Published March 24, 2026 in Artificial Intelligence • 11 min read • 
Audio available
When will we get our first one-person unicorn – a company valued at $1bn that has a single employee? Sam Altman and his CEO friends are confident it will happen soon, thanks to the revolution in generative AI. But do we even need that one person? With agentic AI, we could soon see zero-person corporations that operate according to a core mission that may or may not be visible to us. Let’s call it an algorithmic corporation.
Tech platforms have pioneered a model in which the vast majority of labor is provided by contractors, not employees. DoorDash’s states that the company has 23,700 employees and eight million “dashers” (non-employee drivers) – that’s 340 gig workers for every employee. Similar figures hold for Uber and other platforms. Hiring, evaluation, compensation, and firing of gig workers are managed by algos on the app, not human managers. Big Tech seems to be allergic to creating stable jobs, and the stock market rewards every layoff notice. A zero-employee corporation would be Wall Street’s dream come true.
But what would a zero-person AI unicorn look like? The closest approximation might be a “paper clip maximizer” (PCM) – an AI that relentlessly pursues the goal of making as many paper clips as possible, regardless of the consequences. Philosopher Nick Bostrom proposed this thought experiment two decades ago, and it has animated discussions about the ethics and direction of AI ever since. The PCM’s ruthless optimization leads it to exploit resources without limit, manipulate people without regret, and eventually turn this planet (and maybe others) into paper clips and gray goo. For the algorithmic corporation, substitute shareholder value for paper clips.
Here, I will describe some of the likely features of the emerging algorithmic corporation. The main elements are already in place; the question seems to be not whether, but when we will see employee-free enterprises. And yet we are not obligated to submit to this gloomy vision.
Creating an enterprise typically requires capital, labor, supplies, distribution, and a legal entity. Each of these is accessed through rule-bound markets, and the rules are largely set by national governments. Because each government sets different rules (creating a specific “institutional terroir”), we see very different ways of structuring business around the world. That’s why, for instance, ride-hailing companies look so different in the US, Germany, India, and Indonesia.
In 2016, I published a law review article that speculated about the “web page enterprise.” If you’ve ever right-clicked on a web page, you see that underlying the lovely image that appears on your screen is a bunch of ugly code. The code orchestrates inputs and creates outputs: calling on APIs, pulling information from SQL databases, querying user inputs, and implementing algorithms to combine these inputs and create the page you see. (A note to the nerds among you: yes, I left some things out.) A web page is a lot like a firm, coordinating various factors and producing an output. Indeed, many online retailers are nothing but a web page, taking orders, forwarding them to vendors or delivery services, collecting payments, and distributing them to providers, like a virtual vending machine.
As I put it then: “Nikefication turned the corporation into a nexus-of-contracts, organizationally separating design from production and distribution. Entrepreneurs grew skilled at assembling contractors into a virtual enterprise. More recently, we have seen Uberization, which allows on-demand labor to be contracted by the task via online platforms. Uberization threatens to turn jobs into tasks, to the detriment of labor. Every input into the enterprise becomes possible to rent rather than to buy, and employee-free organizations are increasingly feasible.”
What would a zero-person AI unicorn look like? The closest approximation might be a ‘paper clip maximizer’ – an AI that relentlessly pursues the goal of making as many paper clips as possible, regardless of the consequences
One part of my speculation reliably got me into trouble: the dystopian idea of an online auction for labor. In this scenario, a major retailer seeking to cut employment costs staffs its stores via an app where independent contractors bid for shifts. The low bidders would win the shift, and wages would vary from day to day according to market conditions, perhaps with surge pricing on weekends. This idea got me laughed out of many rooms as both silly and creepy. What civilized society would ever allow such a thing?
And yet today in the US, there are several “Uber for nursing” apps in which exactly this scenario plays out in hospitals seeking nursing staff by the shift. As the Roosevelt Institute’s report describes: “On-demand nursing companies such as Clipboard Health and ShiftKey encourage workers to take part in personalized pay schemes by bidding against each other. On ShiftKey, Ashley not only expresses her availability for a shift but bids for one against peers by indicating the lowest hourly rate for which she will work. To win the shift, she lowers and lowers her rate until it’s well below a living wage.” Uberization makes app-based recruiting feasible; it’s a trivial technical matter to turn it into an auction, whether for retail workers, nurses, kitchen staff, or college lecturers.
