleasure of speaking to Mr. David Arnoux. David Arnoux How are you today, David? I'm super. I'm excited about our chat. David Arnoux Likewise. Likewise. Thanks for doing this. I saw some of the stuff that you were building when I attended one of Hong Li's Recruiting Brain Food Weekly calls. You were building some stuff in the go-to-market space. I think we were playing around with some LinkedIn tools that you've built.
[00:01:51] And so I was just fascinated by the tour of some of the projects that you were working on. So I was anxious to get you into a deeper conversation about some of the things you're building and the philosophy and some of your career arc and some of the multiple things that you're doing right now. So I thought you could just start with just you giving a bit of background about what got you to this point and some of the projects that you're working on. David Arnoux
[00:02:15] One of the first interesting businesses I started, I think, that relates to this story was a growth tribe. It was about 10 years ago. We built it in 2015 or 16 because basically nobody in all Europe specifically, we were very targeted at Europe, was teaching marketeers how to actually do data and growth properly. David Arnoux That was a thing 10 years ago. And we ended up training about 35,000 people across 15, 16 countries before I rolled off. David Arnoux
[00:02:41] But what's kind of crazy is that I'd say the bet we made back then was right, but we massively underestimated the timeline. So we thought marketeers needed to become more technical, growth mindset, experiment based, learn SQL to a certain extent, maybe dabble in Python, learn experimentation. David Arnoux David Arnoux David Arnoux David Arnoux David Arnoux David Arnoux
[00:03:12] David Arnoux David Arnoux David Arnoux David Arnoux David Arnoux David Arnoux David Arnoux David Arnoux David Arnoux David Arnoux hundreds and today the tooling is agents not dashboards and now I run three things side by side
[00:03:41] something called humanoids with my co-founder Walid that's our AI native venture studio out of which we push a bunch of SaaS and we also have partnerships with corporate clients who we build for and share on the revenue and on the on the equity we build SaaS products with agents from day one so AI native and then there is Gen AI Circle which is the community that I run and co-run actually with the members we're like 470 now I think it's 470 of the heaviest Gen AI adopters in
[00:04:10] across the globe at the moment and it's our signal in this sort of way in this cascade of noise I would say and even the signal is noisy nowadays because it goes so fast and then there's just what I call hey are new which is the my fractional practice and that's where I work with maybe four or five at the same time b2b or b2c clients at a time on gtm go to market which is fancy way of saying sales and
[00:04:36] marketing and AI strategy and the honest version I would say of why I run all three it's like a flywheel so they feed into each other the studio breaks things and I learn from it the community shows me what the actual frontier is catch up what's hype what's not hype what people are doing and the fractional clients they it's where I get to play around with somebody else's CRM and strategy and work with
[00:05:03] stakeholders in like real world deployment data without all three I think I'd be a board or flying flying blind yeah no that's amazing and yeah I'm sure it's a lot to juggle all three I'm sure you've found ways to sort of cross pollinate you know ideas and concepts you know across them so would you just in terms of how they may interact I mean if someone comes up with an idea you know within the
[00:05:29] the gen AI circle you know community you could technically just feed that into you know humanoids or find another way to help them you know bring that idea to to fruition to commercialization or whatever the goal is yeah so I actually so no in the sense that the I'd say the biggest blocker at the moment for growing a business is actually audience or access to some type of proprietary data
[00:05:57] and or having some regulatory captor capturing some weird industry that no one's ever heard of so ideas are a dimes a dozen now I don't think it's the revenge of the idea gal or guy I think it's the revenge of the person well it's not even revenge distributions everything nowadays and audiences more and more important than ever so what we usually do is we try to partner with people or
[00:06:22] individual organizations that either have a massive CRM or some sort of data that nobody else has access to or know about some weird boring process that no one's ever heard of but that is a massive market like boring businesses yeah that's where we kind of focus at the moment and the idea of the three sort of this trifecta of these three is that like the fractional consulting it's like b2b clients gtm
[00:06:48] strategy ai integration that's where you see the deep client problems show up so from time to time you see a deep client problem that could potentially be a venture but usually these clients they just they just want me to fix that issue and they're not into venture building so it's usually like people paying me you know five figures a month to help them not get killed by ai if I put it in really weird terms and help do stakeholder management with security with cio with compliance all of that
[00:07:17] stuff the gen ai circle is the community arm I really believe audience and community are remote nowadays because we don't know what's true or what's false online anymore dead internet theory la at least there there's a similitude of real humans even though sometimes you interact with what their agents are telling them say but we keep it real we keep it really humans we have sort of a manifesto around that and what's great there is we do one brain trust session a month members share work
[00:07:45] playbooks we publish over 100 of them so far we're going to try to reach 200 and funnily enough nowadays playbooks it's not so much reading the playbook and implementing it a lot of the time is you download the scale or the md file or we always have a copy for llm button and then you sheet it to your llm and ask your llm or your whatever your setup is is this where it's me implementing how can I adapt it so yeah so it's kind of compounding loop client surface is a hard
[00:08:11] problem I bring the question to the community five members have already solved it we extract the playbook and potentially the studio builds a product around it if it's if it's big enough and I use that product back in client work and round and round it goes until we're all out of a job jokey but uh we're gonna get into that for sure but yeah no it's fascinating I mean I'm sort of a
[00:08:36] guy with no shortage of of ideas but I definitely appreciate the fact that you know going into businesses and understanding where there's some you know nascent sort of need and challenge that needs to be solved is is the ideal you know starting point right and so so I certainly recognize that there's a lot of opportunity for folks like you to lead those types of engagements and to show what's
