The AI Maximalist Playbook: James Raybould on Building at the Speed of Ideas

Dennis Yao Yu: Welcome to AI for business leaders podcast. Welcome to AI for Business Leaders podcast. I am the founder and CEO of the Other Group, a strategic advisory firm that partners with commerce and AI technologies to scale go to market growth. Today I'm very excited to have James Raybould on the show. James was the former SVP and general manager of Turing Intelligence, the enterprise consulting and AI deployment division. And ah, a company that's valued at $2.2 billion that uh, works with frontier AI labs like OpenAI and helps enterprise transform AI from proof of concept into what they call proprietary intelligence. Before touring, James spent years at LinkedIn as an executive. He played a key role in two major acquisitions. The $1.6 billion LinkedIn.com deal that became LinkedIn Learning and the Glenn acquisition that brought employee engagement tools into the platform. James, welcome to the show.

James Raybould: Thank you, my friend. Excited to be here.

Dennis Yao Yu: Yeah, excited to have you. Just abundance of wisdom. So we're going to dig right into that brain.

James Raybould: We'll find out. We'll find out. We'll let the next 45 minutes, but I'll take the nice intro. I'm sure there's a lot of positive

Dennis Yao Yu: feedback that's going to come back from this, but, uh, maybe just kind of set the stage, give us a little background. Like you really had a really interesting career arc, right? And from bing consulting to 13 years of LinkedIn, different functions from like sales strategy to product management. Then you left for fractional exec work and eventually ended up at touring. So walk us through the thread that really connects those moves. Are there common problems that you're trying to solve?

James Raybould: I think I just like business. So I was one of those people. Some of you listening, watching, may have an mba. And so most people think of like reading business cases as like, oh, I have to read this business case. Like business school for me was the most fun series of my life because every day, every day I would get to, uh, you know, put myself in the shoes of the executive who's up late at night thinking about what to do about this or about that. So I just like business. If you look at my, if you want, I published all my ratings and reviews for all the books I've read. And if you look at my books from, I don't know, probably 2005 to 2020 or so, it's almost all nonfiction business. And so I just like business problems. I think it's just some people like gardening, some people like, you know, uh, the Raiders. I like business stuff. So I just, I mean, Bain Consulting is a Fun, broad way of getting exposed to a lot of stuff. And then I got lucky with LinkedIn. And so I do think one thing about careers, as you know, I'm sure as well as I do, is you just get lucky. And so I, uh, ended up joining LinkedIn because a guy from Bain had joined LinkedIn already, but he was debating between two companies. He was debating leaving LinkedIn. He was. I was leaving Bain to join LinkedIn or Quinn Street. M my guess, you haven't heard of Quinn Street. Maybe you have. By the way, it's a cool company. I'm not. But you probably haven't heard of Quinn Street. And so he was deliberating. He's like, oh, I'm going to join LinkedIn because I think the leader I'll be working for is going to help me grow. So he got lucky. So he got lucky by picking because again, he joined Quinn street. Who knows? But I got lucky because I picked him. Does that make sense? So it's like one of those things where you don't really, you can't foresee what's going to happen but you use your judgment like, well, he's got good judgment and so I'll just follow him. And then yes, to your point, the last. I was at LinkedIn for a long time and worked at various different versions of LinkedIn from pre IPO startup trying to figure out product market fit to we're growing crazy rates and getting acquired by Microsoft and all of us funds things. And so I think the through line is just, I don't know, curiosity. I like digging into things. I think, you know, more recently I like building things. I think the. I really, I mean, I'm sure we'll talk about AI shortly. Like I just. There's never been a more fun time than, you know, March 2026 to build things. And so I think anyone likes to build things, especially digital things. It's just, it's like you couldn't pick. It's like, you know, it's like the gold rush from like 1849 or you know, the whatever pick era you want to pick. That is that right now. And that's why it's, you know, it's, it's so much, so much fun.

Dennis Yao Yu: Yeah, I think that's also. We talked a little bit about this before this call. Right? Where just the excitement around this era, uh, uh, like you, like you mentioned, it's gold. It's like gold rush. If you're a true builder right now, it's really builder's paradise. There's Just so many different possibilities. You can go. But one quick sort of segue before we go into AI you talked about the books you read, right? In terms of business. And so what's one business that you remember, one business book that you remember that really is a good application to where we are right now? Just call it sort of AI era

James Raybould: M. It's a good question. The best AI book it's not. I mean there's a book.

Dennis Yao Yu: It doesn't need to be AI just the application because you use the word curiosity.

