You can't retrofit your way to AI-native. The companies getting results started by changing their beliefs about what management is for. Here's what's changing.
This is the piece most AI roadmaps skip. Everyone budgets for the tools, almost nobody budgets for rewiring how decisions move between people. That 'agreed, but guessing' zone in your diagram is where I've seen the most value quietly leak.
The colliding-work example caught my attention, because it happened in a company where almost everything was already visible, which suggests the gap was decision ownership. From my experience, when two capable people pick up the same problem, it usually traces back to nobody having said who holds the call on that piece of work, and faster tools simply bring that moment forward. The DRI role goes some way towards fixing this, though I wonder how it holds up when several DRIs’ outcomes overlap and each has the authority to pull people in. Have you seen teams use the confirmed record to settle those boundaries, or does that still need to be worked out in the room first?
I think the collisions are inevitable. And if we assign people ownership over more complex problems it’s more likely to occur. Having better coordination infrastructure can help spot those conflicts and help put teams in touch to pool resources and ideas, other times the competition may be the right way to go. As long as it’s seen as a healthy competition and people are aware of it. What stresses people is the surprise and lack of clarity on the rules of engagement. If everyone knows: “hey, there may be other teams tackling the same problem, let’s make sure we make our work as visible as possible so we find each other in the network so we can collaborate.“
The Block, Lovable and Every examples are the companies describing their own results, so I would read the pattern as a hypothesis. Survivors write the field notes.
Here is the part I would test. Every agent gets a named human agent parent. You note that review cost lands on experts who did not write the draft. An agent parent is the same pattern with a title. What happens to the agent when its parent changes teams or leaves?
The coordination bottleneck is the strongest section. A company brain makes information cheap and leaves commitment expensive. Does the layer Afoot is building change who is accountable, or only who gets the record? Happy to connect, let's talk more about it.
The truth is everyone is experiencing the role of manager whether they want to or not. Whether that is manager of people or agents, but increasingly that will be agent. So like a person, when the manager leaves, the agent gets a new manager.
Afoot builds and maintains the shared record, the commitment is made by a person to other people. Afoot just makes that commitment (as well as decisions, etc) visible to those beyond the room and has agents that help the broader network coordinate effectively.
The shift from "managers who coordinate people" to "managers who coordinate people + AI systems" is the real story here. The companies getting this right aren't just adding AI tools, they're fundamentally rethinking what a management layer does. Curious how this plays out in industries where middle management was already thin.
"Production has fallen below the cost of coordinating." A DeepMind panel at INSEAD said nearly the same thing: knowledge moves more freely but still travels along the hierarchy. I now treat capacity and coordination as the two halves of decision quality. Curious whether this matches what you see: https://creatism.substack.com/p/the-ai-reflection-framework-goes
So, how do the "experts" of the future gain their expertise? Judgement is built by learning, making mistakes, and having standards. Who is the creator of the standards for newcos? Their Supreme Founder? There's no way to know whether the What and How are correct without a Who to evaluate and update them. Let's see how it goes when the rest of the experts holding these companies together leave the workforce.
By replacing traditional supervision with AI-powered decision-making, automation, and outcome-based performance, AI-native businesses are rebuilding management. Managers can spend more time on strategy, creativity, and team development rather than on tasks that get done over and over again. The real advantage comes from combining AI capabilities with human judgment, leadership, and collaboration.
The winners in the electrification example "redesigned around" the new technology, and later she names "the belief that a company is its org chart." In my essay Use the f*cking lanes I wrote "Question every requirement, and make every requirement carry the name of the person who asked for it. Then delete." That is redesigning at the level of one requirement, before any agent gets added.
Thanks for this article. Most of the Large organisations are thinking ai as “medicine”. Without thinking on their curent issues about alignement judgement and data/knowledge. It will be very costly if they rely on ai with out taking account their current state. AI is complete advantage for newly established organisations, I am not sure about large scale organisations? What are your thoughts?
The electrification parallel lands. We keep bolting AI onto the same approval chain and then wonder why throughput barely moves. Where have you seen a team actually redesign a handoff instead of just speeding it up?
Excellent insights, clear perspective. Love this piece. Just one minor addition: The title suggests that it takes AI native orgs to pull this off. Not quite: It can be easier for them, yes. But others can also remodel themselves.
Thank you. Yes. I would use it broadly as any company willing to step back and rethink how they do everything. To break free of established beliefs on how roles are defined, goals are set, people are managed, what the work day looks like altogether. I’m not sure AI native have the advantage, the leaders there often come with preconceptions as well. Established firms with already well established management practices and good culture of experimentation and self leadership would likely get ahead faster.
This is the piece most AI roadmaps skip. Everyone budgets for the tools, almost nobody budgets for rewiring how decisions move between people. That 'agreed, but guessing' zone in your diagram is where I've seen the most value quietly leak.
The colliding-work example caught my attention, because it happened in a company where almost everything was already visible, which suggests the gap was decision ownership. From my experience, when two capable people pick up the same problem, it usually traces back to nobody having said who holds the call on that piece of work, and faster tools simply bring that moment forward. The DRI role goes some way towards fixing this, though I wonder how it holds up when several DRIs’ outcomes overlap and each has the authority to pull people in. Have you seen teams use the confirmed record to settle those boundaries, or does that still need to be worked out in the room first?