[00:09:00] possible and then you know the community piece I think it's fantastic because I'm a sort of a collective intelligence advocate I have been for at least two decades and collective human intelligence I should say so but getting the those people and that cognitive diversity together in in a group is great and then you can always complement that or augment that the collective intelligence group with some custom you know AI which you know I'm guessing you've probably explored as well but yeah really
[00:09:30] really fascinating I did find it some of the things that you've created I was mentioning before that you've got a site kind of a fun site called don't build this dot com and I was going to take a couple of my post-it note ideas and just throw them in there just for fun it hurts for those those things you should definitely check it out it's basically put in your idea and it'll basically like roast your
[00:09:55] idea and poke holes in it and so I think it'd be a good exercise for people to think you have all these great ideas and you know maybe uh maybe David has something something to say about that the just on these like sort of human you know themes that happen and as we think about the the human and AI relationship I thought we could unpack that a little bit in terms of how people think about
[00:10:20] I think Professor Malek would call it co-intelligence you know humans plus AI as I was alluding to before you know collective human intelligence plus artificial intelligence but then you have this concept of augmentation that says you know humans are great at a lot of things and we don't want to lose the humanity and the human centricity but AI can perhaps make them better have them think more deeply about their ideas their decisions you know all of these other ways
[00:10:48] that AI can help and then you have automation where we're like taking things off of the plate of of a human to hopefully free them up to do other things but I just thought we could you know talk about how you think about those things as you talk to our clients as you engage with the community in the gen AI circle and and how you think about some of the the other projects that you won't work on yeah I have pretty strong opinions on this and I I don't want to create anxiety but I do want to create a
[00:11:17] little bit of fear and a little bit of hope at the same time and I also deep down I think we're all going to be fine like with any s-curve like with any new tech but I do have a really really strong opinion about this because I'm on the bleeding edge and I'm also I also interact with older industries so I also see the laggards a little bit and I also truly believe that there's beautiful industries out there like building bridges or actual physical bridges or really really deep tech that are so
[00:11:45] necessary or care work nurses teachers so it's it's kind of a meta discussion I would say so I'll just come at it with kind of a strong opinion and then people can nuance it for themselves and we could put a little bit of water in the wine so I think most people today are stuck in co-intelligence land or augmentation land and it's comfortable and they think AI is like a faster typing assistant
[00:12:11] still I would say that's that was the 2023 frame and I think it's already obsolete like completely obsolete so the way I see the four words co-intelligence is what Malik popularized AI as a thinking partner the book is amazingly well written I follow Malik on the on LinkedIn religiously anything he posts AI as a thinking partner real useful but it's still you in the driver's seat all right
[00:12:38] augmentation it's what I see most enterprises sold themselves on chat GPT enterprise for like 500 seats you got workshops on prompts engineering that was all the rage in 24 25 still happening today six months later adoptions like 12% and the CFO is like where did the investment go then you've got this wave of automation so that's kind of where the real money lives right now whole workflows running without
[00:13:04] a human in the middle little bit of human in the loop lead scoring email triage content production you've got weekly reporting customer onboarding things like that and token costs blowing up and then CFO asking why are we spending so much on tokens can't you optimize token use and by the way the era of the subsidized frontier models is about to end we're starting to see they're throttling their models and we
[00:13:29] might start paying a lot more for these tokens having said that most of the work that we do can be done on cheaper models or open source models and you only need the frontier models for a few use cases now the one we don't talk about is redundancy so that's the part nobody wants to say out loud the whole categories of work that just stop existing and my read is that a lot of mid-tier knowledge work which actually
[00:13:54] you know is those are the formatting jobs they're like the data moving jobs their report writing jobs they tend to be the entry-level jobs those don't survive the next 18 months and there's been a lot of freezing hires on those jobs now some people are saying that the frontier models are becoming more expensive than entry-level humans and entry-level humans equipped with smart llms are actually cheaper and more efficient than then relying solely on the frontier models but i would say that a lot of the
[00:14:23] work we're asking the models to do don't require the frontier models and i think we need to be honest about the fact that pretending it's all just about augmentation is how you and i end up unprepared so the work that survives i believe is taste judgment relationships and creating new things that don't exist so good taste cultural curiosity domain expertise knowing what's good what's not good
[00:14:50] what has value what doesn't have value judgment building up real relationships on most of the projects where i work you free up time for people who are in gtm function sales and marketing to build relationships with with with clients with customers the the analogy is usually the account executives who takes a skateboard to go see the client rather than filling filling out salesforce or for recruiters it's building up those relationships rather than spending half your day filling out an ats
[00:15:19] whatever it is humans were not meant to fill out an ats right boring nobody wants to do it nobody got into this job to do it however i think there's a real list of jobs and functions that won't exist anymore and we need to be realistic about it and just like any s curve any technological revolution we need to jump on board with this one and the other thing that's kind of how that we see happening is there's actually going to be two internets there's going to be one internet that's human to human