Now, what happens when we introduce AI to this process?
Why did “Uber for everything,” including online labor markets, emerge most enthusiastically in the US? Because America’s institutional terroir is geared toward ruthlessness and rigor. Ruthlessness is evident in relentless efforts to cut costs and keep input prices as low as possible. Auctions and other markets are often the most effective ways to get to the lowest price. Thus, when ICTs such as the smartphone land in the US, enterprises use them to create markets.
Rigor describes the data-driven approach to decision-making in American business. One of the most striking aspects of Silicon Valley is the systematic use of experimentation (A/B testing) to drive business choices – not just “Which headline will drive the most clicks?” but sophisticated design decisions. Which shade of blue yields click-throughs on ads? Which style of artwork gets more subscribers to watch a show? What’s the best ordering of search results to book rental properties? How can you run valid A/B tests when users’ choices influence each other?
With ruthlessness and rigor as orientations, how is AI being implemented today? And how would an algorithmic corporation actually function?
In US labor markets, employees enjoy a host of protections under the Fair Labor Standards Act of 1938 that are not available to contractors, such as a minimum wage, which helps explain why companies have shifted toward contractor models. Law professor Veena Dubal uncovered pervasive algorithmic wage discrimination in the ride-hailing industry. Using the extensive data their platforms gather about drivers and riders, ride-hailing firms can customize the fees they pay drivers, so two drivers sitting next to each other in the airport parking lot might receive different compensation for the same ride. Bonuses and surge pricing also vary by driver, guided by unseen factors known only to the algorithm. It is fair to assume that the same dynamic plays out at the many other “Uber for X” platforms.
The experimental ethos extends to high-skilled platforms as well. Northwestern’s Hatim Rahman and his co-authors describe the “experimental hand” at work on a digital labor platform used by coders, designers, and other skilled artisans, and how experimentation evolved from an occasional, voluntary endeavor to tweak the platform to a pervasive, daily experience built into the terms of service. As Tim Weiss of Imperial College London put it, “We’re all lab rats now,” whether we know it or not.
Agentic AI will automate this process and draw on comprehensive datasets to drive wages to their absolute minimum. Perhaps if you are late on some bills, or your Apple Watch reports that you have skipped your costly medication, the platform will offer you a lower price for your labor. A determined experimenter could quickly uncover your reservation wage, and that is what you will be offered.
Ride-hailing companies also pioneered algorithmic pricing for customers based on their imputed willingness to pay. Your history of ordering and canceling rides, splurging on town cars, or always choosing the cheapest option, including how much you tip and your pickup and drop-off locations, provide helpful insights into the price that will make you say “yes.”
The fast-food chain Wendy’s experienced a backlash when it proposed what critics described as “surge pricing for burgers”. Readily visible prices are standard fodder for outrage, which is why gasoline looms so large in the public’s perception of inflation. But prices accessed via a screen can be exquisitely targeted to individuals. The Groundwork Collective ran a study in which 40 volunteers, connected by video conference, ordered the same items from the same stores online at the same time and received different prices. Instacart, the platform they used, stated: “The pricing tests are short-term, randomized, and designed so that people may see slightly lower prices and some may see slightly higher prices, with the goal of helping retail partners understand consumer preferences and identify categories where they should invest in lower prices.” In other words, they were experimenting to uncover consumers’ willingness to pay.
It is easy to visualize how agentic AI will use this technology to bring every customer to their maximum willingness to pay. Does your phone use suggest that you are going into labor? That ride to the hospital is going to cost you dearly. AI will automate the practices of Martin Shkreli, the “pharma bro” who jacked up essential medicine prices because people with desperate health needs have a high willingness to pay.
