Your Shadow Network is Getting Exposed This Year
AI
Management
Your org chart is a map, and maps lie by omission.
It shows you who has the authority to make a call. But it doesn’t show you who people actually go to when something’s actually broken.
![](data:image/svg+xml,<svg display="block" id="565267359" role="presentation" viewBox="0 0 3000 1376" xmlns="http://www.w3.org/2000/svg"/>)
THE PROBLEM
Organizational Network Analysis, which maps real information flow and advice-seeking instead of formal titles, has been proving this for years: run it against any org chart and the divergence is consistent. A shadow network of trust and competence does the real coordinating work, cutting across departments and titles. The chart was built to answer a different question — who’s accountable — not the one that determines whether work actually gets done, which is who’s competent.
Once you see that, the fix looks obvious: cut the layers that aren’t doing anything but slowing down people who already know what to do. Companies have tried exactly that for over a decade — delayering, flattening, and in its most ambitious form, frameworks like Holacracy, which replaced job titles and managers with self-governing circles. Several of these experiments ended in expensive, public failure.
Hierarchy is, underneath the authority dressing, an information-routing system: a way of moving context from the people doing the work to the people deciding what to do about it, and moving authority back down so that decision can be executed. What killed Holacracy wasn’t the principle — the principle was sound. But in order to make it work, companies tried to replace that routing system by hand, through meetings and governance rounds and objection-handling protocols, instead of with something built to run it at the speed the work actually required.
In 2013, that something didn’t exist. But it exists now. AI agents embedded in the tools a company already runs — the comms stack, the project tracker, the CRM, the codebase — do the job a middle layer of management used to do: filter, translate, and re-route context so the right person has what they need to decide, without a human standing in the pipe. The org chart was a solution to an information problem, but that solution was formed when the saying still went “this meeting could have been a telegram.”
The reason hierarchy takes the shape it does was formalized in 1933, when V.A. Graicunas showed that the number of relationships a manager must coordinate grows geometrically with every direct report — humans can effectively manage 3 to 8 people before information decay turns operations into chaos. Authority flowed downward through the resulting layers so accountability could flow upward without collapsing under its own noise. Strip out the routing function and authority doesn’t disappear with it. Neither does accountability — it just stops having anywhere obvious to live.
That's the actual stakes of automating the plumbing, and it's one of the puzzles that we, at Evonomix, are now tackling head-on.
COMPOUNDING COSTS
The gap between chart and reality isn’t trivia. Map it, and it tends to surface in the same place: individuals two or three levels below VP titles — “hidden influencers” — turn out to be the de facto information brokers, knowledge routers, and trust anchors for entire divisions, independent of what the chart says about them.
“So what?” you’d think. “As long as things are running, who cares whether the formal org chart is perfectly optimized to reflect reality?” And you’d be partly right. But what you might be missing are several hidden costs that compound.
Performance reviews that miss actual contribution damage your cultural climate and lead to attrition.
Single resignations damage performance and engagement for entire teams.
Your most important asset — your data and its derived organizational knowledge — is scattered, borderline unretrievable, and cannot feed your growth.
These effects are rarely touched upon in an executive report. And when they do surface, the org chart can’t tell you who’s responsible. Because no one is. The issues are structural, built into the system from the start.
Not so fast, though
Here’s what “expensive, public failure” actually looked like for those who tried to flatten the org chart without replacing the function.
By 2015, Zappos CEO Tony Hsieh — two years into the company’s Holacracy rollout — ran out of patience with slow adoption. He gave employees a choice: commit fully to the new system, or take a severance package and leave.
18% took the money. 11% more left later, without one.
A few years after that, the system was quietly retired, though it was never officially declared dead.
At Medium, Twitter co-founder Ev Williams adopted Holacracy in 2012 and walked it back by 2016. The official line was that “it wasn’t the right fit at our current stage.” Internal reporting told a more specific story: governance overhead was consuming time that should have gone to product, and nobody could confidently say who was actually allowed to make a call.