[00:15:45] or you and i are still making the purchase decision on a piece of software we want to buy or on a candidate that we want to hire and more and more we're seeing the a to a internet the agent to agent stripe released something where you could actually give stripe access for it to be able to make purchasing decisions for you so we're going to see an a to a internet which is agents buying from agents and decision making from agent to agent directly and we already do it when we buy software except it's
[00:16:11] going to it's going to happen more and more so just pretending it's all about augmentation is how we end up unprepared you've got five you've got time still but it's important to uh to start writing the wave and be interested in the subject and understand it could be as simple as you know what's one task or process or workflow that i can augment myself or automate this week or next week it
[00:16:37] could be as simple as that but i do think we need to be realistic about how powerful this you know promethean technology promethean style technology is it's the fire for the brain basically yeah wow no i think that's a really nice sort of landscape view of all the ways in which we have got to think about this sort of human ai relationship and i guess i feel like some of it where we're stuck
[00:17:03] now i don't know if stuck is the right term but as we think about how to do it right you know i spent a lot of time talking about human centricity focused on responsible not just responsible use but responsible design and so i feel like because everyone's kind of scrambling we don't have a lot of you know legislation at least here in the u.s and the different rules more stringent rules in in
[00:17:30] europe and in other places around the world i feel like people are still sort of running around disoriented not sure what their individual or collective you know place is in you know that department within an organization or the organization as a whole and i feel like people have there's a lot of subjectivity in terms of how you think about human centricity what do i what do we want to allow
[00:18:00] humans to to continue to do should they have the freedom and the agency to do the things that they want to do it was just a new report from this i think it was the center for humane technology and it laid out all these principles about you know what we shouldn't allow ai to do and i was a little bit i mean i i'm definitely a human centricity advocate but some of it gets a little i would say too like fluffy
[00:18:27] in the sense like listen there's no just because someone like maybe somebody likes doing you know filing and shuffling data around and copying and pasting because it's less taxing on the brain and it's just it's maybe how they it's a relaxing you know sort of you know not a highly cognitive you know task and they enjoy doing it do we we take it away from them i mean people are different
[00:18:55] thresholds about what they want to continue doing and where they get job satisfaction from but i don't know to me that's a little naive to think that you know the organization is just going to let you continue doing it at you know whatever the additional cost is to to the organization i mean you've got to you got to recognize that we need to advance a lot of the things that we do to make
[00:19:22] you know to make sort of progress and so if you want to do those those things then maybe go find a job at a company that's more of a laggard than someone who's trying to you know fuel you know sort of rocket ship kind of growth within an organization i just think people are kind of disoriented about where to trust ai to take take things over hi i'm jeremy ames your host of signals a starbold studios
[00:19:51] production i'm taking bold ideas and real challenges submitted by leaders like you and turning them into powerful conversation with actionable results some weeks i'll react in just four minutes others i'll bring on experts for deeper dives so send up a signal and i'll throw you a line and so yeah i guess i'm just curious to get your reaction to that yeah so i mean it's almost like asking a truck driver to use a paper map nowadays rather than google map right just because they enjoy it so it's it it
[00:20:20] whereas where to trust it that's kind of a okay so we're talking about accountability and responsible ai and of course that's a cursor it's a cultural cursor it's an individual cursor so it's difficult to sort of give a a blanket statement about this i can give you my actual rules just sort of how i approach it so anything that touches the outside world goes through a confirmation step at least when we're building it out so sending bms posting publicly paying anything signing anything the agent
[00:20:49] drafts a human reviews and then the agent executes after approval so we've learned this the hard way a few times after the script that dm'd people that i had already manually messaged for example except when you do a little bit of hardening you can then just set up the right guardrails and a sample check or a check from time to time so typically when you start off it's anything that touches the outside world goes through confirmation step and then once you start trusting it and then there's continuous quality
[00:21:17] reviews two is audit logs so for every meaningful action time action source target if an agent does something stupid you have to be able to trace it so we use markdown files in a memory folder every action logged boring but kind of critical and then three is of course pre-flight checks before an agent does outreach cross-reference against the data store before it makes a claim it cites a source la-di-da-di-da
[00:21:44] and then the fourth one is like a daily lock file so high frequency tasks like outreach they get a daily cap and a lock file so the agent literally cannot run again until tomorrow and then five is the agent never executes destructive actions without explicit human input so you don't it doesn't kill everything even with all those guardrails you're never totally safe as you're always safer but you're never totally
[00:22:07] safe and then deletions force pushes money movement sending mass anything there should always be like human in the loop to be able to confirm however once you start scaling you realize that you can't always be the human in the loop if you're we have one venture that's in the six figures now and we're running it with one person basically so the human in the loop would just be validating stuff all the time so at some point you need to kind of trust your systems and you can never be fully safe you can be
[00:22:34] safer and our cursor our north star is a little bit is this delivering value to the end user essentially if it is then then we keep pushing it if it isn't and again it's a cursor right it's not black or white it's nuanced not a dichotomy so i think the bigger frame for me is that responsible ai is not a philosophy debate it can actually be a checklist and it's the same checklist you'd be building for
