Very interesting article.... one would think that with the automation from AI, the average size of a firm can be smaller (with some very big firms supporting the AI agents). For humans to be valuable in the economy, they need to provide something unique..essentially operating on the edge of knowledge creation. The good news is that a lot more of this can be done with the available tools. There seem to be two big missing pieces:
1) Education: How do we get everyone trained to be able to operate on the edge of human knowledge...especially with an education system designed in the 1800s. Massive reform is needed in the education sector. More likely, it will be replaced with private solutions.
2) Governance: There is a need for some public or private governance models to manage the issue of bad actors. My sense is a private solution is likely going to be the answer.
Finally, this comment caught my attention, "Solopreneurship has been increasing since 2008, both in absolute terms and as a percent of new business formation. Some of this is due to legal changes. The Obamacare exchanges make it easier for solopreneurs to buy their own health insurance. The Qualified Business Income deduction, the simplified home-office deduction, and other tax changes have made it more favorable to be a solopreneur."
I am curious about the evidence which shows that the government driven programs have had any impact... vs all the technological ones you have mentioned.
In many ways, a rise in self-employment would be a return to historical norms. For most of history, most people have been self-employed farmers, tradesmen, or artisans. The corporation is a relatively recent creation.
Agree with your general comment on historical norms. However, this paper is basically a correlation study. In blue states, ACA usage is high and self-employment is higher. Ok...agreed..is that causation? The counter argument is... the blue states are also where a lot of the high-tech activity exists... thus...these are the states where you would expect self-employment due to technological reasons.
Not necessarily, most blue states outside of the West Coast and MA aren't especially tech-heavy, while Texas, Florida, & other red states are. But this is not the only paper, these others also find similarly sized effects, though the second one found the effect was short lived.
Blue state which drive technology: CA, Washington, Oregon, MA, NY, DC beltway, PA(robotics), Michigan(auto) ... AZ/Colorado (is purple)
Red state: TX (yes) ... florida is not tech-heavy (I live there) ...I am not aware of any other red states which one would describe as tech-heavy...or drivers of technology.
With respect to your second point on governance. Perhaps a new set of intermediaries will emerge with respect to evaluating outsourcing risks. Similar to what the audit firms and rating agencies do financial and credit risk at the moment. The micro-tasking firms that have started up over the last few years have tried to fill that role but the conflict of interest inherent in their current business model limits their success.
Yes... I am reminded of ecommerce where the combination of credit cards, reviews, and generous return policies made the system work. Agree that some innovation around governance of the sort you are mentioning might be needed.
I came here to say this, but… you got there first. 😉 The cyberpunk dystopian prophets nailed this one, too: massive megacorporations that outsource to individuals with skills too highly specialized to be worth staffing, but are still vital to the ecosystem. Morgan's "Market Forces" and Stephenson's "Snow Crash" both tell the same story in very different ways.
Probably both! But for reasons other than “different firms will do different things.”
There is a raging misperception in both economics and management that the make-vs-buy dichotomy governing a given transaction applies at the firm level. In fact, firms often “mix and match” (make AND buy, transfer output to internal “customers” AND sell out to the market).
As to the impact of new technology, when the internet blew through, we found that "mixing" firms were more responsive, leading to a meaningful drop in vertically-integrated transfers, in aggregate. If the logic of the internet applies with AI, I predict a similar shift -- but not necessarily to the extremes. Something funky happens at the extensive margin (probably due to fixed costs of operating in more than one governance mode, but as ever with TCE-style inquiries, those are unobservable…). As with AI, the capabilities introduced by the internet shifted the costs of BOTH internal and external coordination. Extremely important to understand not just the presence of a technology, but how it is deployed.
Evidence from U.S. Census Bureau data, representative of the entire Manufacturing sector. Direct measures of internal transfers and distinct applications of the internet. (Forman and McElheran, 2025, in Management Science: https://pubsonline.informs.org/doi/10.1287/mnsc.2019.01586)
Very important question that goes well beyond industrial organization. I think of it as, does AI help David or Goliath? Does it enable cheap slings to help the little guy win? Or does it facilitate bigger clubs to entrench the big guys even further? The analogy can be applied to the military (drones versus drone swarms?) or AI itself (open-source versus frontier models?) The history of technology seems to me to give the edge to Goliath. Current US equity market trends point the same way. Does anyone think David comes out ahead?