James Raybould: That's true. I. I think the. There's a book called Conscious Business by a guy named Fred Kaufman who was a. He worked at LinkedIn for a little bit, he worked at Google. He's basically an executive coach, but he writes all these kind of maxims and principles and theories. It's a little bit like I'm sure you know Ray Dalio and he's written his principles, et cetera. The Bridgewater founder. It's not the same, but it's the same of just like a first principles approach to all types of things and how you operate individually in a company. And I think the in some ways in this AI world where now it's all about how do you give context and how do you give instructions and how do you give the right goals. In some ways I want to kind of like mainline like great, here's the book now build all the agents that I need to take all these different ideas and all these principles. In fact, it hadn't occurred to me until this conversation, but maybe I will now. In fact, probably after this grab whatever I can like copyright appropriate uh, grab from his book. I think I have the Kindle version so I can like take some of my Kindle highlights, feed it into Claude code and then start saying, hey, this is a principle I believe in. I want to kind of put automate and ensure I'm keeping front and center and executing on. Let's build this into my kind of agent stack. Uh, once you have all the principles and the foundations, AI is just taking those ideas and taking those principles and automating them, amplifying them, whatever phrase you want to use. But it starts with like what is your kind of your grounding? And I think the to getting a little geeky, you know, sort of he has a lot of evals. Like if I use his book as like my sort of set of evals, that would be a great foundation. That's yeah. Conscious Business. Fred Kaufman K O F M A

Dennis Yao Yu: N I need to check it out

James Raybould: for sure, it's better dense. So again there's like oh hey, I read that book in an afternoon. This isn't one of those books but if you kind of invest the time, it's uh, it's, it's excellent.

Dennis Yao Yu: Is it worth for me to just summarize it with clock code?

James Raybould: No, I think, I think you got, I think it's, I think it's one of those books where. I think there's some books where you can get the kind of. The tldr. This is the one where I think it's actually worth doing the uh, the full. But it's fun now is actually, I mean my favorite thing now with reading is I read the book and then I spend, I don't know, 10 minutes, 30 minutes, an hour just chatting with. If I'm doing video, if I'm doing voice, I'll use chat cbt. If I'm doing everything else I'll use Claude and just like hey, just read blah blah blah. I want to get more information on this or hey, do people like this? Same thing with films. Like I'll often try and understand some stuff about a scene or some stuff about behind the scenes. I feel like the media is the starting point. But then AI obviously you can do it with humans too. There's nothing wrong. But AI just knows more. And so you and I could have a conversation about a film but maybe you haven't seen it or maybe you haven't gotten old background. AI has more background context about everything than we have anything. And so I find it's sort of a. Even if I do the reading or the watching kind of quote offline, then there's still that whole like second level reinforcement that I think is so much fun.

Dennis Yao Yu: Yeah, absolutely. So it's no surprise that you're AI forward, right? You said that you assume almost every decision you make is probably better with AI bold stance. Don't know. Depends on who you talk to. AI maximalist perhaps. So for leaders listening to this that uh, are still kind of treating AI as a nice to have or side project. What does it actually mean to be AI forward in practice? At least for you personally?

James Raybould: I think it means that almost every use case I have in my, in my day, not every single one could be better. So example, I'll give you a few. I have an 8 year old and 11 year old, third grade and fifth grade. So they're, they're in school and thankfully they're doing pretty well. But they'll bring back homework sometimes and then you know, they'll get a few wrong. And the old way of doing it's like, oh, well, good try. The new way is I've built an agent where I can take the piece of paper and it shows the questions, it shows what they got wrong. And I can put into my agent and say, hey, can you give me 10 more questions like these? Or, uh, if you want even more nuance, can you give me 10 more questions that are like the two that my kids got wrong? And so again, and with 8, 11, I can probably just about explain it myself. But we're getting to the point where I don't know if I'm probably a year or two away from all of a sudden we'll be doing Pre Cal the self, you know, Like, I'm sure I did it a long time ago, but like, I don't remember actually how to do this. And then if I can. So, you know, so I, I can relearn it myself or I can just give it to my kind of AI agent friend to do that. That's like. It's like a small example or, you know, give another one. I'm training for a. An event called Hyrox for. Think of it as kind of a blend of marathon running and CrossFit. The easiest way of thinking about it. And so now what I'm doing is I'm big. Same thing. I'm like, oh, cool. I gotta build up my training plan and figure out how to do a bunch of stuff. So I'm building again, an agent who's gonna ingest all the YouTube videos and all the stuff as a kind of the corpus of content. And then it'll become my coach. I'm like, hey, I did this workout yesterday. What should I do today? Or hey, here's my week went, hey, next week. In the same way we all pay for coaching. Coaches are great. Human coaches are still great. But all they're doing is pattern matching what they've seen before with different athletes and different research and all that kind of stuff. If I give my AI agent all the workout results from the last bazillion races of hierarchs, plus all the YouTube videos of people who I think are really expert, plus physiological, uh, journal stuff, well, why can't it be equally good? And so I guess, like, you know, my point is, like, you wouldn't necessarily think of like, oh, AI tutoring for your kids or hierarchy coaching for yourself or figuring out strategy for, you know, companies I work with. But there it's all the same of like, wouldn't it be Cool. If there was a source of knowledge and a source of insight out there that could take literally everything the world knows. Certainly, uh, publicly and privately, if you're feeding it stuff and then run it through, you know, algorithms that are way more advanced than my brain can possibly handle, uh, and be my, my, my Sharpa, uh, my guide, whatever, you know, what you want to call it. And so I think back to your comment at the beginning. I just can't think of many things where AI doesn't make me better. I'm not saying it has to be AI does the whole thing. You know, there's like, I could do it myself, AI could do the whole thing. But for, at least for the next few years, we'll be in the, you know, me plus AI is the best solution. At some point, AI will be better. Like, AI will literally be like a better writer than I am, period. And so, in fact, all my suggestions and all my edits will be making it worse. But we're not there yet.