[00:23:01] managing a team of like 22 year olds fast capable occasionally reckless need structure i don't know if i should use age maybe we should use personality types because there's a lot of reckless 55 42 year olds out there as well but treat your agents to treat your agents the same way i do think that are you okay for me to go on on this because i have like a mental model i'd love to share on what heavy
[00:23:28] opt heavy adopters are doing within our group or within my network and you can see how far people are taking this at the moment but maybe you want to just rebound yeah yeah no i think that would be great to go through that and then you may while you do that you may be answering my what my next question was going to be but so yeah please what's the next one so i can maybe like dive into it at the same time well i was gonna ask you a question about just some some of the personal things that i'm doing
[00:23:55] and how it relates to your your the philosophy that you just outlined that i think you were gonna elaborate on but i did want to ask about you know how your philosophy has been sort of embedded into the your humanoids you know visit the businesses that you support yeah you know through through that so i think one thing i always like to look at what people are doing i think it's that mark andrews in quotes look at what developers are doing on the weekend and invest in that i like to look at what people are
[00:24:25] doing in the gen ai circle the heavy doctors are doing in the gen ai circle and then use that as my sort of north star cursor compass for what's coming next and how i should think about the world so inside inside our gen ai circle we have like 400 470 of like the heaviest adopters that i know founders operators agency owners technical marketeers not everybody's super mature right on the technical
[00:24:52] side a lot of them are immature but they're playing around with it tinkers and there's a whole group of what we call ai adoption and implementation that their job is to translate what the technology can do and bring it into their organization and some of the patterns i see and this is how we should think about this at responsible ai piece is that they run ai headlessly so we're not talking chat windows and that's how i run my businesses i don't have chat windows i don't even have an interface anymore i run
[00:25:21] my businesses from the terminal and i'm not a developer from the terminal from the command line interface so if you see my screen today it's it's a wide screen and i've got a bunch of terminals so it's like a bunch of command line interfaces and that's where i chat with my skills with my agents with my data sources even with my email with my google calendar we had conversations via my email i never opened the email interface it was always through that command line interface there's data
[00:25:50] coming in from the different products that i run i don't look at the dashboards from the different tools that we have whether it's post hog or mix panel or google analytics i pull the data and i and i interact with the command line interface and i create my own dashboards i have cron jobs scheduled background agents so these frontier frontiers people their work happens most of it is over for the
[00:26:16] ones who are further than what i'm doing their work happens overnight one member built and and she was like a executive at a pretty big boring and boring enterprise heavily compliant firm she built a morning briefing pipeline so by the time she opens her laptop the agent has already pulled emails summarize priorities drafted replies and chewed her day took her about nine months to get this validated in a
[00:26:42] rink fan in a sandbox environment with in an azure environment because she's a she's she's on the microsoft stack the two is i they build their own skills a scale is a markdown file which is just a text file with instructions no code so heavy adopters they productize every workflow that they do twice most users just chat heavy adopters they skill it we you build a skill so if you see yourself doing
[00:27:08] something three four or five times you turn that into a skill that your system cannot use which is a fancy way of saying a repeatable set of instructions and then three i would say is that they have this sort of second brain architecture i've actually started commercializing my second brain 22 years of how i approach work and go to market and you build it as what's called you have a memory folder plus integrations
[00:27:34] into the different tools that you use so the ai knows your business it has in my case it has access to my slack my notion meeting transcripts not from client work because i'm not allowed to do that but from like our internal ventures to the crm when i start a task the ai briefs itself it already has all the context i don't need to brief it on anything which is a cyber security nightmare by the way so there's different ways to approach that we could talk about the lethal trifecta isolation isolation isolation
[00:28:03] how you approach that but that's kind of the frontier there's three other things that these people do they orchestrate sub-agents so not one ai but a swarm with an orchestrator that delegates and parallelizes so one member runs what he calls wiggum loops overnight he spawns 20 parallel agents on a problem checks results in the morning sees if it's good enough to push out that check result is where
[00:28:29] the human in the loop and the domain expert is and the taste is and the evaluation but that's it the rest is sort of spawned by agents yeah i think that's it and two other things that are important is these frontiers people they've ditched sass replaced 80 of their stack with custom-built micro tools so of course they need a bit of maintenance they need to learn how to maintain it etc etc and they eat their own dog food they use things they're building on themselves first way before clients and then
[00:28:57] last thing is that now we're voice first so walking talking dictating into agents now that you've got tools like whisper flow and some people they actually do everything from their logitech mouse where basically one of the buttons is to talk to your whisper flow so you're dictating to your computer and the other button is yeah sure to your cli it's like yeah sure go ahead yeah sure go ahead and then so i mean we have real numbers like there's one guy what one guy in the in in the in the community
[00:29:24] he's uh he's a doctor running a clinic 3 000 facebook followers in january 40 000 in april 2 million plus view post clinic booked solid through january he doesn't have a contact team he's agents and for me that blows my mind on what's possible nowadays so that's where we're going and i know we say we're in a bubble in this group etc etc but we also were saying two years ago that agents hallucinated and
[00:29:51] we would never be agentified and we can never have you know a reliable picture of will smith built by the by the image agents and we just get used now to the fact that we've got photo realistic images at at the touch of our fingers so this is definitely where we're going and your moral compass your responsible ai campus needs to be built based off of those facts it's happening it's being pushed now