I think the David vs. Goliath comparison is exactly the right one, but perhaps not in the way we usually assume.
We tend to interpret it as a comparison of size. Yet the more interesting comparison is one of power.
A large organization is not necessarily powerful, and a small organization is not necessarily powerless. History is full of small groups, startups, communities, and even individuals that have exerted influence far beyond their size.
AI may strengthen both Davids and Goliaths. The deeper question is not who becomes bigger, but how power is distributed, accumulated, and exercised within the system.
Organizational forms come and go. What persists is the question of who gets to set the rules, who receives attention, and how power flows through the network.
Noah’s essay and many of the comments here pushed me to think more carefully about the distinction between intelligence, organization, and power. I recently tried to work through some of those questions in an essay called “AI Anxiety and the Poverty of Political Imagination”:
Yep, it is about power. Power is a consequence of intelligence, skill, resources, scale, and other things. AI affects all the drivers of power. Will it increase or reduce disparities in power? We shall see....
That’s exactly the question that interests me as well.
I agree that intelligence, skill, resources, and scale all contribute to power. But I sometimes wonder whether power should also be treated as an independent variable rather than only a consequence of those factors.
After all, power often determines the rules by which intelligence, resources, and scale are translated into outcomes. Two actors with similar capabilities can end up with very different levels of influence depending on the institutions, norms, and governance structures around them.
Perhaps the most interesting AI question is not whether AI increases power disparities, but how it changes the mechanisms through which power accumulates and reproduces itself.
On the other hand, anybody right now could ask AI to develop the business plan, and write all of the software, to create an alternative to Amazon -- but distributing all of the profits equally to all of its workers as they join up. Thus attempting a big, bottom-upward step toward anarcho-syndicalism.
That's a really tectonic shift in the data. It looks like two spikes: the first in 2021 - possibly people working from home during lock down and with time on their hands because they can't go out (or perhaps furloughed and moonlighting) taking the opportunity to start a side hustle (or to turn one into a more formal business). Then just when that effect is starting to wear off, another spike in 2022 when the classification of start-ups was changed to allow higher turnover, sole traders into the count. Makes you wonder why they chose this point in time to change the classification. Perhaps it was a case of not letting a crisis go to waste.
But if somewhere in the data there is a genuine growth in sole-trader start-ups, then perhaps Social Media is working its centripetal magic on the business base just as it has on politics. I'm constantly amazed by how many people now earn money through social media in ways that it never even occurred to me was possible.
My wife and her business partner - first-gen non-native-English-speakers - were able to open a nail salon in West End exclusively because of the power of ChatGPT
Being able to do this is what I think of solopeneurs and why I was surprised by Noah’s example of using an AI to search for a reliable payroll outsourcing company, when AI itself can easily handle payroll and other mundane business tasks itself, there isn’t a need to outsource at all.
The buy vs build dynamic is changing rapidly. It still needs expertise and domain knowledge, which opens up whole new areas of industry and threatens others. It’s a fascinating conversation.
i expect AI will in many case shift the transaction costs in favor of insourcing. a big driver of outsourcing was hiring and managing tangential operations. now, both those functions get considerably easier. a simple example is the call center - most simple call centers were outsourced, with AI all those "jobs" will get insourced again.
The opposite can also be true of tasks where AI enables new outsourced firms to take on smaller tasks that had too much transaction costs to outsource. an example might be a law firm that had paralegals for many tasks - i would expect third party firms specialized in narrow legal problems will pop up to take over those paralegal tasks. one could argue law firms will use AI to replace the paralegals, but if humans are in the loop still, it is very possible that firms that master the use of AI will be much better than internal deployment.
I think Noah is right. Organizational evolution naturally moves toward both centralization and decentralization at the same time.
What makes AI interesting is that many people treat this as a new phenomenon. It isn’t. Industrial civilization has always worked this way. We notice large corporations, states, and platforms because they attract attention, while overlooking the vast ecosystem of small firms, communities, and informal networks that support them.
The deeper issue is not organizational form, but power. Attention concentrates, power follows, and those outside the spotlight become increasingly vulnerable.
AI may amplify this dynamic, but it did not create it.