Dennis Yao Yu: Yeah, yeah, I agree. And, um, I think the interesting part, sort of taking your analogy of the instance of, uh, teaching your kids about schoolwork, what I found, at least with my kids, they're learning math very differently the way I learned it. So Claude, or whatever model we're using, they can take into consideration what are some of the newer framework methodology of teaching that coincides really align with the way they're being taught at school. That I don't want to confuse them. Right. So I think that's really valuable in the sense of whether that's for health or not.

James Raybould: I'm not a poker player. I know. Are you a poker player? Do you follow poker? You know, do you follow, like, the World Series of Poker and all kind of stuff?

Dennis Yao Yu: Not so much.

James Raybould: I don't play at all.

Dennis Yao Yu: Yeah.

James Raybould: But essentially what's happened in the last 10 or 15 years is there's this new generation of players. You know, the traditional generation of players, you know, they played analog. You know, they played a lot, but they played analog. The new generation who have been used to playing, you know, 10, 10 hands at a time online, um, so essentially they've gotten the reps in. And so I think, again, back to what you're saying. Traditional education or most things great. I get a few reps, I do my homework assignment, I do 10, I do 10 questions, and so do you. But if you give me a hundred, I will now learn it faster. In the same way Kobe Bryant took more shots. That's why he became Kobe Bryant. I'm sure he had genetic gifts, he took up a lot of shots. And so I think we're in that phase where before you had to wait for things like, do I have access to a coach, do I have access to the gym, do I have access to whatever it might be. In a digital world, you have endless access and it can be completely customized to whatever you're trying to learn. And so again, same thing with, with, with coding. Right now, half of my learning to code is actually doing it. The other half is just like having clawed code. Tell me as I'm going like, hey, I'm building this. What am I doing wrong? And it's like, uh, blah, blah, blah, like, what would make it easier? How would I make you if, uh, I brought in the world's best coder? What questions would he be asking you? And so you're essentially having this kind of interactive experience with these agents where not only are they building stuff for you, but they're actually helping rewire kind of how you think about making them more successful. It's almost like a, like a reinforcing loop of like you're kind of. There's a meta layer of like, the better I get at uh, working with you, this agent, the better I will then more productive I will be in a week. And so I can, you know, so it's like sometimes I just, I just want it built and I don't care about all the details. It's just, let's just build it right now. Sometimes I go into a kind of learning mode of like, hey, let's take two hours rather than 30 minutes and I then learn more and then, which means the next time I'm doing it, the 30 minutes could become, you know, 20 minutes and 10 minutes, etc. So it's also fun. Learning is back to having kids. Seeing, uh, an 8 year old or 11 year old kind of light up when they have that kind of, that moment of learning delight where suddenly like the light bulb goes off. Like it's the same thing with buildings. M. I'm like, oh my God, I finally hooked up the Google auth to my app. And like, you know, it's not always as easy as it's like, just do it. But you get there and now I can. And the second, then you've, once you've done it once, you can do it again, and you can do it again until you figured that there's a whole kind of like, you know, there's gotta be some pain and some, you know, some toil to get there. But once you've done it once, then it's just so much easier to do it again. And so I feel like I'm in that phase. Like the more, the more of those moments I can kind of get A, it's really fun and it's really kind of like a dopamine hit. But B, you're kind of building all these building blocks that then make the next build or uh, the next experience you're trying to bring to life that much faster.

Dennis Yao Yu: Right, right. Yeah, no, totally agree. Taking a bit of a pivot from personal to like enterprises. Um, you have worked inside an AI company serving enterprises. So you saw what separates what succeeded and what's failed or what got stuck and was some kind of bottleneck. Now that you've stepped away a bit more, what patterns are more clear in hindsight? Like what are the ones that have failed? For example, building agents, like what are the enterprises getting wrong?