[00:30:15] there's sort of ethical issues there's social issues that need to be addressed there's also environmental issues the only the only thing i can say about those is that with every new technological s-curve for some reason we forgot about those issues because we are as humans we have otf optimal it's called optimal foraging theory we will choose the path of least resistance if it makes our lives easier and we
[00:30:43] will choose convenience over almost everything else whether it's security the environment etc etc so it's our responsibility for people who are on the cutting edge of this to push the conversation in in that direction while also you know riding the wave at the same time because from a capitalistic point of you for better for worse it makes sense whether it's on the job markets or on the on whatever
[00:31:08] market you're on at the moment sorry there was a lot of things at the same time but uh i hope it makes a bit of sense it does it makes a lot of sense and i think it was important for people to get the context and sort of the landscape of all the things that need to be considered because i think to your last point just in terms of people finding the path of least resistance you know i spend a lot of time with earlier adopters and and not just the current workforce but the future workforce right i talk to
[00:31:38] you know students not just college students but you know k through 12 you know primary school as you might call it but you know are we teaching the right habits are we teaching people you know to learn these skills but also make sure they're learning the the material they're not outsourcing their critical thinking they're not just finding shortcuts right whether it's you know doing or having ai do
[00:32:03] their homework or write their essay or whatever it is and so i think it is human nature and that's why it's important to start this i guess it's in some ways a new form of of digital literacy that you've been doing for for quite a while now i think a couple of the other things i've wanted to mention is like when i i mean i personally get stuck sometimes thinking about when i'm gonna
[00:32:31] where and when i'm going to use ai in everything that i do whether that's with my advisory practice and the outreach that i do the research that i do if i don't have their contact information i don't want i don't think i should be you know scraping it from places where i'm not authorized to scrape it from if it's making those connections like you were talking about with your second brain system well
[00:32:57] i've been working with with claude to try to you know put that together but i want to pull in information from a bunch of communities that i'm a member of but i'm not the owner or the administrator of those communities so i can't just go in and scrape data from you know a circle not your genai circle community but like the circle community platform school school with a k is another
[00:33:21] community platform mighty networks is another community platform i mean i'm a member of far too many communities it sounds like i should be a member of yours instead and get rid of all the others but but uh you know i have all these sort of i've i've fragmented my my own sort of ai you know upskilling and engagement kind of universe right and and too much so and now i'm trying to bring it all back
[00:33:49] together so that i understand the nature of my relationships and and to keep tabs on all these things and so i mean it's it's basically you know social media you know you know expanded to all these other new places and i think community is really important to your original point around you know you're not in this alone there's other people who are building things maybe you can find new collaborators new new
[00:34:14] partners new new clients whatever it is but but yeah how do you how do you bring it all together so i'm having issues i get into these internal you know deliberations with myself around what is appropriate for me to do in terms of outreach whether that's finding you know podcast guests like you or it's a you know potential advisory clients or you know potential marketing partners whatever it is
[00:34:39] and then even on the post-production side you know how much of the you know video editing or social posts do i want to basically take off my plate and how do i get to the point where it's producing things that are that are trustworthy but are also like i want to be transparent with what i do if ai is helping me do certain things in that context i want to be transparent about that and also make sure
[00:35:06] that if i'm writing on my substack for example that people are expecting that that writing those ideas those thoughts come from from my brain not you know written by ai show notes for a podcast is one thing but if something is literally trying to be you know me talking about you know what i'm doing what i'm thinking and how i operate you know that needs to be authentically me so i wound up in these you know
[00:35:33] battles with myself trying to figure out where is the line for me personally of where i use ai yeah it makes sense i think all those questions are more like moral ethical questions they're not really tech they don't really have anything to do with generative ai because you could you could do all those things theoretically before the gen ai wave and before the release of chat gpt in november 2022 you could already scrape you could already that's that was more like automation now it's just it's a
[00:36:01] little bit faster it's a little bit smarter and you could do multiple of these at the same time but those questions would um would have arisen 10 years ago five years ago and then with regards to writing i would say like again if i talk about the loom again the mechanical loom there were riots about the loom in 1830 in dion for example end of the day i hate saying end of the day the the point of the
[00:36:26] loom the designer still had the same sort of idea for what they wanted to produce whether it was a rub or whether it was a sheet so that the the idea for the design was there it's just you could produce a lot more of them and a lot faster so this technology actually gives us wings and removes the need to have always developers or automation experts in the way i'm talking about bits not not atoms so i think
[00:36:49] that that's more like legal ethical moral question gdpr considerations ethical considerations and that just really depends on you or on the company or on the country uh in on the country that you're in and typically you'll have some people who are more aggressive okay but let's go the ends justifies the means and then other people who are more like kind of on the fence as to whether