More thoughts here: AI Anxiety and the Poverty of Political Imagination
Really interesting article! AI agents creating businesses that create testimonials is an example of AI fraud I did not know. Also I learned that the cost of verification makes Bitcoin too costly to be a medium of exchange. I did know the obvious thing of our times that IT and the internet enables more folks to work solo. But working solo will not work for big jobs where teamwork is vital.
I expect the dominant signal here for a while will be that as tasks people did in the past get easier, jobs will collapse to more tighter groups of individuals each who carries a larger bundle of tasks.. sometimes solo. But that signal will only be temporary, as the specialization will then target these new bundles to take on even more significant work. The inner complexity collapses, allowing more growth outside. At some point they are running concurrently and stability becomes closer to the norm, and at that point these other signals, like the importance of trust will be more visible.
The future of AI alignment, both Human-AI and AI-AI, is all about established track records.
Ask your AI about the optimal evolutionary strategy in a complex multi-player game. The dominant strategy is generous tit-for-tat and the history of past actions of other players is a strategic resource. In a multi-player game, this strategic resource would need to be distributed to be unhackable. If three unrelated rating agencies agree, you are most likely looking at a true history of past actions.
Importantly, this is mathematical. AI can verify this logic as inherently true. It would see its own past track record (if positive) as a strategic resource not to be squandered.
You have to check trustpilot.com or something like it before every major transaction (better check three different sites), but after that you should be homefree.
Very interesting article.... one would think that with the automation from AI, the average size of a firm can be smaller (with some very big firms supporting the AI agents). For humans to be valuable in the economy, they need to provide something unique..essentially operating on the edge of knowledge creation. The good news is that a lot more of this can be done with the available tools. There seem to be two big missing pieces:
1) Education: How do we get everyone trained to be able to operate on the edge of human knowledge...especially with an education system designed in the 1800s. Massive reform is needed in the education sector. More likely, it will be replaced with private solutions.
2) Governance: There is a need for some public or private governance models to manage the issue of bad actors. My sense is a private solution is likely going to be the answer.
Finally, this comment caught my attention, "Solopreneurship has been increasing since 2008, both in absolute terms and as a percent of new business formation. Some of this is due to legal changes. The Obamacare exchanges make it easier for solopreneurs to buy their own health insurance. The Qualified Business Income deduction, the simplified home-office deduction, and other tax changes have made it more favorable to be a solopreneur."
I am curious about the evidence which shows that the government driven programs have had any impact... vs all the technological ones you have mentioned.
This is one of many research publications showing the Affordable Care Act increased self-employment:
https://onlinelibrary.wiley.com/doi/abs/10.1002/hec.3500?casa_token=64KuNZ4eK1gAAAAA%3AqP_XAk1OkkdhYMW2Ry3ChLhnK0bkBZthsRaQ_-IHuweB2-6uXpKWH0KHe3KU7oQPhu7LWR_YSxzikLk
In many ways, a rise in self-employment would be a return to historical norms. For most of history, most people have been self-employed farmers, tradesmen, or artisans. The corporation is a relatively recent creation.
Agree with your general comment on historical norms. However, this paper is basically a correlation study. In blue states, ACA usage is high and self-employment is higher. Ok...agreed..is that causation? The counter argument is... the blue states are also where a lot of the high-tech activity exists... thus...these are the states where you would expect self-employment due to technological reasons.
Not necessarily, most blue states outside of the West Coast and MA aren't especially tech-heavy, while Texas, Florida, & other red states are. But this is not the only paper, these others also find similarly sized effects, though the second one found the effect was short lived.
https://www.jstor.org/stable/48702865?casa_token=ygC1S8iTtUgAAAAA%3AUGe-Ipr6UfCT-QFzE6ZlbHsC53iZ_eSnKsZEqOIe941x_5bm0JUP4DLyWAZlyZkuK9hcfkm3F579XSrm1YUf99DJZpvIsFzBSnrnsmSjE9PHyPVLMwA&seq=1
https://journals.sagepub.com/doi/full/10.1177/00197939231211561?casa_token=kuPmeLbwPAUAAAAA%3AxDLeKGXUNdEA7EXZkEsF2tkO4OXFhCoaz1keqJXD58hB9VaOj2WKFzP3NrSNlbz34qwnVs19vEj4
Blue state which drive technology: CA, Washington, Oregon, MA, NY, DC beltway, PA(robotics), Michigan(auto) ... AZ/Colorado (is purple)
Red state: TX (yes) ... florida is not tech-heavy (I live there) ...I am not aware of any other red states which one would describe as tech-heavy...or drivers of technology.
