Using enterprise AI to improve the legibility of the value chain and the visibility of work outputs will enable better, more agile and autonomous management systems
As a practitioner leading one of those pioneering organizations, built around the principle of radically distributed autonomy, where there are no managers, anyone can make any decision (in a structured way, of course), I don't subscribe to the view that it automatically translates to better AI adoption.
In progressive organizations, the hard part is not the tooling that would broker between different people/teams. It's how people act and why. To stick with the Disco example, their internal-currency brokerage system is just a technical solution used to facilitate an entirely different collaboration system. The collaboration system (how people act) is the advantage, not a virtual currency marketplace.
In fact, Morningstar, Favi, Buurtzorg, Nucor, etc., don't seem to be poster children of the AI revolution. I see it more as a correlation between a specific industry a company operates in, rather than their organizational models.
In our—Lunar Logic—case, we are on the AI bandwagon, but not *because* of our organizational model. It's because of the business we're in. In fact, for all our "fancy" pioneering stuff, we use such old-school tooling as sticky notes, shared spreadsheets, announcement boards, and communication.
I'll be more than happy to change them when it's advantageous. So far, it has not been the case.
Hi Pawel and thanks for taking the time to share your valuable experience in a progressive company. Of course, the most important added value in most cases comes from people - how they connect, collaborate and create value together - and that is not something anybody should seek to automate away.
I don't believe that autonomous or agile management is the factor that predicts AI adoption, but rather a service-oriented rather than process-oriented approach, where underlying support services and workflows are modular in a way that can later become automated where appropriate. Larger firms operating manual top-down management often struggle with this.
For me, the most interesting architecture for an organisational operating system is a shared platform of basic services and functions that can orchestrate and automate the simple, boring stuff to allow more autonomous and agile team structures working on top do the platform to focus on human ways of working without bureaucracy getting in the way. But many companies still struggle to define, write down or digitise these basic processes, and so people waste too much time in the 'system' rather than working.
As a software development group you probably take much of this logic for granted, with shared libraries, project documentation and perhaps also deployment and repo tools that automate some of the invisible background activities. But for less tech-confident larger firms with complex management structures, there is not so much awareness of how to automate the back-end in order to free people up for real work on the front-end. Ultimately, I think organisations are like a form of human software, which means we need to think about architecture, platforms, maintenance, etc. if we want them to keep evolving and improving, and also if we want the 'system' not to get in the way of human value creation and collaboration.
Organizations which design systems (in the broad sense used here) are constrained to produce designs which are copies of the communication structures of these organizations.
That, however, suggests the opposite relationship. Decentralized organizations might be better suited to adopt AI-led architecture, which, as we look at it today, seems to be decentralized/autonomous.
The qualifier Conway used remains in place "organizations which design systems." Many progressive/pioneering organizations don't design systems, at least not as their product. That's why they won't be poster kids of AI adoption.
Anyway, you're right with the fact that operating in the context of software gives us a native understanding of some decentralized architectures. I wouldn't have considered that (which probably makes it a canonical privilege).
I would add a separate comment on designs that promote radical "agentism" for every single individual (like Disco). I know a story of such a company. They ended up in radical individualism, where every part (every individual) was optimizing for a local optimum. As we know from systems thinking, the system is more than the sum of its parts, and local optimizations may be detrimental to the performance of the whole system.
I share your enthusiasm. AI is the crucial missing ingredient to replace outdated management models. Exciting times ahead.
As a practitioner leading one of those pioneering organizations, built around the principle of radically distributed autonomy, where there are no managers, anyone can make any decision (in a structured way, of course), I don't subscribe to the view that it automatically translates to better AI adoption.
In progressive organizations, the hard part is not the tooling that would broker between different people/teams. It's how people act and why. To stick with the Disco example, their internal-currency brokerage system is just a technical solution used to facilitate an entirely different collaboration system. The collaboration system (how people act) is the advantage, not a virtual currency marketplace.
In fact, Morningstar, Favi, Buurtzorg, Nucor, etc., don't seem to be poster children of the AI revolution. I see it more as a correlation between a specific industry a company operates in, rather than their organizational models.
In our—Lunar Logic—case, we are on the AI bandwagon, but not *because* of our organizational model. It's because of the business we're in. In fact, for all our "fancy" pioneering stuff, we use such old-school tooling as sticky notes, shared spreadsheets, announcement boards, and communication.
I'll be more than happy to change them when it's advantageous. So far, it has not been the case.
Hi Pawel and thanks for taking the time to share your valuable experience in a progressive company. Of course, the most important added value in most cases comes from people - how they connect, collaborate and create value together - and that is not something anybody should seek to automate away.
I don't believe that autonomous or agile management is the factor that predicts AI adoption, but rather a service-oriented rather than process-oriented approach, where underlying support services and workflows are modular in a way that can later become automated where appropriate. Larger firms operating manual top-down management often struggle with this.
For me, the most interesting architecture for an organisational operating system is a shared platform of basic services and functions that can orchestrate and automate the simple, boring stuff to allow more autonomous and agile team structures working on top do the platform to focus on human ways of working without bureaucracy getting in the way. But many companies still struggle to define, write down or digitise these basic processes, and so people waste too much time in the 'system' rather than working.
As a software development group you probably take much of this logic for granted, with shared libraries, project documentation and perhaps also deployment and repo tools that automate some of the invisible background activities. But for less tech-confident larger firms with complex management structures, there is not so much awareness of how to automate the back-end in order to free people up for real work on the front-end. Ultimately, I think organisations are like a form of human software, which means we need to think about architecture, platforms, maintenance, etc. if we want them to keep evolving and improving, and also if we want the 'system' not to get in the way of human value creation and collaboration.
Conway's Law comes to mind:
Organizations which design systems (in the broad sense used here) are constrained to produce designs which are copies of the communication structures of these organizations.
That, however, suggests the opposite relationship. Decentralized organizations might be better suited to adopt AI-led architecture, which, as we look at it today, seems to be decentralized/autonomous.
The qualifier Conway used remains in place "organizations which design systems." Many progressive/pioneering organizations don't design systems, at least not as their product. That's why they won't be poster kids of AI adoption.
Anyway, you're right with the fact that operating in the context of software gives us a native understanding of some decentralized architectures. I wouldn't have considered that (which probably makes it a canonical privilege).
I would add a separate comment on designs that promote radical "agentism" for every single individual (like Disco). I know a story of such a company. They ended up in radical individualism, where every part (every individual) was optimizing for a local optimum. As we know from systems thinking, the system is more than the sum of its parts, and local optimizations may be detrimental to the performance of the whole system.