Both companies had removed the org chart. Neither had built a working replacement for what it had been doing, however badly.
Not every attempt to make self-management work at scale ended in disaster. And successful stories carry insights that become more salient the more AI agents revolutionize the workplace.
SUCCESS STORIES
Haier
Haier is a Chinese home appliances manufacturer with 80,000 employees. In 2012, CEO Zhang Ruimin dismantled the traditional management structure and reorganized the entire company into roughly 4,000 micro-enterprises — internal startups, each with its own P&L, its own strategy, and the freedom to hire, partner, and compete. The model is called RenDanHeYi: the integration of people and goals.
Accountability no longer flows upward to a VP who reports to a SVP. It flows outward to the customer. Each micro-enterprise lives or dies by the market it serves. That is a more brutal accountability mechanism than any performance review cycle — and it’s also a structure that scales beautifully once each micro-enterprise can run its own market read and coordination with an AI agent instead of a back office.
Buurtzorg
In 2006, a Dutch nurse named Jos de Blok founded Buurtzorg on a single organizing principle: give small teams of nurses everything they need to run themselves, and get out of the way. Today, Buurtzorg employs around 15,000 nurses organized into self-managing teams of 10 to 12. There is no middle management layer. The back office supporting all 15,000 nurses is around 50 people. Overhead runs at roughly 8% — against an industry average of 25%.
Buurtzorg built that 50-person back office by hand, over a decade, because de Blok wanted it and had the discipline to enforce it. Most companies don’t have a founder willing to fight that fight. AI-supported operations gets you a version of that ratio without the decade or the discipline — which is either the best argument for doing this deliberately before it’s forced on you, or the reason most companies will end up with Buurtzorg’s overhead numbers without ever making Buurtzorg’s choice.
Basecamp and GitLab
Neither company reorganized into circles or micro-enterprises. What they did was quieter and more transferable: they systematically replaced the functions that management typically performs with documented systems anyone in the company can access and use. GitLab’s company handbook runs to thousands of pages and is publicly available — covering everything from how to run a meeting to how to make a hiring decision. GitLab went public in 2021 with over 1,300 people, fully remote, no offices, without adding the management layers that typically accompany that growth.
The lesson is not that they had no management functions. They had all of them. They had just moved those functions out of people’s heads and into systems. GitLab’s handbook is what that move looks like when a company builds it manually, page by page, over years. An AI knowledge layer is what that move looks like when a company doesn’t have years — the same function, compressed from a decade of deliberate documentation into whatever your agents can index this quarter.
Three companies. Three industries. Three different implementations. Read together: a checklist for what has to be true before you let AI take over the routing function your hierarchy currently performs.
the checklist
Accountability must be anchored to something external. In all three cases, employees are not accountable to a manager but to a market, a patient, a shipped product. An AI agent can move context instantly. It cannot manufacture a market to be accountable to. If accountability in your org currently lives inside the hierarchy rather than outside it — tied to a manager’s approval rather than a customer’s outcome — automating the routing layer won’t fix that. It will just remove the last thing standing between your teams and no accountability at all.
The people doing the work must already know what good work looks like. Buurtzorg’s nurses are trained professionals. GitLab engineers know what good code looks like. Autonomy handed to people without a strong craft or professional foundation does not produce self-management — it produces anxiety and drift. Same with AI: an agent can hand someone a decision faster than a manager could. It cannot hand them the judgment to make it. If that judgment isn’t already distributed in your organization, faster access to context just means faster, more confident wrong answers.
Management functions must be replaced by systems, not eliminated. Managers do real work — distributing context, resolving resource conflicts, carrying institutional knowledge, making and recording decisions. In every successful case, these functions still happen via a P&L structure, a handbook, a software platform, and a professional training framework. AI is a candidate system for this — arguably the most powerful one yet available. But it’s a candidate, not a guarantee. Bolting an agent onto a broken structure automates the dysfunction; it doesn’t remove it.
If the primary function of management layers has always been information routing — and AI can now handle that by excellence — then management-as-a-career-path is a historical artifact.