James Raybould: I think the biggest thing is not. It's kind of funny because in our personal lives you don't have to always have the spec fully figured out because you can kind of figure out as you go and oh, I forgot to add this, just add it. It doesn't work. The ones who pick a very specific problem could be customer support, could be email automation, whatever it might be. Mhm. It's a very specific use case and then you build that rather than trying to. I think what happens is the CEO often comes in and looks at you know, uh, a function and says great, let's AI ify the whole thing. And then you spend a couple weeks and you have months trying to. And then you just realize like no agents are still at the point where they're very, very good. If you give very precise, you know, if you constrain it. And obviously what's happening now and what's so exciting about AI is all of a sudden with multi agents and agent orchestration. If I can constrain one agent, then I can connect it to a second agent, connect to a third agent. All of a sudden you actually can do more complicated things. But I think the, the companies that uh, pick that precise use case get it right and don't try to, you know, go from for example, I'll give you uh, you know, confidential. But my, but my buddy works at a large asset manager institutional wealth kind of, you know and, and uh, there's a question around. Well, do we ERP with AI? Can't we just like replace you know, an Oracle or SAP or work to companies like do it all AI? It's aspirational. It's exciting. But, you know, there's certain things AI is very good at, you know, precisely defined, reasonably small, and then, you know, stitch it all together. ERP is the ultimate kind of monolithic underpinning of everything. And so I think you have this weird tension where some CEOs and some executives don't, quote, get it, and they're like, hey, let's just not do an erp. We'll, like, build a whole thing in AI, which sounds great, but it's actually like a really, really bad idea. And then there's some who are like, I tried ChatGPT nine months ago and it hallucinated once, therefore it can't be reliable, and therefore I can't use it for anything. And both are obviously wrong. And obviously, over time, the person over here who's delusional will be more and more accurate, and the person who's over here will be more and more kind of out of touch. But the right answer is to pick to your question is those who pick a use case, make it failure defined, and also have very clear evaluations or evals. The QAI is, I think, what we're all realizing, you know, reinforcement learning. I'm sure you've read about, you know, all the deep SEQ stuff. And the more you can give. If you go back to some of the examples of, you know, the famous examples of beating humans at chest or beating humans at AlphaGo and Go and AlphaGo and all this stuff, because they had a very, very, very clear feedback loop of like, great, there's rules. And now it's just like, if I try every single thing, at some point, you figure out, like, what the fastest path through is. Same thing with some of these enterprise use cases. Is, uh, the reason why customer support is the one with, you know, whether it's Decagon or Sierra, all these companies, it's obviously, it's one of the spaces, you know, next to coding, that's probably been the most impacted already. And it's because there's a fairly clear what does success look like? And some things like, I'd like to build an AI agent for strategy. Very reasonable, by the way, but kind of hard to have the training data set or the evaluator because you don't know if it's what works. And maybe I can put in, like, all this historical strategy data, but you still don't know. When you and I decide to. Great, the other group's going to pivot and do this. Cool. It might take you two years to know that customer support you know, within 12 hours of rerouting some of the tickets from humans to AI, et cetera. So I think in some ways, the quicker you can get a feedback loop going and the more you can build a really quantitative eval, the easier it is to kind of have confidence. The more fluffy and the more nebulous and the more, um, unconstrained it is in some ways, like, oh, it's so cool. But all of a sudden then you spend a couple months and it doesn't quite work and you don't know if it's working because you're like, well, what metric? And you're like, I don't know. And so I think we're in that phase where we're still in the. I would recommend pick use cases where there are metrics that you can fairly clearly see whether this is working versus those fluffier, more nebulous ones where I think in a year or two that may be different, by the way, but for right now, I think it's pick use case with metric and you fairly narrow and then stitch together, you know, from one narrow sliver at another one and another, and then all of a sudden you've actually taken a whole flow and rebuilt it versus trying to completely rebuild all in one go.

Dennis Yao Yu: Right. I think this is great because a lot of customers or a lot of people we speak with, I think the confusion really comes in when it comes to like, oh, what, what should we utilize AI for? So obviously customer support, as you mentioned, you know, it's a, uh, it's a, where people have tested, it's tried and true. There's also software development, there's also backup office operations. Some of things are more metrics oriented. You can actually, you know, bringing in a little bit more that, uh, you can measure the effectiveness without disclosing any sort of client names and so on. Can you kind of walk us through? Because I think the other challenge for a lot of folks is that they, once they identify area they want to try or they want to implement AI on, they may not understand the sort of end to end process. What that actually could look, uh, what actually look like? Whether they hire a consultancy like touring or they're doing it internally. Can you give us an example that we've seen that's been successful? What does that end to end process look like?

James Raybould: I think the biggest thing actually is for most folks who hire, you know, let's say you're used to hiring a Accenture or, you know, systems integrator, et cetera, there's usually a Failure to find. We're migrating from SAP to Oracle or we're, you know, building something. And I think with, with traditional software, there actually is kind of like an endpoint where it's like, cool, you know, it's like kind of building a house. At some point the house is built and you have the foundations and blah, blah, blah, blah. I think with AI, what's interesting, and I think this is something we were learning at, uh, Turing and more broadly, is it keeps going. The models are always changing, the weights are changing. So it's one of those things where. And there's like a fine tuning element if I build software and if you give me a spec and I build software, I can like, yep, done, done, done, done, done, done, done, done. And like, and it's almost like, great, here's your pizza. And the AI is more like, great. You're constantly changing the ingredients and you're constantly fine tuning. And so it's one of those things where you have this. The frustrating thing about building the AI right now is you get it built pretty quickly, but then actually the time from it's built to. It's really, really working. It could be a day or two, or it could be a month or two. And that's where I kind of. I wish I had a kind of a rubric for you of like, for these use cases, one day, for these use cases, two months. But I think it's almost like the. It's just, you know, it's not, quote, deterministic. And so therefore it's something where that tuning element and that constantly improving. It's also, you know, with these models, if I build you a piece of software, it's not really getting better. Does that make sense? Like, you know, we build multiple databases together. Great. It's built a year, you know, a year goes on. Maybe I can run it more cheaply because, you know, the cloud prices came down. Like, it's the same product with AI as, you know, we've all seen my favorite curve and you could link to it, you know, is the, uh, meter curve showing how long a task in human time, AI can do. And, you know, and every seven months it's doubling. So, you know, so when ChatGPT3 came out, I guess three years ago, it could do, you know, a couple seconds, like quick, hey, answer a quick question. We're now up to like half a day. So basically it can now do tasks that take us half a day. And it's doubling every. It was every seven months. Now it's more like every Four months. So basically we're getting to this point where the improvement curves are so steep that I've, quote, finished the build for you in March. But A, I've got to tune it for a while, but B, by May, the models may actually be 40% better and then by September, the models may be 200%. So it's almost one of those things where it's like building a house, but the house needs to be ready to. It's like, you know, built a different way, where all of a sudden we're going to actually, the house is going to get three times bigger in the next six months. We're not quite sure in which direction. Maybe it's going to go out, maybe it's going to go the other direction, maybe it's going to go up. We're not quite sure. The, you know, I do. I think. I think the. I think the, uh, human architecture is a very, very bad analogy. I think it was a good analogy for building software. I think it's a very, very, very bad analogy because actually everything's going to be malleable and everything's going to be flexible. So in some ways it's less that you have to build it right once and say, great, here's your erp. Here's your. Here's your database. It'll work for the next 10 years. Like, no, no, it's like an ongoing thing. And so again, back to your question. If you hire a consultancy and then they go away and you don't have the right folks in place, it's sort of. It becomes out of date very, very quickly. It's like, it's like, great, I built you a bunch of. Bunch of TikTok videos and then I stopped producing TikTok videos. TikTok videos. Like, you need to keep doing them because, you know, memes change and timeline change. You know, it's not timeless. And so I think this isn't saying you have to, you know, hire a, uh, consultancy forever. I'm not saying that either. But the handoff process is different when it's build me, you know, migrate my erp. Great, thank you. You're done to. You're never, you're never done. You're always, you know, adapting that house and adding rooms here and there or, you know, having it expand in all these different directions.

