About this talk
In this talk, Alexey, an organizational design consultant, explores how companies can effectively leverage AI technology to enhance productivity and competitiveness. He emphasizes that simply adopting AI without rethinking existing processes will not yield significant improvements. The speaker discusses the importance of minimizing transaction costs and enables a cycle of discovery and delivery in organizations. He also presents insights from his book, '10x.org', advocating for a shift towards multi-learning within teams, allowing individuals to acquire diverse skills and remain relevant in the evolving tech landscape. By embracing these strategies, organizations can better integrate AI and foster innovation rather than merely applying it to legacy structures.
Full transcript
Something good. Welcome back everyone. Hope everyone's ready from lunch. You can get some food in their stomachs. Everyone ready for the next round of talks. We have a really cool talk again. Like all of them. Um, I don't know. Have you guys wondered how some companies just like to sprinkle AI on everything and then wonder what magic will happen without actually thinking about the processes behind it
and what to do to make these things work and actually be productive? Well, I have for sure. Our next speaker, Alexey, is going to talk a lot about things like this. So, big round of applause for Alexey. >> [applause] >> Am I Am I Am I Yes, I'm here. Hello everybody. Welcome. Um, sorry for interrupting this nice blues music to talk about um, this topic. But, I
hope you don't mind. Um, my name is Alexey. I'm an org design consultant and I help my clients to be smarter about intelligence. I have hair, but I wore this kippah because otherwise I don't see you. There's these lights. But, I want to see some of you. And now I do. You see me well, right? You hear me well. Fantastic. Let's get going. So, you said you've
got an AI AI software development life cycle. Okay, but which one? So, um, this is not really a technical uh, topic. This more managerial uh, topic. Um, and essentially the question is a broader question is, how can we actually benefit as as an organization from all this great AI uh, technology that everybody is now using? Right? So, this talk is about um, that. Yesterday I spoke at
a meetup and it took me 90 minutes. Today I have only 30. So, let me try to make it really crispy and precise. And the the crispiest form I can come up with is this talk being recorded? No? Uh it's only you and me, right? So, I would say we guys are [ __ ] That's the sharpest way I can say, and I will I swear I'm not going
to swear anymore in this talk. Uh so, I hear uh my clients say, "Okay, we already have AI SDLC." I'm like, "Are you done? And which one of those?" And by the way, like if you remove AI, is it the same old or is something changed? And more importantly, I mean, do you see real improvements? So, this is the question here. And historically, you know, consultants they
have this internal secret that never promise more than 10% improvement to your customers because otherwise they will be embarrassed if you actually can improve this that company more. But now, I think 10% improvement is not enough uh because AI is there. We know it's fast at doing many things, and maybe if not this week, then next week a new model will drop. And it will actually be
way faster than anybody of us. And if we humans just come with a 10% improvement, we're not comparable. So, we out of job. So, this talk is essentially how to stay um relevant, if that makes sense, in your organizations as a individual topic. And for our organizations, how actually to stay relevant because all these startups with five Mac minis will eventually under catch you. So, I'm here
with a um on a book uh trip. There's a book, and I will be after this talk, if you want, in this hallway bit bit between this dark place and that brighter place somewhere in between I'll be with a book if you want to have a look. And the book is called 10x.org. So, we are saying if you really want to compete with the eyes as individuals,
as developers, and as employees, you really need to do something differently. And the book is doing actually pretty good. Some people appreciate it. So, I'd like to give you like some gist of it. So, I have this friend Fabian Wesner. He's a CTO from Berlin, quite a known person in small circles, and he's a AI innovator. So, what he's doing, he actually wrote like a 50-40 page
specification, and every time when a new model drops, he actually gives this spec to his team of AIs and checks actually how well they are at delivering that whole e-commerce shop. So, he tested different models, and actually GPT 5.5, as you know, is pretty awesome. This is not my point. My point is that AIs are actually good at doing a lot of cool stuff these days. And
I talked to Fabian recently about my book and other stuff we're doing together, and he actually asked me, like, "Alexei, like, how did you guys write how how did you guys write this book about AI and AI implication on management because AI it's such an emergent, fresh, novel topic. How can How can you ever put it in a book, right?" And the answer is, yes, of course,