READY OR NOT
What survives is accountability. But accountability, separated from its coordination wrapper, doesn’t distribute the way a hierarchy does. It concentrates where the expertise is. And here is the uncomfortable implication most organizations haven’t yet sat with: a significant portion of what your managers currently do are professional functions — judgment calls, quality standards, domain decisions — that got delegated upward because that’s where the authority was, not because that’s where the knowledge was.
AI will not solve that. It will expose it. Once an agent can hand a room everything a meeting used to take an hour to establish, the only thing left worth arguing about is who in that room actually knows what to do with it. An org chart was never built to answer that.
The organizations that will navigate this well are not necessarily the fastest AI adopters. They’re the ones that already know where their real network is — where work actually gets decided, who the actual knowledge anchors are, and which of their management functions can be systematized versus which require genuine expertise. Most companies haven’t run that audit. AI will run it for them, whether they’re ready or not.
Here’s the part that should actually keep you up at night: the companies in the success stories above ran this audit voluntarily, on their own timeline, over years, with a founder or CEO willing to absorb the transition cost. You don’t get that version. Every SaaS product your company already pays for — the CRM, the ticketing system, the codebase, the doc tool — is getting an agent bolted onto it this year, on your vendor’s release schedule, not yours. The shadow network you didn’t map is about to become visible.
The only question left is whether you’re the one reading the results, or the last one to find out what they said.
Featured Case Studies
Featured Case Studies
Your Shadow Network is Getting Exposed This Year
AI
Management
Your org chart is a map, and maps lie by omission.
It shows you who has the authority to make a call. But it doesn’t show you who people actually go to when something’s actually broken.
THE PROBLEM
Organizational Network Analysis, which maps real information flow and advice-seeking instead of formal titles, has been proving this for years: run it against any org chart and the divergence is consistent. A shadow network of trust and competence does the real coordinating work, cutting across departments and titles. The chart was built to answer a different question — who’s accountable — not the one that determines whether work actually gets done, which is who’s competent.
Once you see that, the fix looks obvious: cut the layers that aren’t doing anything but slowing down people who already know what to do. Companies have tried exactly that for over a decade — delayering, flattening, and in its most ambitious form, frameworks like Holacracy, which replaced job titles and managers with self-governing circles. Several of these experiments ended in expensive, public failure.
Hierarchy is, underneath the authority dressing, an information-routing system: a way of moving context from the people doing the work to the people deciding what to do about it, and moving authority back down so that decision can be executed. What killed Holacracy wasn’t the principle — the principle was sound. But in order to make it work, companies tried to replace that routing system by hand, through meetings and governance rounds and objection-handling protocols, instead of with something built to run it at the speed the work actually required.
In 2013, that something didn’t exist. But it exists now. AI agents embedded in the tools a company already runs — the comms stack, the project tracker, the CRM, the codebase — do the job a middle layer of management used to do: filter, translate, and re-route context so the right person has what they need to decide, without a human standing in the pipe. The org chart was a solution to an information problem, but that solution was formed when the saying still went “this meeting could have been a telegram.”
The reason hierarchy takes the shape it does was formalized in 1933, when V.A. Graicunas showed that the number of relationships a manager must coordinate grows geometrically with every direct report — humans can effectively manage 3 to 8 people before information decay turns operations into chaos. Authority flowed downward through the resulting layers so accountability could flow upward without collapsing under its own noise. Strip out the routing function and authority doesn’t disappear with it. Neither does accountability — it just stops having anywhere obvious to live.
That's the actual stakes of automating the plumbing, and it's one of the puzzles that we, at Evonomix, are now tackling head-on.
COMPOUNDING COSTS
The gap between chart and reality isn’t trivia. Map it, and it tends to surface in the same place: individuals two or three levels below VP titles — “hidden influencers” — turn out to be the de facto information brokers, knowledge routers, and trust anchors for entire divisions, independent of what the chart says about them.