Dennis Yao Yu: So what do these companies do then, at that point? Let's just say they want to get something started. Obviously, they don't have the internal expertise, for example, on AI. They don't have AI engineers, and they Hire some consultancy to come in and do strategy as well as implementation at some point. Like you said, they're probably not going to hire consultants forever and they want to bring it in house is the best strategy for them to start looking for talent internally and then have that sort of handoff. Much more, you know, sort of a smooth transition.

James Raybould: You're going to have many guests on the show. I'm going to be much more on the kind of AI maximalist, et cetera. So I guess, you know, I guess my feedback would be I'm probably unbiased, maybe I'm overly optimistic, is I think everyone at this point remember how 20 years ago, or maybe, maybe it was 25 years ago. Oh, I don't want to be digital. I'll continue to transcribe, to dictate my emails to, uh, my secretary, again being very out of date on the terminology I'm using. So. And then, you know, and occasionally there's still people who have their emails printed out. Like, no one does that anymore.

Dennis Yao Yu: Mhm.

James Raybould: So my point is like, we're already there. AI is, it's like that, but instead of it taking 20 years, it doesn't take, you know, it's gonna take three or four years. I think the. If I were an exec who was not who like, doesn't, quote, know what to do, I think it starts with like, understanding yourself. I think, I think AI is something where you actually have to, you have to feel it. I feel like, you know, it's one thing to see a demo and it's super cool, but like, at least for me, and maybe same, I'm curious. Same thing for you. It's when you see it for yourself, you're like, holy shit. I just gave Claude code this piece of data. I'm pretty good at data analysis and it would take me an hour in Excel to create a bunch of pivot tables and write a bunch of scripts, et cetera. And it crunches it in 30 seconds and again and in, in two months you'll crunch it in 25 seconds and in six months you'll crunch it in 15 seconds. You can choose, you can just kind of see where the quote puck is going. And so I think, I think the world of, uh, I think if I were, you know, my colleagues, et cetera, execs, I think you just gotta get your hands dirty. I think you've got to get, quote, close to the metal. I think the whole. You can hire Bain, great. I used to work at Bain, but go hire Bain. Yeah, but you've Gotta understand that. You gotta feel it yourself. It's kind of one of those things where. And I think the nice thing about AI too is I know you are, you're a tinkerer as well. Doing it in an enterprise. Yeah. There's security and privacy. You know, there's a lot of stuff that's important and kind of quote hard for the right reasons. In your personal life, you can build that AI tutor I described a few minutes ago for my kids or I can build a, you know. Do you remember the movie Clueless where Sharon Horowitz, who's she. She builds a wardrobe builder where she can see where what's, you know, what's her top, what her bottoms, do they match? My, my 8 year old the other day was said, hey, my clothes don't match today. I don't, I kind of don't like them. And I'm like, great, we're going to build you an app where we'll take a picture of all your stuff and we'll just be able to kind of do that like. And I built the basics of doing that in one hour and it wasn't even one hour of hands. Like it's like great. Write, get, get it. Write, write something. Wait four minutes as Claude co does his thing. Write one more thing. Wait. So of uh, actual like my time, it probably took 10 minutes and again, is it perfect? No, but so I think at that point of like you just see the, the gap between idea you have and functioning product is now in minutes or hours. Again, I'm not saying you can completely transform your go to market in minutes or hours. That's delusional. But you get the same sense of like, all the reasons you can't are things like human change management or system stuff like you actually could. But it's all kind of the inertia of the way we've built things for the right reasons because we didn't know AI was quote coming. But I think what's so interesting now is seeing which CEOs and which companies are going to do kind of quote complete rewrites. Like basically have the conviction to not kind of incrementally, you know, you know how like our, you know, our, our skin sheds again. I think, I think like every month or two like all our skin is like, it's like gradually been shed before. They're like, no, no, I'm like rip off all the skin in like a month or in a day. And like, and like, and I, I, I, and I don't know the answer yet. But I think the, you know, like, almost everyone's still kind of. I'm m. Kind of like, you know, very into first principles. Like, you know, first principles are not all these coordinating steps between different humans to get a procurement done. Like, you actually break down, you know, what you need to do. AI or, you know, if you, if you had all the humans together in a room, they could do the whole thing in five minutes. The reason it always takes a month is because someone has sent to someone and they're, they have a queue and then someone has sent someone like, you know, so all these layers of checks and balances. Again, I'm not saying approvals are bad. I'm just saying the way we do it as humans is that you tend to kind of have three or four stages and three or four gates and each one has delay and another delay and another delay. And that's why building software takes a while, because teams have to coordinate, blah, blah, blah. But if you can just get it, you remove all this coordination, you can build things much, much, much faster. I think that's why. Back to your personal lives. Once you see what you can do in an hour. And again, I know you're an example of this too. If you can build this in an hour, surely your effing company can build something in a month. Does that make sense? Whereas right now it's like, that'll take us six months.