some things are nice and fresh, and like every week we actually have these new models, new features, new harnesses. It's like an amazing time to be alive and active in the field. Yeah, we all need to actually appreciate the hard work of those engineers. Some things are super fresh. Some things are not that fresh, and maybe these are our organizations which actually had been there when AI
came to be. Anybody's working in a non-AI native organization? Org which which is by definition org that existed before this wave of AI. Right? And it had been developing itself, like grew organically, some compartment here, some department there, some individual here, some expert there, and it was functioning, and it's still a functioning place, maybe even a happy place. But the question is, like, can you really compete
in your org um with maybe this person? Anybody knows who this guy is? Who is he? Yeah, I said Peter Steinberger, which is the creator of OpenClaw, right? Uh so, like, a super cool developer, right? And this is actually a real photo I found the way he works. So, as you can see, he just spawns agents, and he just gives them a task, and, you know, clarify
some things. Think most of us work like this at night when we are outside of our old old slow legacy organizations. So, our org cannot really compete with that unless we do something different in the org. That's the message, and I think this guy is like 100x compared to like a normal developer I used to be. Well, he's also super smart, but he's also using intelligence on
top of his smartness to make it like a double smart, if that makes sense. And you can say, "Yeah, this is just one guy with five computers." But this is not a topic I talk about Anthropic, but just to give you an example and you don't need to read all that text, but it just I asked Gemini, give me the release notes of all the big things
Anthropic has shipped from February to March. And these are the things model harnesses, design like like wow. You can actually ship 100 x compared to other orgs even at scale. And how many employees Anthropic has? Like a thousand something? See, some companies know how to do things differently. I'm over excited. I need to drink some water to cool my nerves. Um as I said, this talk yesterday
took 90 minutes. I've only 23 left. What does companies do differently? Um a metaphor, okay? So, a bird has two wings and one wing, which is very different, is this one. And it's a bit of scientific. In the book, we try to make it more like clear and simpler, but one thing they do these companies, they minimize transaction costs. So, they're trying to make sure that cycle
is very cheap so that you can make a lot of cycles. If you've been familiar with with the lean startup MVP, like build, measure, learn, right? That cycle and it's not just discovery to delivery thing we used to do before in the Agile transformation field, it's a discovery to delivery to observability, getting metrics, and operations to actually improve what we've just shipped based on those metrics. And
then back to discovery with better hypothesis. So, these companies are good at doing that cycle super super efficiently. Another thing and a bird requires two wings to fly, obviously, is it's not only fast at looping, they also found this way to meet meet meet meet to minimize what in finance is known as switching costs. It's the cost required you to stop what you have been building and
switch to something new and different because now the value is here. Okay, so you go fast in your lane, build, measure, learn, and once you discover that now value is there because you've shipped enough value here, you go there, you open up a new repository, you get surprised, you talk to your AI, and now you're building something new. And if you have two wings, you can actually
be at least 10x times better than any normal organization. Let's go Let's go dive a bit deeper on that. But this is actually the essence. That's the formula. That Does this make sense? And you know, like if you are going to quit now to have some a coffee, uh the message is this is the formula, and you can use AI actually also to help you do all
of it. Okay? So, back to our traditional And in the book the book is written as a business novel where like mixed with with a theory. And the business novel, there is this three persona, and Eric, he he is engineering manager, and he takes his colleagues, Hannah, head of HR, and Paula, head of product, to a tour to his IT wing of the company to show how
AI is already being used. And he says, [snorts] "We have AI SDLC." And they're like, "Show us." So, he shows and tells, "What used to take days now take hours, sometimes minutes." That makes sense. And now he approaches a developer, young developer Debbie, super smart um girl, and she says, "AI allows me to finish my front-end work in a fraction of time. That makes sense? Is it
also your experience? Can do things faster with AI? Yeah. it's not a tricky question, right? But the head of product at Apollo she's not satisfied because she sees that despite everybody is using AI she knows the road map and road map is way behind as before. So those local improvements do do do not really compound at a global scale and she she says everybody seems faster but