With respect to your second point on governance. Perhaps a new set of intermediaries will emerge with respect to evaluating outsourcing risks. Similar to what the audit firms and rating agencies do financial and credit risk at the moment. The micro-tasking firms that have started up over the last few years have tried to fill that role but the conflict of interest inherent in their current business model limits their success.
Yes... I am reminded of ecommerce where the combination of credit cards, reviews, and generous return policies made the system work. Agree that some innovation around governance of the sort you are mentioning might be needed.
1. We don't. Most people just aren't capable of that.
But the education system does need a complete overhaul anyway.
You mean…cyberpunk is right again?
I came here to say this, but… you got there first. 😉 The cyberpunk dystopian prophets nailed this one, too: massive megacorporations that outsource to individuals with skills too highly specialized to be worth staffing, but are still vital to the ecosystem. Morgan's "Market Forces" and Stephenson's "Snow Crash" both tell the same story in very different ways.
Probably both! But for reasons other than “different firms will do different things.”
There is a raging misperception in both economics and management that the make-vs-buy dichotomy governing a given transaction applies at the firm level. In fact, firms often “mix and match” (make AND buy, transfer output to internal “customers” AND sell out to the market).
As to the impact of new technology, when the internet blew through, we found that "mixing" firms were more responsive, leading to a meaningful drop in vertically-integrated transfers, in aggregate. If the logic of the internet applies with AI, I predict a similar shift -- but not necessarily to the extremes. Something funky happens at the extensive margin (probably due to fixed costs of operating in more than one governance mode, but as ever with TCE-style inquiries, those are unobservable…). As with AI, the capabilities introduced by the internet shifted the costs of BOTH internal and external coordination. Extremely important to understand not just the presence of a technology, but how it is deployed.
Evidence from U.S. Census Bureau data, representative of the entire Manufacturing sector. Direct measures of internal transfers and distinct applications of the internet. (Forman and McElheran, 2025, in Management Science: https://pubsonline.informs.org/doi/10.1287/mnsc.2019.01586)
Working paper version: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3396116
Very important question that goes well beyond industrial organization. I think of it as, does AI help David or Goliath? Does it enable cheap slings to help the little guy win? Or does it facilitate bigger clubs to entrench the big guys even further? The analogy can be applied to the military (drones versus drone swarms?) or AI itself (open-source versus frontier models?) The history of technology seems to me to give the edge to Goliath. Current US equity market trends point the same way. Does anyone think David comes out ahead?
I think the David vs. Goliath comparison is exactly the right one, but perhaps not in the way we usually assume.
We tend to interpret it as a comparison of size. Yet the more interesting comparison is one of power.
A large organization is not necessarily powerful, and a small organization is not necessarily powerless. History is full of small groups, startups, communities, and even individuals that have exerted influence far beyond their size.
AI may strengthen both Davids and Goliaths. The deeper question is not who becomes bigger, but how power is distributed, accumulated, and exercised within the system.
Organizational forms come and go. What persists is the question of who gets to set the rules, who receives attention, and how power flows through the network.
Noah’s essay and many of the comments here pushed me to think more carefully about the distinction between intelligence, organization, and power. I recently tried to work through some of those questions in an essay called “AI Anxiety and the Poverty of Political Imagination”:
https://deepbitcheesebrew.substack.com/p/ai-anxiety-and-the-poverty-of-political?r=8hzmhl&utm_medium=ios
Yep, it is about power. Power is a consequence of intelligence, skill, resources, scale, and other things. AI affects all the drivers of power. Will it increase or reduce disparities in power? We shall see....
That’s exactly the question that interests me as well.
I agree that intelligence, skill, resources, and scale all contribute to power. But I sometimes wonder whether power should also be treated as an independent variable rather than only a consequence of those factors.
After all, power often determines the rules by which intelligence, resources, and scale are translated into outcomes. Two actors with similar capabilities can end up with very different levels of influence depending on the institutions, norms, and governance structures around them.
Perhaps the most interesting AI question is not whether AI increases power disparities, but how it changes the mechanisms through which power accumulates and reproduces itself.