Dennis Yao Yu: Yeah.

James Raybould: And it's just, it's just not, you know, I think it's like how. It's like, how do you readjust to the speed limit? It's gone like, you know, 65 miles an hour to 700 miles an hour. And like, it's like a, it's like a hard adjustment.

Dennis Yao Yu: Yeah.

James Raybould: But that's kind of. You have to almost readjust your brain to what it looks like you're going from like driving car to flying a plane.

Dennis Yao Yu: Yeah.

James Raybould: And we're all.

Dennis Yao Yu: I resonate with that. I resonate with that so much where sometimes I have like multiple sessions going on, multiple tabs going on, multiple agents running back and forth. Right. I also, we talked a little bit about this as well, is that I feel like not only am I training the AI, AI is actually rewiring my brain. The way it's training me, the way I need to think. If we're talking about this is the speed of, um, you know, where we're going to go later on. But the great thing is, kind of to your point, traditionally in organizations, everything's done on projects done sequentially because one depends on another person, or another person, get it done. Or another department. But now you can do everything all at one time. So as an orchestrator, as a, just called a project manager in the sense that you need to be able to context switch pretty quickly and as fast as you can. And I think that's also I'm being trained by AI, no joke.

James Raybould: Also, I mean I'm sure you've seen it to me that the reason why obviously we're recording this M in March, but things have changed a lot over the last four or five months because all of a sudden all these multi agent systems, you know, when you use Claude code, as you know, you say, hey, blah blah blah, it says great and it spins up 2, 3, 4, 5, 6 agents at a time that all, you know, and then they do all the coordination, all the coalescing and all of a sudden, you know, a minute, an hour, whatever it is later, it just comes together and again, does it always work? No, something doesn't work. But it's that same idea again with humans. All the coordination costs take time. And you know, you have to sleep and you have to, you're on DTO and like, you know, blah blah blah, someone's busy. None of those things happen. AI is ready whenever you want. You take a break for a week. It doesn't care. You want to do it at three in the morning because you have jet lag. It doesn't care. So I think kind of it's like, yeah, we have all these human driven constraints about how to build things, how to make changes, how to get things done. Once you remove all those, it can move a lot faster again. I'm not saying, you know, humans are still really important. There's still gonna be lots and lots of things that we have to do that are human. That said, there are a lot of things we do that are routine, rote, repetitive, whatever phrase you want to use that AI can do cheaper, faster and better. And that's a good, that's a dangerous combination.

Dennis Yao Yu: Absolutely. This kind of is a good segue to the next question and I'm very excited to hear your side of this. We talked about the technology there and it's always going to be, it's only going to be better. We'll talk about the process that's there. A lot of times, depending on the size of organization, the bigger it gets, the harder it gets. Which is the people piece and that's the sort of organizational piece. Right. I'll give an example. A year ago I was speaking on a panel at Phoenix and Houston with a group of people who are non technologists, they're teachers, they're blue collar workers and so on. The conversation obviously, obviously was around AI, but every single topic or every single answer I noticed was driven from fear. It was, what's going to happen to my job, what's going to happen to my kids at school? Um, a lot of times, obviously that's more personal basis into the organization itself. And I know a lot of organizations are at that stage. What have you seen or uh, what would you recommend in terms of how to enable and how to help people in this sort of transitional stage into like more AI forward or AI first?