I think the collisions are inevitable. And if we assign people ownership over more complex problems it’s more likely to occur. Having better coordination infrastructure can help spot those conflicts and help put teams in touch to pool resources and ideas, other times the competition may be the right way to go. As long as it’s seen as a healthy competition and people are aware of it. What stresses people is the surprise and lack of clarity on the rules of engagement. If everyone knows: “hey, there may be other teams tackling the same problem, let’s make sure we make our work as visible as possible so we find each other in the network so we can collaborate.“
The Block, Lovable and Every examples are the companies describing their own results, so I would read the pattern as a hypothesis. Survivors write the field notes.
Here is the part I would test. Every agent gets a named human agent parent. You note that review cost lands on experts who did not write the draft. An agent parent is the same pattern with a title. What happens to the agent when its parent changes teams or leaves?
The coordination bottleneck is the strongest section. A company brain makes information cheap and leaves commitment expensive. Does the layer Afoot is building change who is accountable, or only who gets the record? Happy to connect, let's talk more about it.
The truth is everyone is experiencing the role of manager whether they want to or not. Whether that is manager of people or agents, but increasingly that will be agent. So like a person, when the manager leaves, the agent gets a new manager.
Afoot builds and maintains the shared record, the commitment is made by a person to other people. Afoot just makes that commitment (as well as decisions, etc) visible to those beyond the room and has agents that help the broader network coordinate effectively.
The shift from "managers who coordinate people" to "managers who coordinate people + AI systems" is the real story here. The companies getting this right aren't just adding AI tools, they're fundamentally rethinking what a management layer does. Curious how this plays out in industries where middle management was already thin.
the factory electrification comparison made the whole shift click for me, rebuilding the layout instead of swapping the motor.
https://aipeacebuilding.substack.com/p/a-roadmap-for-diplomacy-with-ais?r=ql6co&utm_medium=ios
"Production has fallen below the cost of coordinating." A DeepMind panel at INSEAD said nearly the same thing: knowledge moves more freely but still travels along the hierarchy. I now treat capacity and coordination as the two halves of decision quality. Curious whether this matches what you see: https://creatism.substack.com/p/the-ai-reflection-framework-goes
So, how do the "experts" of the future gain their expertise? Judgement is built by learning, making mistakes, and having standards. Who is the creator of the standards for newcos? Their Supreme Founder? There's no way to know whether the What and How are correct without a Who to evaluate and update them. Let's see how it goes when the rest of the experts holding these companies together leave the workforce.
Directly Responsible Individual (DRI) == Product Owner
Player-Coach == Scrum Master
Individual Contributor (IC) == Developer
Welcome to 1995
FUCK YEAH! The LARGEST office construction company in Moscow fired me after I came to them with an AI replacement solution.
They have 4000 employees. I told them the entire company can be automated with 7 Mac Studios (M3 Ultra 512gb ram)
I came to the C-Suite EMPLOYEES as well as Economic Safety officers, they still didn’t wanna listen. Then due to “lack of performance” was fired. Hehe
Even though they asked me for a solution. They didn’t think I’d build it in four months.
By replacing traditional supervision with AI-powered decision-making, automation, and outcome-based performance, AI-native businesses are rebuilding management. Managers can spend more time on strategy, creativity, and team development rather than on tasks that get done over and over again. The real advantage comes from combining AI capabilities with human judgment, leadership, and collaboration.
The winners in the electrification example "redesigned around" the new technology, and later she names "the belief that a company is its org chart." In my essay Use the f*cking lanes I wrote "Question every requirement, and make every requirement carry the name of the person who asked for it. Then delete." That is redesigning at the level of one requirement, before any agent gets added.
Since the arrival of ChatGPT I found it baffling that everyone started with automation of existing processes. This seemed premature.
Automation should be the last step after rethinking the entire process and questioning every requirement.
Most of the coordination load I see is self-inflicted in exactly that way.
Please do link to your article. Would love to read it.
Here it is: https://usethefuckinglanes.substack.com/p/use-the-fcking-lanes
Thanks for reading.
We have been on thus oath for four years now. It works. Not crazy at all.
Any path working particularly well?
Thanks for this article. Most of the Large organisations are thinking ai as “medicine”. Without thinking on their curent issues about alignement judgement and data/knowledge. It will be very costly if they rely on ai with out taking account their current state. AI is complete advantage for newly established organisations, I am not sure about large scale organisations? What are your thoughts?
The electrification parallel lands. We keep bolting AI onto the same approval chain and then wonder why throughput barely moves. Where have you seen a team actually redesign a handoff instead of just speeding it up?
Excellent insights, clear perspective. Love this piece. Just one minor addition: The title suggests that it takes AI native orgs to pull this off. Not quite: It can be easier for them, yes. But others can also remodel themselves.
Thank you. Yes. I would use it broadly as any company willing to step back and rethink how they do everything. To break free of established beliefs on how roles are defined, goals are set, people are managed, what the work day looks like altogether. I’m not sure AI native have the advantage, the leaders there often come with preconceptions as well. Established firms with already well established management practices and good culture of experimentation and self leadership would likely get ahead faster.