the whole is not getting any better. Well, maybe she's British and she says it's not getting any better. But it's not changing the point, right? I mean global local two different AI adoption is not just a technological thing and there are great talks now in the bright side of this conference about the technical side of things, but social technical is the system. You cannot just spread AI
on top of your existing org structures, silos, whatever and hope it will actually improve everything. There's a system of things and this slash in [snorts] between, that's where the friction is actually happening. And we saw this already with agile transformation. So I've been in the field for 25 years as a consultant. So I've seen that with agile, but now AI is way faster showing us this friction
and this discrepancy. So back to our organization, there's this nice corner office, of course, with a nice mountain view and there's this CEO and he he looks like I'm looking at you now, guys. And he sees people working, everybody's busy, screens, AI agents. And there's some nice values on the wall, of course. But then what he really sees is his metric. And two metrics are like the
cost are climbing, right? Because we know just a burning calories, we also know a burning tokens like crazy. And we hire AI consultants. Don't do that. And other things happening. So the cost are climbing. But the performance as before, maybe even got worse. Because guess what? If everybody outputting faster, the queues between silos will only grow. So the lead time might actually get worse, not better. But
let's not go He sees no improvement in performance. Wow, how come? And I call it one x org. It's one x improvement to before AI. No improvement. Is anybody there with in a one x org? Despite everybody is doing everything, same old. Same old. And I mean, if you don't trust me, there lots of reports, slightly outdated, slightly before Opus 4.6. So maybe we're going to see
some change this summer. But the previous Dora report actually proven that everybody is faster, nothing really nothing really happened. So we will see this discrepancy before between AI native and hopefully AI augmented. So adding AI inside a legacy org to do something better, better. anybody been in the Agile transformation? Anybody old enough to so this? We are not going to actually fake AI transformations as we did
with Agile transformations, I think. It's going to be harder. Why? Because in the Agile times, I as a Scrum Master, Agile Coach, used to say, "Oh, it's important that you guys are more agile now than you were a year ago. Good boys. Keep improving. Continuous improvement one step at a time." With the AI, it's not going to cut it because you can compare yourself to a 10x
org. And if you don't believe it, well, people who work in a 1x org now at night are live coding or AI engineering. And they really see this difference. Night and day. Day and night difference. Make sense? we need to do something something about I'm still been building up the case, okay? Why we're [ __ ] Um Um let's visit one cubicle of that company. Uh in the cubicle,
there's this wonderful guy. By the way, that that's GPT. It's pretty good at drawing, huh? Sloppy? Not sloppy? It's It's pretty good, I think. Um I was surprised yesterday to discover that. So, this guy, he His name is Debian, obviously, because he he he's a database designer. And uh well, I like picking the names which remind you what the role of. That's why we have Eric, Hannah,
Paula, Devi in the book. he works in an organization where there's a strong belief management belief that you really need to leverage expert knowledge. So, if we have this Debian who used to be forever a database designer, we need to keep loading him with DB design work. Does this make sense? Because this is efficient, right? So, this definition of efficiency is Now, guess what? AI Debian discovered
AI or AI discovered Debian. They meet and now, of course, he's a way faster than Maybe not this week, maybe next week. There that a new model out there, which makes his work faster. That is plausible, right? And now he says, "I got an AI as DLC." And his meaning of that is that now I can be more productive in my zone of expertise. Fantastic. So, in
3 days, he finishes database design for the entire year. Yoohoo, a big win. Right? And then he's like, "What am I supposed to do in the remaining 362 days?" of course, he can crack more databases. Right? Or maybe do something else, but do we need so many databases as he can now produce with AI? There's no demand for his work if he actually is going to be
even 10 times And then I don't want to be dramatic, he's gone and his cactus is dead. it's a bit of overplayed, of course, uh dramatic moment, but the question is how to keep people relevant. And making him do more databases is not the way to keep him relevant because nobody needs so many databases. Oh my god. What he's missing, he's missing this wing of a bird.
