> ## Content Index > Fetch the complete content index at: https://www.s3t.org/llms.txt > Use this file to discover other available public pages before exploring further. # The Great Unbundling: How AI is reshaping firms & employment - URL: https://www.s3t.org/the-great-unbundling-how-ai-is-reshaping-firms-employment/ - Published: 2026-11-06T06:28:00.000Z - Updated: 2026-11-06T06:27:59.000Z - Author: Ralph Perrine - Tags: Unemployment, Jobs, AI *AI is beginning to challenge the economics that made large corporations and traditional employment structures necessary. As individuals and small teams gain the ability to orchestrate increasingly capable AI agents, executives and investors face a fundamental question: Where will work, capital, and economic power migrate—and who will capture the value?* S3T PodCast The Great Unbundling 11-6-26 0:00 /1162.1616326530611 1× Last week we looked at how AI will remake workflows and capital flows. This raised a crucial question: If AI causes corporations to divert from Coase's original theory (that firms exist because transaction workloads and costs prohibit an economy comprised solely of freelancers), then what happens to employment? Today we want to look at early thinking and signals that can help us start to understand what is likely to change, and what can we start to learn now about those changes and how to get ready for them. ### Smaller entities -->Bigger results The [Commanding Heights](https://commanding-heights.com/?ref=s3t.org) essay we noted last week predicts that companies may become comprised of smaller players doing bigger things. We are starting to see examples: - Mithril, the biopharma recently acquired by Anthropic was literally one person - Cody Tranbarger - orchestrating a set of AI agents to discover a new treatment for pediatric muscular dystrophy. [Full story here](https://x.com/Prof%5FOak%5F/status/2115185849424048564?ref=s3t.org) and Technical Report for the newly discovered treatment [here](https://www-cdn.anthropic.com/52e1f5d50c47ca3818f56704c27cf19e59b62e77.pdf?ref=s3t.org). - Cisco gives a 27 year old former startup founder[ unlimited ai tokens in order to refactor 98% of Cisco's legacy code](https://www.tomshardware.com/tech-industry/big-tech/cisco-gives-27-year-old-engineer-unlimited-ai-tokens-to-rewrite-98-percent-of-the-companys-legacy-code-in-24-months-president-plucks-former-startup-founder-telling-him-to-go-nuts?ref=s3t.org). In the logic of previous capabilities, a project of this magnitude would seem impossible. ![](https://storage.ghost.io/c/c7/1e/c71eed42-6fd8-496e-9a70-8c65b93c91e7/content/images/2026/11/image.png) Excerpt of Mithril [Treatment of LAMA2-Related Congenital Muscular Distrophy](https://www-cdn.anthropic.com/52e1f5d50c47ca3818f56704c27cf19e59b62e77.pdf?ref=s3t.org) ### Single person productivity increases These and more constitute a growing set of experiments **pushing the envelop to see how productive one person can become**: [Cody Tranbarger put it this way](https://x.com/Prof%5FOak%5F/status/2115185849424048564?ref=s3t.org): *"Mithril became a bit of an experiment itself - to find the upper bound of how far it could scale as a single operator + a bunch of Claude (+ a stellar bench of part-time consultants)."* Cody goes on to explain why the experiment matters: *Most drug hunters see a handful of go/no-go calls in a career, & spend entirely too much of the rest trading emails (my inbox can attest)...*Give them the bandwidth to run several programs at once, and the time to focus on scientific judgment, and I’m willing to bet they’ll make better calls.* And at scale, better calls translate to more drugs, higher hit rates, broader patient impact, & likely, a disease or 10 or 100 cured.* In other words, the traditional approach to drug discovery is laden with extra processes and people that may not directly contribute to solving target problems. Mithril's approach condenses this down to a single orchestrator (Cody) armed with AI agents and then selected sources of expertise (the "stellar bench of part-time consultants" noted above). This "[Pipeline in a Person](https://x.com/Prof%5FOak%5F/status/2115185849424048564?ref=s3t.org)" approach aligns with findings of [Yale's "AI Premium" study](https://news.yale.edu/2026/07/13/analysis-380-trillion-ai-tokens-reveals-how-technology-transforming-financial-markets?ref=s3t.org) which noted that that *routine* scientific work would be most impacted by AI: *"Occupations in science are among the most negatively affected, which may seem surprising. There is a distinction to be made here between more routine lab work and the kind of cutting-edge scientific research that occurs at major research universities. *It is the routine work that our analysis indicates will be most affected by AI*.”* Armed with the right skills, AI Tools, and access to expertise, **humans could become better orchestrators of problem solving attempts**, and solve problems more efficiently than traditional corporate departmental structures have been able to. Will these new economics make new kinds of projects feasible? ![A group of people sitting at a bar](https://images.unsplash.com/photo-1721673378502-6d5e9d1f56f7?crop=entropy&cs=tinysrgb&fit=max&fm=jpg&ixid=M3wxMTc3M3wwfDF8c2VhcmNofDMwfHxuYXBraW4lMjBiYXJ8ZW58MHx8fHwxNzkzOTEwMzM3fDA&ixlib=rb-4.1.0&q=80&w=2000) Photo by [Meg von Haartman](https://unsplash.com/@traveleroohlala?utm%5Fsource=ghost&utm%5Fmedium=referral&utm%5Fcampaign=api-credit) / [Unsplash](https://unsplash.com/?utm%5Fsource=ghost&utm%5Fmedium=referral&utm%5Fcampaign=api-credit) ### Doing things we couldn't do: the concept of Total Addressable Work Back when dot.com startups were born as notes on bar napkins, many of those notes featured TAM - the old Total Addressable Market concept - roughly the idea that there was a huge market for a product or service but until (the internet, e-commerce etc.) this market couldn't be accessed in a very efficient or effective way. Let's borrow that idea to think about Total Addressable *Work*. Today there is a percentage of high value work (work that could unlock benefits and value at very large scale) that is literally out of reach due to limitations of current tools, current organizational friction, and current human capabilities. The recent Navier-Stokes other math solutions generated by AI are examples of being able to solve theoretical problems that humans without AI previously could not: *"Roughly 10,000 agents coordinated in a mere 88 hours to produce a proposed solution to something humans had not been able to solve for nearly a century."