[00:37:13] this is okay yes or no uh but it's it's good to bring up these considerations i think you were talking about learning a little bit also and i think these are the things we need to teach because i think that when we were chatting before the call you were asking about like a decade of teaching digital skills got wrong what did it get wrong what do people actually need to learn so not just on the moral side but it's on the tech side and when you were talking about i'm a little bit lost where do
[00:37:42] i need to start i think it's kind of interesting because we used to teach things like python tableau sql basic statistics and then the actual core of what we were teaching was actually a mindset like a growth mindset rapid experimentation hypothesis test ship measure iterate you can use that internally externally for every business the technical skills are always the vehicles for the mindset whether it's the growth mindset or the ethical mindset or the legal mindset and that
[00:38:09] mindset tends to compound across every single tooling shift so the people who came through i used to have that's what we call it's carol dweck's mindset because that's exactly the loop you need on ai what's different right now isn't what people need to learn it's how learning happens so it used to be like
[00:38:36] read a book take a 12 week course practice for six months linear slow kind of expensive now learning is a markdown file. That's crazy to me. Real example, somebody in our community, they publish a skill for headless outbound. So it's a markdown, because you were talking about outreach. It's a markdown file with instructions, maybe like 100 lines. You copy it, you drop it in your second brain. 10 minutes later, you have their workflow running on your data. You didn't take a course,
[00:39:06] you didn't watch a tutorial, you downloaded their thinking directly into the system. Now, some people, they're going to read the markdown. What does this thing actually do? Other people, they're actually just going to push it and see what the outcome is. Other people, they're going to take it, understand what's in there and make it theirs. So it's really that sort of, it depends on who you are as a person. But it's a fundamental shift that's happening right now on how skills propagate. It used to take months for someone to figure this out. So now everybody
[00:39:33] has access to it in hours. One of our community members, even me, like I just released my GTMAI second brain. I'm basically giving people everything that I know in the form of markdown files, and they could just plug it to their data and go to town with it. That's what's being productized at the moment. And so the architecture at the moment, it's either codex or cloud code or cloud at the base, a memory layer, just a bunch of skills files. And then
[00:40:01] the brain holds everything it needs to know about your business in that memory file, your ICP, your offer, your pipeline, your client history, the different API integrations you want to have, MCP servers, connecting it all to your real data sources, Slack, Gmail, Drive, Notion, your CRM, blah, blah, blah. It's pretty crazy when you think about it. And so when you pull all of those data sources, you're stacking risk. And there's a concept called the lethal trifecta, originally from
[00:40:30] Simon Willison. And it was popularized in our community by Louisa Lima. And it's the idea that if a single agent has access to three things at once, private data, untrusted content, like web pages or emails, and the last one is the ability to communicate externally. So private data, untrusted content, like web pages or emails, and the ability to communicate externally, that's lethal for you. You have like a prompt injection waiting to happen. Someone can hide
[00:40:59] instructions in a web page today or in an email that your agents read, and suddenly it's like exfiltrating your CRM to an attacker. So it's not just a legal or ethical risk. It's also sort of a business risk. Missed revenue, late releases, customer return. Those aren't just major business issues, they're people issues. On Higher Power Radio, we dig past symptoms to the root leadership,
[00:41:25] hiring, and process problems that hinders company growth. Real founders, real diagnostics, and actionable takeaways that you can plug into your business today. Peer lessons to fuel your company's growth. Subscribe to the Higher Power Radio Show and start plugging your people issues in your business today. Higher, H-I-R-E, Power, P-O-W-E-R, Radio, R-A-D-I-O.com, or here on the Work Defined Network.
[00:41:50] And usually for that specific case, the architectural answer is isolation. So you split agents by capability. The agent that reads your inbox, that doesn't have permission to send messages, the agent that crawls the web, doesn't touch your CRM, the agent that posts publicly doesn't read your private memory. You compartmentalize. So that's kind of the moat. But from a more, what you're struggling with, your ethical demons, that's really a question of, is this the right thing
[00:42:18] to do yes or no? And I think that's very much a personal one. Yeah, no, that's totally fair. I think some of it is still technical. So my second brain is kind of a mess. You know, I've got it built in. Well, yeah, I was going to make that comment too, but I thought I'd just stick with not turning this into a psychology psychiatry session.
[00:42:46] The, so yeah, the second brain, I got it started. It's, you know, I've got a whole segment of my Notion database is there, but it's basically saying all your stuff is all over the place and where you need to, I don't know, I guess I just need to, you know, invest more time in that. Like I said before the show, like I'm trying to do six things at the same time. I can't build the chief of staff and do the outreach, you know, agents and do the second brain or whatever all simultaneously.
[00:43:15] I want to keep those things in context as I sort of build my little universe of agents and subagents. But, but yeah, I need to, I need to invest time because that can help me take some things, you know, off my plate that don't get into those moral and ethical, you know, questions. But I, but I love what you've laid out so much, even on the responsible AI side, because where I get concerned, especially in the, because I spent a lot of time in the talent technology space and every
[00:43:45] decision absolutely has to be made by a human, even if AI is in decision support role or many of those roles. But the concern has been, as you add more tools to your stack and you have more agents coming in. It's not just how that age particular agent behaved, it's that agent to agent, you know, communication and how data is passed and, you know, probably ties to some of those, what you were
[00:44:12] just describing the three sort of lethal, you know, pieces. And so, so it's just very encouraging that you've already, you know, sort of addressed that in a, in a very logical way. And so, so that's encouraging to me that we've got ways to do that and to have the right level of, of oversight and. Yeah. Right-ish, I would say. You never say you can be safer.