He's not on the whole cycle, not on the whole value cycle. He just does one part of the cycle. And of it doesn't create any value if you just speed up one part of that. The whole thing needs to be sped up and you need to remove the bottlenecks and you need to speed up the whole cycle. You guys understand that, I'm sure. So, he is lacking
this wing of a bird. Now, in another part of our wonderful office is this great team. It's a search team. Anybody work on a search team? So, if yes, that is like completely unrelated. don't take it personal. Um so, it's a search team in the building and maintaining e-commerce search search functionalities. They are like gurus at search, and I know it's a hard field, so I'm not
making any jokes about the expertise. They are the search team, the best in the company. So, if there is something to do with search improvements on this e-commerce website, it is this wonderful team that is give getting this work. And of course, the core belief in this part of the company is that we need fast flow. And how do we make this team a fast flow We
put what? Team topologies on top. Spotify model, whatever, and we actually keep this team running in this narrow lane, streamline team, right? So, they if they can focus only on the search, only one lane, fantastic, they'll be faster than anybody else. And this is true. And guess what? They discovered AI. And AI makes them 100 times faster doing search. Oh my god. What's going to happen next?
You guys know, right? Same question. Is there enough work for them? Do customers always want improvement of search like forever, unlimited? Is it unlimited demand on that? Well, if it's a search company, maybe, but they are in a product development company. They are in product R&D, where search is just one functionality of the customer journey. So, now they actually have one wing there. Hopefully, they can utilize
the skills, the multi skills to work on the whole I've only 10 minutes left, I know. But they lack in the other wing, and the other wing is they do only search. Not because they are stupid or because they are born search engineers. No, because the company puts on this lane because they believe is that it is the fastest way, most efficient way to do things. And
in finance there is this term, the law of diminishing returns. Like doing more of the same eventually will stop making any sense. So, at some point more databases, more search, even faster doesn't really justify the investment. So, here's our guy and he's like sees this local improvement, but globally doesn't really compound. What he can I do? Cut costs cuz that's the only thing he see, and we
know how they cut costs, And we also know these companies were actually over staffed during COVID times and before. So, actually this guy can go can can can come to his stakeholders and say, "We're going to fire 20% of our people because we are so successful with AI." Isn't it that ironic? So, this is actually happening now. Um sadly. So, how do we stay relevant? In the
book we found this nice metaphor we we put and So, everybody has now a Ferrari. Every developer, every team, they rev the [snorts] engines, they are super fast, they have these consoles with AI agents. Fantastic, super fast, expensive too, but good looking. But if you look in the company every Ferrari is just staying in the traffic jam. Nothing is really moving. A very expensive traffic jam, I
tell We call it a Ferrari effect. people like this metaphor and use it everywhere now on the internet. So, a lot of companies just are bunch of expensive traffic jams with Ferraris burning oil and tokens. And don't want to be dramatic, but you might be gone, too. How do we keep them relevant? Please read. A wonderful quote. We have this discrepancy between the technology we have at
hand and the creepy old organization that were built on a 18th century models. Can you believe that? Adam Smith and stuff, Taylor. So, how do we keep them happy and busy and relevant? Not just busy. Keeping them busy is is easy, right? Just give them the same work. How do we keep them relevant? So, I have only like 7 minutes, so I need to go to the
solution space now. But, I apologize if that took so long, but I had to build up my case. Was it okay? Are you guys partially agreeing with that? Fantastic. Thank you. Um no, really. I need to check because your reality, your context actually is a king, might be different. If it's not applicable, you guys on the lucky side of things. You should be on the other part
of the conference now. All right. Not in the dark dark room. Alexy, focus. You have only 7 minutes. So, if there were these like dials you can turn right and left, I think there are two major dials we need to learn to work on. And it's not easy. It is org design topic. We need to turn the dial right on instead of utilizing people with a primary
expertise, database design, front end, back end, Java, whatever. I used to be a Java developer. Um And we need to allow those people to grow other skills. That's the only way actually to keep them relevant. Other skills, more skills. Do we have somebody who programs more than one language? Like well enough? Keep your hands. Two? Three? Languages? Four? Fluent in four? There's a hands. Five? Six? Anybody
with with with with with with with the hands still? Five? You were five? You can have my book. Seriously. But you know that this stuff. This book is is about how we multi learn. How we can we acquire more skills to stay relevant. Okay? That is makes sense? Another and another um dial is from keeping teams fixed on lanes for fast flow, allow them to switch and
follow the value and learn. This is possible. Before AI, we were building companies like I'm like consultant. So I've been doing that to help companies to help teams actually open up new repositories and open up new components and learn new things and contribute guess what to the whole product, not just to a little search lane. It is it was possible It should be possible now even better.