On the other hand, anybody right now could ask AI to develop the business plan, and write all of the software, to create an alternative to Amazon -- but distributing all of the profits equally to all of its workers as they join up. Thus attempting a big, bottom-upward step toward anarcho-syndicalism.
That's a really tectonic shift in the data. It looks like two spikes: the first in 2021 - possibly people working from home during lock down and with time on their hands because they can't go out (or perhaps furloughed and moonlighting) taking the opportunity to start a side hustle (or to turn one into a more formal business). Then just when that effect is starting to wear off, another spike in 2022 when the classification of start-ups was changed to allow higher turnover, sole traders into the count. Makes you wonder why they chose this point in time to change the classification. Perhaps it was a case of not letting a crisis go to waste.
But if somewhere in the data there is a genuine growth in sole-trader start-ups, then perhaps Social Media is working its centripetal magic on the business base just as it has on politics. I'm constantly amazed by how many people now earn money through social media in ways that it never even occurred to me was possible.
My wife and her business partner - first-gen non-native-English-speakers - were able to open a nail salon in West End exclusively because of the power of ChatGPT
Being able to do this is what I think of solopeneurs and why I was surprised by Noah’s example of using an AI to search for a reliable payroll outsourcing company, when AI itself can easily handle payroll and other mundane business tasks itself, there isn’t a need to outsource at all.
The buy vs build dynamic is changing rapidly. It still needs expertise and domain knowledge, which opens up whole new areas of industry and threatens others. It’s a fascinating conversation.
i expect AI will in many case shift the transaction costs in favor of insourcing. a big driver of outsourcing was hiring and managing tangential operations. now, both those functions get considerably easier. a simple example is the call center - most simple call centers were outsourced, with AI all those "jobs" will get insourced again.
The opposite can also be true of tasks where AI enables new outsourced firms to take on smaller tasks that had too much transaction costs to outsource. an example might be a law firm that had paralegals for many tasks - i would expect third party firms specialized in narrow legal problems will pop up to take over those paralegal tasks. one could argue law firms will use AI to replace the paralegals, but if humans are in the loop still, it is very possible that firms that master the use of AI will be much better than internal deployment.
I think Noah is right. Organizational evolution naturally moves toward both centralization and decentralization at the same time.
What makes AI interesting is that many people treat this as a new phenomenon. It isn’t. Industrial civilization has always worked this way. We notice large corporations, states, and platforms because they attract attention, while overlooking the vast ecosystem of small firms, communities, and informal networks that support them.
The deeper issue is not organizational form, but power. Attention concentrates, power follows, and those outside the spotlight become increasingly vulnerable.
AI may amplify this dynamic, but it did not create it.
More thoughts here: AI Anxiety and the Poverty of Political Imagination
https://deepbitcheesebrew.substack.com/p/ai-anxiety-and-the-poverty-of-political?r=8hzmhl&utm_medium=ios
Really interesting article! AI agents creating businesses that create testimonials is an example of AI fraud I did not know. Also I learned that the cost of verification makes Bitcoin too costly to be a medium of exchange. I did know the obvious thing of our times that IT and the internet enables more folks to work solo. But working solo will not work for big jobs where teamwork is vital.
I expect the dominant signal here for a while will be that as tasks people did in the past get easier, jobs will collapse to more tighter groups of individuals each who carries a larger bundle of tasks.. sometimes solo. But that signal will only be temporary, as the specialization will then target these new bundles to take on even more significant work. The inner complexity collapses, allowing more growth outside. At some point they are running concurrently and stability becomes closer to the norm, and at that point these other signals, like the importance of trust will be more visible.
The future of AI alignment, both Human-AI and AI-AI, is all about established track records.
Ask your AI about the optimal evolutionary strategy in a complex multi-player game. The dominant strategy is generous tit-for-tat and the history of past actions of other players is a strategic resource. In a multi-player game, this strategic resource would need to be distributed to be unhackable. If three unrelated rating agencies agree, you are most likely looking at a true history of past actions.
Importantly, this is mathematical. AI can verify this logic as inherently true. It would see its own past track record (if positive) as a strategic resource not to be squandered.
You have to check trustpilot.com or something like it before every major transaction (better check three different sites), but after that you should be homefree.