* (from [*Technological Progress and the Meaning of Work*](https://aleximas.substack.com/p/technological-progress-and-the-meaning)) The Mithril and Cisco examples above (especially the Cisco example) suggests to me that AI agents will also allow us to tackle a new universe of *practical* work that we previously couldn't touch. Anecdotally I am seeing other examples of this emergence of “orchestrator“ roles where a single person becomes a manager of multiple agents, consulting with and pulling in select expert humans as needed but single-handedly driving a high volume of work and outputs that previously would have been delegated to a larger team of FTEs. ![](https://storage.ghost.io/c/c7/1e/c71eed42-6fd8-496e-9a70-8c65b93c91e7/content/images/2026/11/image-1.png) So, as depicted above, we have human orchestrators simultaneously directing the work of multiple agents, with the already proven potential to significantly boost productivity. But as we'll see below, there's another flavor. ![](https://storage.ghost.io/c/c7/1e/c71eed42-6fd8-496e-9a70-8c65b93c91e7/content/images/2026/11/image-2.png) ### But wait there's more: Coordination and Contracting get better [This new DeepMind Institute paper](https://institute.deepmind.com/essays/the-economics-of-intelligence/?ref=s3t.org) explores how AI could shape economic transformation, and concludes that productivity increases for individual humans won't be the whole story. The larger impact will occur as AI agents become good at coordinating with each other: *"Gains in the ability of agents to work together within a firm alter how organizations scale. Once coordination reaches a critical threshold, it triggers large, abrupt jumps in market concentration, employment, and wages"* The **Grossman-Hart-Moore (GHM) theory of incomplete contracts** (articulated across 2 papers: [here](https://faculty.washington.edu/mfan/is582/articles/GH1986.pdf?ref=s3t.org) and [here](https://hart.scholars.harvard.edu/sites/g/files/omnuum10571/files/hart/files/jpe%5F1990.pdf?ref=s3t.org)) observes that contracts (between company and vendor or company and employee) are “incomplete”. They do not and cannot perfectly cover every future scenario and or know every issue that could impact performance nor can they specify every ideal remedy. As a result, **a percentage of contractual arrangements are suboptimal.** This has been the problem:Think about the cases you’ve seen where a vendor isn't impressing anyone yet they are not likely to be replaced because it would be so disruptive. Likewise in employee-employer relationships: one or both parties may not be meeting their obligations to each other, yet the arrangement persists. These persistent suboptimal arrangements cause significant inefficiencies. In aggregate, these inefficiencies in practice mean that more people remain employed than a truly efficient process would require. ### Will AI agents have less friction than vendors and human employees? They will but the initial workload to get them working correctly will be significant - at least until standard patterns emerge. The process of creating and testing an agent requires a high degree of specificity, across *multiple* disciplines. Two agents that coordinate a simple task of managing a person's travel schedule and arrangements might not have a lot of risk or complexity. But a set of agents that coordinate investments or insurance instruments across multiple finance partners and corporate clients would be a different story. The process of creating, testing and managing those agents will require expertise from legal, liability, financial, compliance, as well as technical sources. Setting up this kind of contracting will require the same due diligence and cross functional expertise required for a large complex contracting or partnership arrangement between corporations. At least until we see the emergence of a standard set of proven patterns. ### Will large firms disappear entirely? I don't think so. But I do think that the structure and work of large firms - and what makes them durable - will change significantly. The assignment is pretty clear for large firms and smaller firms that aspire to grow: Find "Scarce, controllable positions of *Economic Power"* and leverage these positions to"capture the gains of *Increasingly Capable* and diffused intelligence" (from the [Commanding Heights](https://commanding-heights.com/?ref=s3t.org) subtitle). This suggests we could see a large decomposition of the Fortune 1000 and even the PE held assets, as management teams "learn how to drive on ice" ...suddenly skid into unfamiliar terrain, lose control, and collide with changing economic realities. It seems to me that there will be an increase in small 1-10 person firms that are expertise driven and token hungry. They will live by the maxim: **"refuse to be put in a position of having to wait for others to play through their mistakes."** These smaller entities will be in high demand as the larger firms recover and realign, shed their dead weight and look to acquire or partner with small efficient engines that can get them back on track. Whatever on track means at that point. --- *Opinions expressed are those of the individuals and do not reflect the official positions of companies or organizations those individuals may be affiliated with. Not financial, investment or legal advice, and no offers for securities or investment opportunities are intended. Mentions should not be construed as endorsements. Authors or guests may hold assets discussed or may have interests in companies mentioned.* *(c) 2026 All Rights Reserved. No part of this may be copied or shared without permission.*