“So what?” you’d think. “As long as things are running, who cares whether the formal org chart is perfectly optimized to reflect reality?” And you’d be partly right. But what you might be missing are several hidden costs that compound.
Performance reviews that miss actual contribution damage your cultural climate and lead to attrition.
Single resignations damage performance and engagement for entire teams.
Your most important asset — your data and its derived organizational knowledge — is scattered, borderline unretrievable, and cannot feed your growth.
These effects are rarely touched upon in an executive report. And when they do surface, the org chart can’t tell you who’s responsible. Because no one is. The issues are structural, built into the system from the start.
Not so fast, though
Here’s what “expensive, public failure” actually looked like for those who tried to flatten the org chart without replacing the function.
By 2015, Zappos CEO Tony Hsieh — two years into the company’s Holacracy rollout — ran out of patience with slow adoption. He gave employees a choice: commit fully to the new system, or take a severance package and leave.
18% took the money. 11% more left later, without one.
A few years after that, the system was quietly retired, though it was never officially declared dead.
At Medium, Twitter co-founder Ev Williams adopted Holacracy in 2012 and walked it back by 2016. The official line was that “it wasn’t the right fit at our current stage.” Internal reporting told a more specific story: governance overhead was consuming time that should have gone to product, and nobody could confidently say who was actually allowed to make a call.
Both companies had removed the org chart. Neither had built a working replacement for what it had been doing, however badly.
Not every attempt to make self-management work at scale ended in disaster. And successful stories carry insights that become more salient the more AI agents revolutionize the workplace.
SUCCESS STORIES
Haier
Haier is a Chinese home appliances manufacturer with 80,000 employees. In 2012, CEO Zhang Ruimin dismantled the traditional management structure and reorganized the entire company into roughly 4,000 micro-enterprises — internal startups, each with its own P&L, its own strategy, and the freedom to hire, partner, and compete. The model is called RenDanHeYi: the integration of people and goals.
Accountability no longer flows upward to a VP who reports to a SVP. It flows outward to the customer. Each micro-enterprise lives or dies by the market it serves. That is a more brutal accountability mechanism than any performance review cycle — and it’s also a structure that scales beautifully once each micro-enterprise can run its own market read and coordination with an AI agent instead of a back office.
Buurtzorg
In 2006, a Dutch nurse named Jos de Blok founded Buurtzorg on a single organizing principle: give small teams of nurses everything they need to run themselves, and get out of the way. Today, Buurtzorg employs around 15,000 nurses organized into self-managing teams of 10 to 12. There is no middle management layer. The back office supporting all 15,000 nurses is around 50 people. Overhead runs at roughly 8% — against an industry average of 25%.
Buurtzorg built that 50-person back office by hand, over a decade, because de Blok wanted it and had the discipline to enforce it. Most companies don’t have a founder willing to fight that fight. AI-supported operations gets you a version of that ratio without the decade or the discipline — which is either the best argument for doing this deliberately before it’s forced on you, or the reason most companies will end up with Buurtzorg’s overhead numbers without ever making Buurtzorg’s choice.
Basecamp and GitLab
Neither company reorganized into circles or micro-enterprises. What they did was quieter and more transferable: they systematically replaced the functions that management typically performs with documented systems anyone in the company can access and use. GitLab’s company handbook runs to thousands of pages and is publicly available — covering everything from how to run a meeting to how to make a hiring decision. GitLab went public in 2021 with over 1,300 people, fully remote, no offices, without adding the management layers that typically accompany that growth.
The lesson is not that they had no management functions. They had all of them. They had just moved those functions out of people’s heads and into systems. GitLab’s handbook is what that move looks like when a company builds it manually, page by page, over years. An AI knowledge layer is what that move looks like when a company doesn’t have years — the same function, compressed from a decade of deliberate documentation into whatever your agents can index this quarter.