So in the book, don't have time for that at all. We introduce this simple map two by two. One dial is skill dial, other dial is scope of work dial. And we believe the only way to keep people relevant is to help them to go to go diagonal on the map. I just got a signal, I'll be kicked out from the stage in 5 minutes. Guess who's
going to kick out who. Um no, seriously, there's another speaker into be respective. Uh does this make sense? We need to unpin unpin teams from what they do and allow them to learn. And if it was not too dramatic, this is dramatic. AIs are coming from this corner of the map eating our simple single task jobs. And if we want to survive, we need to explore unknown
territories map of work. Now you can say, "Yeah, but but but it's too costly, too disruptive. Our people cannot know everything. Cognitive load." You know all these old good bad arguments about why we need to keep our as they are. please read. So, it's not about extreme, it's not about "No, we cannot learn because learning everything is hard." You don't need to learn everything, just need to
learn another thing adjacent to what you did before. It is already a good step into to direction of multi-learning. If you learn that other thing, open up another team's repository, are you getting worse at your previous skill? I don't think so. I'm a skier, pretty good skier. I can also snowboard. Doesn't make me a bad skier. I play I play bass guitar. I also learn some solo.
Doesn't make me a worse bass guitarist. I tell you, it makes me actually better. So, growing more skills actually is not loss of your expertise. It's actually way to keep it relevant. Good news here. Good news here is that this stuff is not new at all. This is 1986 Harvard Business Review, the paper. You can look it up. It's open. This is the first paper which introduced
the concept of multi-learning. And they actually studied very innovative by that time companies, and they proven those two Japanese guys, Takeuchi and Nonaka, proven that the only way to be innovative and to outpace, outlearn your competitors, is to allow your people to multi-learn. Not stay in one lane, multiple lanes, multiple expertise. We can do it. We can do it. By the way, this is the paper which
was actually introduced Scrum. And I know Scrum is super um hated now among developers, but that's not because Scrum is bad. It's just because your organization abused you with Scrum, I think. But but the ideas are pretty bright. So, multi-learning is old. It's there. We know how to do it. It's also now. It's my final um slide, more or less. Three pictures. What do you guys see
there? Who are those guys up there? Our grandmothers, grandfathers, right? And they were multi-learning, you can say. You had to be a hunter and a gatherer, a killer, and a lover, whatever. You know, you had to make up fire, you know, build and kill a mammoth at the same time. So, actually, our brain has developed it like jokes aside, for 6 million years, our brains were evolving
to become multi-learners. It is true. Look it up. Jiminy do research. And only only bloody 200 years, we were doing this local optimization. You needs to stay at a machine and just do one thing because some capitalist guy thinks it's better for him. [ __ ] them. So, where am I? I clicked something. No. No, go back. Excuse me. So, we need to go back to our strengths to
be multi-learners. It is the the time to do it now. And the good news, final good news of this talk, is that AI can actually not only just push us away, can can can also be used to help us to learn stuff. AI's not just doers, they also be great teachers if we know how to use them and why. Uh that's the way to do it. In
the book, uh we have nine principles, and and half of them are about multi-learning, how we actually embed it in the organization. So, if somebody is interested, find me in the whole between these rooms. And one last quote in the book is redesign, then AI. Don't spend AI in the existing org. First, rethink what you want to get better at, and then AI. Then you will get
at least 10 x improvement out of that. Thank you very much. >> Thank you, Alexy. What an energetic talk. I loved it. I love bias too much energy I can tell Always. Good. We need that. We need that. This is a long day. We need the energy to go on, right? Yeah, I love it. So, how long did you work on this book? Uh, am I still
Yeah, you're good. It was two and a half years and we were writing one book and then the third author came and we actually had to write completely another one. Because he said Uh, so this this is us Uh, writing the book. And the guy in the middle is Craig Larman and he wrote before this wonderful books. A lot of books. And he says, I will join
you because I like your [ __ ] but one one condition, it needs to be about AI. Not just management. I'm like, why? He he says, because AI is going to really change the way we think about organizations. And now that we wrote this book for two and a half years, I truly believe that it was a good idea. That's amazing. So, some of your co-authors are serial authors.
And you're in good company. That's amazing. jamming like crazy. That's the way to write a book, right? That's beautiful. so much. I need to be kicked out of here. thank you so much. Big round of applause for Alexy, please. >> Thanks.