Since the early 1990s, many corporations have followed a practice of Nikefication, contracting out core parts of their production process. But while some companies have an arms-length relationship with their vendors, others are much more hands-on. Patrick McGee describes Apple’s supply chain in China as a 21st-century inversion of Ford’s famously integrated Rouge Plant. All the parts are made by vendors, and final assembly is done by Foxconn. (Visualize the Pompidou Center in Paris, where many of the essential parts of the building are visible on the outside.) But Apple is firmly in control and dispatches its engineers to component factories to ensure quality and the adoption of advanced production processes. Moreover, the spatial organization of the Chinese production economy means that Apple has multiple competing vendors for any given part, ensuring that it pays the lowest possible price. Apple gained the control benefits of vertical integration with the cost benefits of outsourcing, making it the world’s most profitable business.
Agentic AI can take a page from Apple’s book by mapping out supply chains and creating virtual auctions at each step. As with labor, AI-enabled data gathering can drive vendors as close as possible to their reservation price. Similar scenarios could easily be envisioned for distribution and logistics too, with AI agents able to autonomously manage stock levels and “optimize” storage and delivery options.
AI does not have a body to do us physical harm, but an AI-controlled corporation with a labor platform could hire humans to do us harm.
The US is distinguished by the fact that corporate law is made by the 50 states, not the federal government. Businesses can incorporate in whatever state they please (most big ones choose Delaware) because, under the US Constitution, states recognize contracts created by other states. Thus, states compete to provide business-friendly laws that attract corporate customers. Delaware gets around a quarter of its annual state budget from fees related to its incorporation business.
One consequence of this system is that it’s easy to create an anonymous business entity – even one controlled by AI. Legal scholar Lynn LoPucki summarizes the consequences: “The algorithm can exercise the rights of the entity, making them effectively rights of the algorithm. The rights of such an algorithmic entity (AE) would include the rights to privacy, to own property, to enter into contracts, to be represented by counsel, to be free from unreasonable search and seizure, to equal protection of the laws, to speak freely, and to spend money on political campaigns.”
AI does not have a body to do us physical harm, but an AI-controlled corporation with a labor platform could hire humans to do us harm. As LoPucki reports: “One commentator has proposed assassination brokering as a possible AE service line. It is not hard to imagine an AE – the identity and location of its autonomous algorithm shielded by an anonymous LLC – matching human assassins with customers and laundering its fees through layers of shell entities using the wide variety of anonymous payment systems currently in development.”
And, rather worryingly, “corporate charter competition, combined with ease of entity migration, makes it virtually impossible for any government to regulate algorithmic control of entities.”
At the end of A Christmas Carol, Scrooge asks the Ghost of Christmas Yet to Come, “Are these the shadows of the things that Will be, or are they shadows of things that May be, only?” In American capitalism, a good rule of thumb is this: whatever nightmare scenario a depraved venture capitalist (VC) can come up with will be implemented within six months. But are we doomed to an AI-driven world of relentless markets for everything, overseen by inhuman algorithms? Or can we, like Scrooge, depart from this path?
Governments are not helpless in the face of threats to their institutional terroir. Labor laws can specify that anyone who meets certain minimum standards counts as an employee and is accorded basic rights of fair employment. Antitrust and consumer protection laws can limit price discrimination. Corporate laws could require transparency for owners and directors, so that non-human entities are not eligible to be proprietors. We’ve lived through the disasters of unregulated social media for society. Perhaps we can approach AI a bit older and wiser, and warier of the optimistic promise of VCs.
While humor boosts likability and success with human recruiters, it risks backfiring with AI interviewers. Candidates must adopt a ‘bilingual’ strategy to avoid being penalized, say Jamie Gloor, Cecily Cooper, T Bradford...
Audio available
Michael Wade and Massimo Marcolivio offer eight ways for healthcare companies to deliver better outcomes for patients and providers. ...
How CHROs can help the workforce navigate AI-driven change by building trust, improving the employee experience, and leading cultural change in HR and talent management. ...
Companies seeking to automate middle management risk eliminating capabilities that algorithms cannot replace. Leaders must identify tasks requiring practical and ethical judgment....
Explore first person business intelligence from top minds curated for a global executive audience