James Raybould: I think it's come back to just seeing what it can do for you. So again, like, you know, there's definitely fear and there's, you know, the, there's, it's unquestionable that society is going to change very, very quickly and very, very meaningful. Like unquestionable at this point. Like, I don't know if it's 18 months or I don't know if it's five years, but you know, it's going to, it's going to foundationally shift. That said, if we focus on all the things it can do for us, then you're at least kind of, you're excited by it. There's still the existential, will there be jobs? And that stuff's very hard. Like, I'm not an economist. I don't know. I think everyone's workflow right now has the same. I gotta check my. There's a lot of these like things like no one likes doing a bunch of these things. So I think, I think if I, if you could, if you remove the existential stuff, which again is hard to do. But like imagine if, you know, your, your boss suddenly is like, hey, I'm hiring you an assistant to do this for you. I'm hiring you a chief of staff. Everyone's actually like, that sounds fricking amazing. It's just, it is that. It's just a different version of that. So it's like, you know, the framing of why not take advantage of all the, you know, again. Yeah, I like the analogy of just imagine your boss or your CEO like, hey, good news, everyone. We're giving everyone your own ea. We're giving you your own whatever it might be. Everyone act like that's amazing. M. So at least, you know, take advantage of. And of course, you know, if there's the. And uh, by the way, you know, in a year we're going to Remove half of you, that's different. But you still have to kind of. I think what happens is some people shut down and when you shut down, you don't take, you don't get the benefits. But then you have a kind of this existential dread following you along the whole time and which may happen along the way. There's going to be all these enhancements, all these improvements again, professionally and, or personally. And so I think just, uh, again, I'm a maximalist, I'm obviously an optimist, whatever phrase you want to use. I'm someone who, I think also I find friction very frustrating. Like, just like in my life, like things like, oh, traffic lights, you know, with traffic lights where you get the traffic light and there's no other car around and then it hasn't, you know, it's not using the sensors. Like you're just waiting for the effing thing to change. Right. Those things just, you know, some people are like, oh, that's cool, I'll just wait. That's like not kind of how I'm wired. So I think I'm unusually, uh, I don't know, AI friendly because I just see in my life all this friction. I see that, uh, ah, I've had this to do for the last eight days and it probably had only taken me three minutes, but it's been there. If I could have just given that to my uh, agent, which I will now, and say, hey, can you follow up with Patrick, get us time next week, probably on Thursday, it would have been done. And so I think kind of the. Yeah, I, I guess I think you have to focus on. There's going to be. There are, there are, you know, longer term downsides that are kind of almost unknowable. But to ignore these short term, medium term. But there are also a lot of long term amazing things too. But you get the idea to basically put your head in the sand and not embrace the most powerful expansion of our intelligence ever. Yeah, that keeps getting better. Again, it's not like the aliens landed and gave us this cool thing. It's like the aliens landed and gave us this cool thing and then every month the cool thing gets 25% cooler. And like that's kind of what AI is like today. So.

Dennis Yao Yu: Yeah, well, what is that saying? They said today's AI is going to be the worst version that you ever use.

James Raybould: Yeah, AI is the worst it'll ever be. Exactly. And again, exponential curves are hard for us to process. You've seen. And that's why I think back to the conversation we had earlier. If you're on the front lines and you're kind of close to the metal, uh, I've seen how Claude code has evolved, and, you know, clawed code has only been around for a year, but in the last three or four months, every single person who's using it regularly has had the same reaction, like, choose your paternity. This has now gone from really good to, oh, my God, you and I can now build five different things in a weekend that would have taken each of which would have taken a team a month and again. And then you think, oh, great, that's happened in the last four months, and four months from now, maybe you and I can both build five things in an hour that would have taken a month. So it's just like, you can just see the. The. The progression. And so that's saying to me, the. The best proxy, in some ways to me, for kind of executive readiness, company readiness, whatever you want to call it is. Are. Are the people close to the metal? Are they actually using the tools? Not like, occasionally putting in, like, a quote, quote, quick copilot query. That's not AI today. That's like AI 2023. AI today is you're automating, you're amplifying, you're building agents that make your life personally, professionally better. And as we both know, it's not very hard. It seems super scary code. Oh, do I have to be an engineer? No, you have to be able to write English. You have to say, I'd like to be able to do this. And then it says. And then it says, what should we use? Hey. Huh? Do you want to use Vercel, or do you want to use Cloudflare? I'm like, I don't care. What do you think we should use? It says, let's use Vercel. Cool. And it does that. So, again, like, you know, I'm not a techie. You're. I don't think you're a techie either. You don't need to be a techie at this point, as long as you know what you want and you know what the use case is. You know, what good looks like. The how doesn't matter so much. Again, for enterprise, it's a little different. Enterprise, you still need, you know, real engineers, but for getting something from 0 to 1, it's now measured in minutes or low hours for most things. Maybe a day or two, something robust. Is there anything you've built recently that actually took, if you'd been focused, huh, For a day or Two, it wouldn't have taken you more than a day or two. I think cowork from Claude took 10 days for them to build. Claude code is the shell and the foundation. They built it in 10 days.

Dennis Yao Yu: That's incredible.

James Raybould: That's nuts.

Dennis Yao Yu: That's crazy. Yeah, that's just incredible. It's just a speed to market and kind of to your point. And that also pushes along all the iterations and features and everything that says sort of going forward. I mean, we talked a little bit about this as well. Like just for fun, you know, being a go to market for like 20, 25 years. One of the biggest pain points I see is a lot of times, uh, the functions are silo, whether it's like sales or marketing. Marketing doesn't talk to sales all the way to like partnership to customer success and so on. Granted, we're a smaller sort of. But, you know, we were able to build different agents solving that issue because that context, as we know, the importance of the context being able to understand and learn from across the functions. That should be a flywheel everybody talks about all the time, right? Oh, Go to market team should be a flywheel. But in practice, it is really hard because they're humans. Humans have metrics and they have, they have guardrails and they have things that they want to protect, perhaps even on the personal agenda standpoint. But with agents, they don't care.