[00:44:41] What came to mind when you were talking about those, those three lethal things was like, isn't that, I mean, that's why I haven't touched anything like open claw because I feel like that's exactly what it's doing. And all of a sudden it's going to be saying things I didn't tell it to say, buying things I didn't tell it to buy, making assumptions, losing context and where I have somehow lost control. So I don't necessarily need to go down a full open claw, like rabbit hole, but
[00:45:09] I feel like some people, some heavy users are obviously leaning into that. And so maybe there's just something that I, that I've missed with, in terms of how to lock that down. The discussion we're having today will seem obsolete in three years. Just think about what you share with, through your credit card information. So MasterCard data on what you're buying that then gets used and sold by ad agencies to retard it. You predict whether you're at risk of cancer,
[00:45:38] you know, so we're already sharing so much. We're tracked continuously. We have ID cards, it's credit card information. And I think that these discussions we're having today will almost be like asking in three years would be like, if we were having a moral and ethical discussion about whether I should use Waze on my iPhone today, knowing where that information's being stored and
[00:46:05] tracked and what is, what some large big data companies are using that information to do. I think that's, I think that's what's going to happen. And that's why my suggestion is always just to get on board so that people who do have these ethical dilemmas and don't just go brute force, open claw across the board are actually just as powerful as the brute force people. Because usually you get into analysis paralysis or ethical paralysis. And by the time you've made the decision to actually use a technology, people who don't ask themselves these questions,
[00:46:35] they're like miles ahead. And that's why I think people today should get serious about a few things. If I could give a few tips about what they should look into is potentially cloud code is really cool. If you're technical curious, it's not a coding tool or just cloud in general. So cloud code, if you're technical curious, co-work, if you're not, this is like your operating system, not just another chat, just go look into it. Two is like what's called MCP integrations,
[00:47:02] which is a fancy way of saying API integrations for your cloud code or your co-work. Pick your top three of tools you're using, like your Slack, your Notion, your Drive, your DMAIL, whichever one holds your actual work, that's your minimal viable second brain. And then three, the important one is pick one workflow you do every day, write it as a, as a markdown file, run it once with, with AI, and then you save the markdown. That's a skill. You just productize part of your work.
[00:47:30] And then the framework I'd give people is the 10 levels of this. We use it inside the community and you're trying to get from level one to level two, to level three, to level four, and we can share it in the show notes later on. But I think that's, that's maybe good, like the first steps on how to, uh, how to approach this. And another thing that I think is important is to, to identify where the friction points in your day and in your week. So again,
[00:47:57] trying to identify where is it that I'm spending time and effort on things that should be done by an agent. Anyway, I didn't sign up for this, whether it's filling out an ATS or whether it's, you know, filling out a spreadsheet, generating a report. I think you dropped off for a few seconds, but I just, you just kept going on, on that piece. And so there's also something that I'm seeing at the moment is that I think SaaS is in trouble. I posted in our community last month that I would
[00:48:23] never purchase a CRM ever again. So if I was in hiring and starting from scratch, I'd never, I'd never buy an ATS ever again. Might be a bit of an exaggeration, but it got the most engagement of anything I've ever written in months because people felt it. The math on individual stacks is getting more and more interesting. Average professionals spend about 5,000 bucks a year on SaaS subscriptions, average professional. And we've had community member go, community members
[00:48:51] go from 12 tool stacks costing like 500 bucks a month to 40 bucks a month replacing in our world of GTM. It's like replacing clay, Jasper, Ahrefs, Gamma, Zapier, Superhuman, Otter, Gravely. And that's pretty crazy. And it's not just cheaper. It's a better fit, custom built for your exact workflow. So it's also like, if we keep automating ourselves into this world of agentic abundance, which companies are going to be left to actually pay for our services.
[00:49:19] So that's a little bit the question. And one important caveat is I'm not saying replace everything. It's just tools that facilitate human collaboration like Slack, Figma for teams, real-time stuff. You tend to keep those, but because AI is bad at coordination, good at production. But the bigger frame is, I think we've entered like the era of personalized tasks, personalized CRMs, personalized dashboards, personalized productivity tools. And I hope recruiters and recruitment, because I think that's a little bit of your audience, jumps on the bandwagon of this one
[00:49:49] as well. It's cheaper, better fit, more secure, no vendor lock-in. And the first step is really what I was talking about to identify what's a task that's a bit mundane, a bit annoying that you do on a daily or a weekly basis. And just get curious and talk with your LLN. How can I improve this? How can I augment this? And it'll kind of show you the way how to do it in your flavor and in your style. Yeah, I definitely, I'm glad you said that about CRM because I've been poking around at certain
[00:50:18] ones that I think could work for me. But then I'm like, wait, what would I be paying for that I can't do with some of the things that I'm learning? And don't I want to just continue doing what I'm doing and learning what I'm learning and using some of these tools? I do get concerned that, you know, by trying to save the, say it's, you know, 20 bucks a month for, you know, one thing,
[00:50:43] am I spending that in tokens, you know, trying to build something myself? So you've got to like really understand the complete, the total cost of, of ownership, just like if you were making any, you know, build versus buy decision, right? But, but I think to your point, yeah, as a solopreneur, like, do I really need a CRM, like an official CRM as a monthly subscription? Because, you know,
[00:51:10] there's only a couple key pieces of functionality that I need and if I can recreate that in my own way, then why would I need that? But for, for a bigger team, you know, for a medium to large size organization, I think you've got to think about, you may not want to replace like a system of record, you know, something that's, you know, secure, private, has a lot of confidential information,
[00:51:37] is orchestrating a lot of other things. So maybe you don't want to replace like your ERP system, for example, but for every more niche, you know, tool, you've got to sort of rethink what is the best approach to get the same, you know, functionality, the same value that I could, while also bringing the humans along to almost upskill them in the process. So now can they,
[00:52:03] do we have more people that could build things, you know, on our behalf? Do we have more ways and newer ways to think about as we redesign the organization itself and what we do, you know, that immediately should have you thinking about what, you know, could we actually use AI for this? And I think that helps you move up the sort of AI, you know, maturity scale to say, look, we're not just doing some of the things that you alluded to earlier, David, around, you know,
[00:52:32] sort of, you know, just automating, you know, routine things and, and, you know, getting tasks and certain activities off people's plates. But how do we really use AI to grow this organization, to find new opportunities, to overcome the challenges that have plagued the organization for a long time because you didn't have the bandwidth or, you know, the investment available to tackle some of those other things?