Three companies. Three industries. Three different implementations. Read together: a checklist for what has to be true before you let AI take over the routing function your hierarchy currently performs.
the checklist
Accountability must be anchored to something external. In all three cases, employees are not accountable to a manager but to a market, a patient, a shipped product. An AI agent can move context instantly. It cannot manufacture a market to be accountable to. If accountability in your org currently lives inside the hierarchy rather than outside it — tied to a manager’s approval rather than a customer’s outcome — automating the routing layer won’t fix that. It will just remove the last thing standing between your teams and no accountability at all.
The people doing the work must already know what good work looks like. Buurtzorg’s nurses are trained professionals. GitLab engineers know what good code looks like. Autonomy handed to people without a strong craft or professional foundation does not produce self-management — it produces anxiety and drift. Same with AI: an agent can hand someone a decision faster than a manager could. It cannot hand them the judgment to make it. If that judgment isn’t already distributed in your organization, faster access to context just means faster, more confident wrong answers.
Management functions must be replaced by systems, not eliminated. Managers do real work — distributing context, resolving resource conflicts, carrying institutional knowledge, making and recording decisions. In every successful case, these functions still happen via a P&L structure, a handbook, a software platform, and a professional training framework. AI is a candidate system for this — arguably the most powerful one yet available. But it’s a candidate, not a guarantee. Bolting an agent onto a broken structure automates the dysfunction; it doesn’t remove it.
If the primary function of management layers has always been information routing — and AI can now handle that by excellence — then management-as-a-career-path is a historical artifact.
READY OR NOT
What survives is accountability. But accountability, separated from its coordination wrapper, doesn’t distribute the way a hierarchy does. It concentrates where the expertise is. And here is the uncomfortable implication most organizations haven’t yet sat with: a significant portion of what your managers currently do are professional functions — judgment calls, quality standards, domain decisions — that got delegated upward because that’s where the authority was, not because that’s where the knowledge was.
AI will not solve that. It will expose it. Once an agent can hand a room everything a meeting used to take an hour to establish, the only thing left worth arguing about is who in that room actually knows what to do with it. An org chart was never built to answer that.
The organizations that will navigate this well are not necessarily the fastest AI adopters. They’re the ones that already know where their real network is — where work actually gets decided, who the actual knowledge anchors are, and which of their management functions can be systematized versus which require genuine expertise. Most companies haven’t run that audit. AI will run it for them, whether they’re ready or not.
Here’s the part that should actually keep you up at night: the companies in the success stories above ran this audit voluntarily, on their own timeline, over years, with a founder or CEO willing to absorb the transition cost. You don’t get that version. Every SaaS product your company already pays for — the CRM, the ticketing system, the codebase, the doc tool — is getting an agent bolted onto it this year, on your vendor’s release schedule, not yours. The shadow network you didn’t map is about to become visible.
The only question left is whether you’re the one reading the results, or the last one to find out what they said.
Featured Case Studies
Featured Case Studies
Your Shadow Network is Getting Exposed This Year
AI
Management
Your org chart is a map, and maps lie by omission.
It shows you who has the authority to make a call. But it doesn’t show you who people actually go to when something’s actually broken.
THE PROBLEM
Organizational Network Analysis, which maps real information flow and advice-seeking instead of formal titles, has been proving this for years: run it against any org chart and the divergence is consistent. A shadow network of trust and competence does the real coordinating work, cutting across departments and titles. The chart was built to answer a different question — who’s accountable — not the one that determines whether work actually gets done, which is who’s competent.
Once you see that, the fix looks obvious: cut the layers that aren’t doing anything but slowing down people who already know what to do. Companies have tried exactly that for over a decade — delayering, flattening, and in its most ambitious form, frameworks like Holacracy, which replaced job titles and managers with self-governing circles. Several of these experiments ended in expensive, public failure.
Hierarchy is, underneath the authority dressing, an information-routing system: a way of moving context from the people doing the work to the people deciding what to do about it, and moving authority back down so that decision can be executed. What killed Holacracy wasn’t the principle — the principle was sound. But in order to make it work, companies tried to replace that routing system by hand, through meetings and governance rounds and objection-handling protocols, instead of with something built to run it at the speed the work actually required.