James Raybould: They just execute whatever you need them to execute. Obviously, a big piece of go to market's gonna be voice and tone and voice and tone and star. Cause like, you know, we write these principles and we write these. You know, these are seven elements of voice and tone. And then, you know, the, the CMO or people have to like, you know, review and constantly. Like that's not quite. See now, you know, the companies have documented in really clear. You know, again, back to that kind of like the Fred Coffin from the beginning. If you have very clear guidelines and very clear evaluations or, you know, evals, then all of a sudden building a. Hey, here's the document we created. Does this sound like your company? No. Does this sound like, you know, so I think there's kind of. I'm in that phase, especially right now, where building reviewers couldn't be easier. Does this deck tell a story, Connect to someone's paypoint in the right voice and tone? Does this email? Does this blog post? Does this, whatever it might be. I built this silly thing that you may have seen, like a LinkedIn profile analyzer, where it's going to analyze your LinkedIn profile, et cetera. That re reviewing stuff is as simple as what's the rubric we're using? And then just give me the material. And so companies here, as long as you have a clear and then what's happening now is people like who people who are never clear about. We don't have all these, you know, clear onboarding um documents. We don't have these clear voice and tone documents, we don't have these clear. Here's how we build. If you don't have those things, A, it's probably chaos where you operate with humans but B now you can't give these things to agents. So in some ways the companies now that are kind of going to thrive the most are like those who've documented and have actually like you know, at your point before, like now you're learning. Oh if you know how the future is evolving. Documentation, orchestration, they're all the same flywheel. And so in some way and I can help you. It's like, hey, what do you need? Like hey, here's what I got. And they're like, okay, that's a good start. Now I need these three documents or hey, this is good. Like it's just this iterative quick back and forth and you know, it quote knows what it needs. Sure. And then it's just, you just have to give it to it, you know, for the best company. Like cool, we already have that and the whole thing takes hours for those like, oh, we don't have a way of defining our voice and we don't have a way of defining our go to market story. That's going to be tough to create uh, automated PowerPoint builder thing if you don't have your actual messaging 100%.

Dennis Yao Yu: And also kind of adding to your point too, if a company doesn't have this voice or tone or messaging or positioning, this is the best time to create it right now too, right? Using AI as more partner.

James Raybould: Now we create a V1. Now we can take. Great. Here's what I think. Now run through here are eight competitors. Go take a look and see are we differentiated? Are we saying, are we all saying the same things? I think it's never easier. Like great, just run through all their stuff too. You're all saying the same thing. We help you optimize, we help you be more effective. It's never been easier to a review what you're doing but B evaluate uh, what others are doing too and see how they're talking about themselves and what they're doing or who they're hiring on LinkedIn, like all these things. Competitive intelligence has never been easier. And we'll get. And as agents get over, it's like, oh, well, still a human has to go and review a bunch of stuff very soon. Every blog post, every LinkedIn post, every X post, everything that not only the companies, but um, the employees of these companies are doing, you can kind of just automate, et cetera. It's like, it's. Yeah, I mean it's wild again, but it's the kind of thing where there's so many things that a company would do if it were cost prohibitive. Like you would actually hire like hundreds of people. Like just go, go monitor these sites, go read everything on Reddit, go do all these things. And you didn't do it historically because you couldn't do it. Now you can do all these things. I think what's uh, again, back to your point about AI is like sometimes people only focus on the cost reduction, which that will happen, but the amplification, the expansion of. I can now do all these different things. I would never have thought about doing this before. Like, I would never have built a wardrobe picker for my 8 year old. But if it takes, if it takes 20 minutes, I'm gonna do it. So I think kind of, you know, the barrier to so many things in life is that would take so long. That would be too expensive. I'd be better off doing this thing. So now you have to choose. In some ways, the hardest thing about M. AI is gonna be, you know, strategies. You know, it's like making decisions. I, uh, can do A but not baby with AI. Once you can actually do A and B and C and D and E and F, in some ways we're going to have this m cacophony of stuff. The companies that, uh, keep it, ah, uh, what's the word? That direct it, well, will thrive. The companies will just Suddenly just create 97 new agents. But they're all kind of. There's no coordination layer. Won't. It's like, yeah, it's a great problem to have where it's like, oh, I have too many things I could build. What should I build? You know?

Dennis Yao Yu: Right, right. Then the judgment comes into play. I feel like, Jason, we can talk about this for hours, obviously, and we have before.

James Raybould: So, you know, we'll probably revisit again

Dennis Yao Yu: in about six months and then we're going to have a whole new layer of information on A.I. uh, but this has been extremely valuable and I'm sure for the listeners out there as well, so thanks for taking the time to chat with us. For anybody that want to reach out to you, since you have such a great LinkedIn presence and then your insights always been very valuable, should they just follow you on LinkedIn or is there somewhere else like you on it?

James Raybould: Joe uh, I know X is like making a huge comeback in the AI community so I'm in theory I should be on X, but I'm um. Yeah, LinkedIn's the best place. I worked there for 12 years. It's kind of like my home base for sure. Thanks for listening and ah, we'll catch

Dennis Yao Yu: you on the next one. To everyone listening if you found this valuable, subscribe to the podcast AI for business leaders on Spotify, Apple podcast, YouTube and leave a comment. Please also share with others navigating the future of AI. Thanks for listening and we'll catch you on the next one.