[00:52:59] Yeah. And just on the software piece, I mean, it's happening at the moment when it's time to renew, like that's the questions the CTOs, CFOs, CMOs are asking, like, why are we paying per seat for this tool? When we can whip something up and harden it with this small team that we built in one or two, and usually one or two devs involved. And I'm not saying everybody's vibe coding stuff into production, but more and more, the economics just don't make sense for paying multi-year contracts for enterprise solutions, where you can build something that's custom fit, that works better,
[00:53:29] and it's actually easier on the adoption side for your personnel, because we built it in our company style. So it's got the look, the feel, and it fits exactly their needs. So this buy versus build, for me, it's completely shifting towards build. And I think SaaS has been driving on this abstraction layer and this pricing. They got away with really high pricing and this seat-based pricing models, usually, that just don't make sense anymore, because now it's going to be token-based and
[00:53:58] yeah, I don't need the 200 seats, this revenue recognition or this revenue visualization tool anymore. The team can actually whip it up. And this is happening at the moment. These are discussions I'm in. Why are we paying for this? Let's stop paying for this. There's no reason to pay for this. So that premium is completely gone. And what SaaS needs to do to survive is to go MCP or CLI. And unfortunately, they're going to have to reduce their prices by like 70%. I think that's what's going to happen.
[00:54:26] Yeah. Well, that whole situation, I think, is really intriguing because I was just reading some stuff. Andreessen Horowitz was just writing about vendors like SAP and Workday and the pricing models. I mean, I've talked to smaller vendors in the talent HR space where they've already said, we're going to think about outcome-based pricing, looking at the value that we're
[00:54:52] generating. So a little bit different than token-based pricing, but that's certainly a factor as well. And then the other is that if you think you can just switch from humans to agents and not have them adjust their pricing model accordingly, then you might be in for a rude awakening. But if you think you can just get rid of five seats and replace it with five agents and they're not going to do anything about it, you are probably mistaken.
[00:55:19] I agree. I completely agree. I agree. And we could talk about it again and again, and it is a lot of prediction, but this is just... And the proof's in the pudding of the valuation to these businesses. And this is a reckoning that was going to happen anyway. And then there's the doomsday theories about the fact that you're automating your way out of jobs and therefore salaries and therefore budgets to actually buy the things that you want to be building.
[00:55:44] But maybe it was the reckoning that we needed to move a little bit more to atoms again and move a little bit more to circular economy, pay teachers better, pay under... Find people to actually be under underwater welders, start building actual bridges again, you know, just go back to the world of atoms a little bit. Maybe it's a reckoning that we needed, but I think tech's going to take
[00:56:08] a big, big hit. It's massive opportunity, but I think there will be less players in the space moving, moving forward. Could be wrong, but that's what I'm seeing on the, on the edges at least. Yeah, no, I think those are great observations. I also think to your point about people rethinking, you know, careers, career paths, do I even, why am I applying to jobs? I now feel confident that I
[00:56:34] could, I could build something myself. So more people following an entrepreneurial, you know, path or opportunities in a gig and freelance space where they can just work on multiple projects, that side hustle becomes a real hustle or multiple side hustles. And, you know, like yourself, you know, juggling, you know, a very entrepreneurial mindset with managing sort of three, you know, entities simultaneously. I think we're going to see more of that and a lot more value
[00:57:00] coming from that. But I also think people will, you know, lean into their own capabilities, their own human potential in the process. So I think that's a little bit of a, of a paradox with the advancement of all this technology as the ability for people to actually realize what they're capable of. Yeah. And then if people want to follow the entrepreneurial route, just one tip is build up your audience as soon as possible, your email list, you're following your audience because go to market
[00:57:26] distribution is the name of the game now. So if you build something fantastic, but have nobody to sell it to that, that's, that's an issue. Yeah. David, thank you so much for, for all your time. I mean, such great insight. Thank you. Really impressed with all the things that you're doing. We definitely have more to talk about. I'm confident, um, got some additional advice from me personally. So thank you again for spending so much time with me and, uh, my audience. I know there's a lot of great insight people to take from this. So thank you so much. Awesome. Thanks for having me.
[00:57:56] Absolutely. And thanks everyone for listening. We'll see you next time.
Sponsors