In 2013, that something didn’t exist. But it exists now. AI agents embedded in the tools a company already runs — the comms stack, the project tracker, the CRM, the codebase — do the job a middle layer of management used to do: filter, translate, and re-route context so the right person has what they need to decide, without a human standing in the pipe. The org chart was a solution to an information problem, but that solution was formed when the saying still went “this meeting could have been a telegram.”
The reason hierarchy takes the shape it does was formalized in 1933, when V.A. Graicunas showed that the number of relationships a manager must coordinate grows geometrically with every direct report — humans can effectively manage 3 to 8 people before information decay turns operations into chaos. Authority flowed downward through the resulting layers so accountability could flow upward without collapsing under its own noise. Strip out the routing function and authority doesn’t disappear with it. Neither does accountability — it just stops having anywhere obvious to live.
That's the actual stakes of automating the plumbing, and it's one of the puzzles that we, at Evonomix, are now tackling head-on.
COMPOUNDING COSTS
The gap between chart and reality isn’t trivia. Map it, and it tends to surface in the same place: individuals two or three levels below VP titles — “hidden influencers” — turn out to be the de facto information brokers, knowledge routers, and trust anchors for entire divisions, independent of what the chart says about them.
“So what?” you’d think. “As long as things are running, who cares whether the formal org chart is perfectly optimized to reflect reality?” And you’d be partly right. But what you might be missing are several hidden costs that compound.
Performance reviews that miss actual contribution damage your cultural climate and lead to attrition.
Single resignations damage performance and engagement for entire teams.
Your most important asset — your data and its derived organizational knowledge — is scattered, borderline unretrievable, and cannot feed your growth.
These effects are rarely touched upon in an executive report. And when they do surface, the org chart can’t tell you who’s responsible. Because no one is. The issues are structural, built into the system from the start.
Not so fast, though
Here’s what “expensive, public failure” actually looked like for those who tried to flatten the org chart without replacing the function.
By 2015, Zappos CEO Tony Hsieh — two years into the company’s Holacracy rollout — ran out of patience with slow adoption. He gave employees a choice: commit fully to the new system, or take a severance package and leave.
18% took the money. 11% more left later, without one.
A few years after that, the system was quietly retired, though it was never officially declared dead.
At Medium, Twitter co-founder Ev Williams adopted Holacracy in 2012 and walked it back by 2016. The official line was that “it wasn’t the right fit at our current stage.” Internal reporting told a more specific story: governance overhead was consuming time that should have gone to product, and nobody could confidently say who was actually allowed to make a call.
Both companies had removed the org chart. Neither had built a working replacement for what it had been doing, however badly.
Not every attempt to make self-management work at scale ended in disaster. And successful stories carry insights that become more salient the more AI agents revolutionize the workplace.
SUCCESS STORIES
Haier
Haier is a Chinese home appliances manufacturer with 80,000 employees. In 2012, CEO Zhang Ruimin dismantled the traditional management structure and reorganized the entire company into roughly 4,000 micro-enterprises — internal startups, each with its own P&L, its own strategy, and the freedom to hire, partner, and compete. The model is called RenDanHeYi: the integration of people and goals.
Accountability no longer flows upward to a VP who reports to a SVP. It flows outward to the customer. Each micro-enterprise lives or dies by the market it serves. That is a more brutal accountability mechanism than any performance review cycle — and it’s also a structure that scales beautifully once each micro-enterprise can run its own market read and coordination with an AI agent instead of a back office.
Buurtzorg
In 2006, a Dutch nurse named Jos de Blok founded Buurtzorg on a single organizing principle: give small teams of nurses everything they need to run themselves, and get out of the way. Today, Buurtzorg employs around 15,000 nurses organized into self-managing teams of 10 to 12. There is no middle management layer. The back office supporting all 15,000 nurses is around 50 people. Overhead runs at roughly 8% — against an industry average of 25%.
Buurtzorg built that 50-person back office by hand, over a decade, because de Blok wanted it and had the discipline to enforce it. Most companies don’t have a founder willing to fight that fight. AI-supported operations gets you a version of that ratio without the decade or the discipline — which is either the best argument for doing this deliberately before it’s forced on you, or the reason most companies will end up with Buurtzorg’s overhead numbers without ever making Buurtzorg’s choice.
Basecamp and GitLab
Neither company reorganized into circles or micro-enterprises. What they did was quieter and more transferable: they systematically replaced the functions that management typically performs with documented systems anyone in the company can access and use. GitLab’s company handbook runs to thousands of pages and is publicly available — covering everything from how to run a meeting to how to make a hiring decision. GitLab went public in 2021 with over 1,300 people, fully remote, no offices, without adding the management layers that typically accompany that growth.
The lesson is not that they had no management functions. They had all of them. They had just moved those functions out of people’s heads and into systems. GitLab’s handbook is what that move looks like when a company builds it manually, page by page, over years. An AI knowledge layer is what that move looks like when a company doesn’t have years — the same function, compressed from a decade of deliberate documentation into whatever your agents can index this quarter.
Three companies. Three industries. Three different implementations. Read together: a checklist for what has to be true before you let AI take over the routing function your hierarchy currently performs.
the checklist
Accountability must be anchored to something external. In all three cases, employees are not accountable to a manager but to a market, a patient, a shipped product. An AI agent can move context instantly. It cannot manufacture a market to be accountable to. If accountability in your org currently lives inside the hierarchy rather than outside it — tied to a manager’s approval rather than a customer’s outcome — automating the routing layer won’t fix that. It will just remove the last thing standing between your teams and no accountability at all.
The people doing the work must already know what good work looks like. Buurtzorg’s nurses are trained professionals. GitLab engineers know what good code looks like. Autonomy handed to people without a strong craft or professional foundation does not produce self-management — it produces anxiety and drift. Same with AI: an agent can hand someone a decision faster than a manager could. It cannot hand them the judgment to make it. If that judgment isn’t already distributed in your organization, faster access to context just means faster, more confident wrong answers.
Management functions must be replaced by systems, not eliminated. Managers do real work — distributing context, resolving resource conflicts, carrying institutional knowledge, making and recording decisions. In every successful case, these functions still happen via a P&L structure, a handbook, a software platform, and a professional training framework. AI is a candidate system for this — arguably the most powerful one yet available. But it’s a candidate, not a guarantee. Bolting an agent onto a broken structure automates the dysfunction; it doesn’t remove it.
If the primary function of management layers has always been information routing — and AI can now handle that by excellence — then management-as-a-career-path is a historical artifact.
READY OR NOT
What survives is accountability. But accountability, separated from its coordination wrapper, doesn’t distribute the way a hierarchy does. It concentrates where the expertise is. And here is the uncomfortable implication most organizations haven’t yet sat with: a significant portion of what your managers currently do are professional functions — judgment calls, quality standards, domain decisions — that got delegated upward because that’s where the authority was, not because that’s where the knowledge was.
AI will not solve that. It will expose it. Once an agent can hand a room everything a meeting used to take an hour to establish, the only thing left worth arguing about is who in that room actually knows what to do with it. An org chart was never built to answer that.
The organizations that will navigate this well are not necessarily the fastest AI adopters. They’re the ones that already know where their real network is — where work actually gets decided, who the actual knowledge anchors are, and which of their management functions can be systematized versus which require genuine expertise. Most companies haven’t run that audit. AI will run it for them, whether they’re ready or not.
Here’s the part that should actually keep you up at night: the companies in the success stories above ran this audit voluntarily, on their own timeline, over years, with a founder or CEO willing to absorb the transition cost. You don’t get that version. Every SaaS product your company already pays for — the CRM, the ticketing system, the codebase, the doc tool — is getting an agent bolted onto it this year, on your vendor’s release schedule, not yours. The shadow network you didn’t map is about to become visible.
The only question left is whether you’re the one reading the results, or the last one to find out what they said.
Featured Case Studies
Featured Case Studies