Will AI make firms smaller?

Business startupstechnologyTop News Will AI make firms smaller? File Photo: File Photo: Illustration Shows Ai (artificial Intelligence) Letters And Robot Hand Miniature Relevant News Will AI make firms smaller? 8 July 2026 Holidaymakers beware: What the EU’s new biometric borders mean for Cyprus travelers 8 July 2026 Two more cloudy days until skies clear on Friday 8 July 2026 newsroom 8 July 2026 FacebookXWhatsAppEmailPrintViber Demetrios Zoppos on technology, innovation, and Cyprus’s place in the world that’s coming. A hundred years idle Let me begin with the fax machine, because it is stranger than it looks. Alexander Bain patented an electric printing telegraph that could reproduce an image over a wire in 1843, three decades before Alexander Graham Bell patented the telephone. Then it did almost nothing for a hundred years, resulting in mass office adoption only in the 1980s, not because anyone had improved the idea but because the conditions it needed had finally assembled: cheap electronics, an agreed standard, and a telephone network already built and paid for that it could ride for free. A capability can be complete and still lie idle for a century, waiting not for a better version of itself but for the world around it to be ready. Artificial intelligence has the same shape. The first mathematical model of a neuron dates to 1943, a hundred years after Bain’s patent; the algorithm that lets such networks learn was written down in the 1970s and largely ignored. AI waited, as the fax did, not for a cleverer idea but for two things others would build for their own reasons: vast computation and vast data, both delivered by the internet. It belongs to a small class of technologies, alongside electricity, the semiconductor and the cloud, that do not merely perform a task but change how nearly every other task is done. The interesting question about such a technology is not what it is, but what it does to an economy once it arrives. For that, the best guide is an economist writing decades before the first computer. Why firms exist In 1937 Ronald Coase asked a question so simple it sounds naive, and won the Nobel Prize for it. If markets are so efficient, why do firms exist at all, rather than individuals contracting with one another task by task? Because using the market is never free. Every transaction carries costs beyond the price itself: finding the counterparty, negotiating, writing the contract, checking that the other side delivered, enforcing the deal when they do not. Firms exist to swallow those costs. You bring an activity in-house precisely when doing it through the market would cost more in searching, haggling and policing than doing it yourself, and the boundary of the firm sits where those two costs balance. This is the thread that runs through history. Whenever a technology changes the cost of transacting, it moves that boundary. The telegraph made possible the large managerial firm; the shipping container gave us the global supply chain; the internet drove the cost of dealing with a counterparty close to zero, letting firms specialise and buy in the rest, while a few companies grew vast by owning the marketplace itself, as Amazon did in retail and Google in advertising. Even the fax belongs here: its long wait was never about the machine, but about the network it needed to carry it. How AI is different AI departs from all of these in one sharp way, and it has little to do with the usual talk of speed and disruption. Every earlier technology lowered the cost of moving something between people, but at every point in a transaction a person still had to do the analytical work: read the contract, weigh the risk, monitor the supplier, check the quality against the invoice. Transactions are expensive, contracts always incomplete and monitoring always costly, precisely because that work could only be done by people. AI is the first foundational technology to attack the cost of that work itself. It is worth being precise about what it does. AI is not intelligent in any meaningful sense, and it is unlikely to acquire taste or judgement. What it does well is perform, at enormous speed and low cost, the brute-force computation and pattern-matching beneath much of what we loosely call thinking: the checking, comparing, correlating and sifting of one document against another. That is exactly the analytical labour that made transactions expensive. AI does not replace the human judgement at the centre of a deal; it strips away the costly work that used to surround it. And because that work sits on both sides of the ledger, inside the firm as the cost of producing and between firms as the cost of transacting, making it cheap disturbs the whole structure at once. Two forces, pulling opposite ways The tempting conclusion is that cheaper cognition means smaller firms. If three people can now orchestrate what a decade ago demanded three hundred, the minimum size of a serious business must fall, and the economy should fragment. There is real truth in this: AI lowers the fixed cost of building a capable operation and the cost of transacting, so buying from the market beats building in-house, and both push towards more and smaller firms. Taken to its limit, this is the logic behind the one-person unicorn, a billion-dollar company run by a single individual directing a fleet of AI agents. It has not happened yet, but it is no longer fanciful. It is Coase’s mechanism carried to its conclusion: if a lone founder can transact at almost no cost with machines that perform every function a company once had to house, the firm recedes until there is almost nothing left inside it but the founder. But the same technology pushes the other way with equal strength, and this is the half the original version misses. AI also drives the marginal cost of production towards zero for anything a machine can generate, and markets with near-zero marginal cost do not fragment. They concentrate. Whoever has the best product serves unlimited demand at almost no additional cost, so scale becomes decisive and a few winners take most of the market. This is the familiar logic of software, and it is why the digital economy, despite low costs of entry, has produced some of the largest firms in history. The incumbents, too, are not spectators: the same tool that lets a newcomer produce cheaply lets an established firm reorganise and defend. Put as a question of firm size, there is no single answer. The threat of entry There is a better answer if we stop asking how many firms a market has and ask instead, as the economist William Baumol did, how easily a new one could enter. What disciplines a market, he argued, is not the number of firms in it but the threat of entry. A market served by a single large firm can still behave competitively, holding prices down and quality up, if entry is cheap and the incumbent knows a rival could appear at any moment. Baumol called such a market contestable. Concentration and competition are not opposites: a market can be concentrated in structure and fiercely competitive in conduct, provided the door is open. This fits the two forces together. The concentrating force may leave many markets in the hands of a few large, AI-enabled firms, but the deconcentrating force, the collapse in the cost of assembling a capable operation, is exactly what raises contestability. If a small team can now build cheaply what an incumbent needed a large organisation to produce, that incumbent lives under permanent threat of entry even while it remains large. The likely result is neither a world of many small firms nor one of comfortable monopolies, but markets that stay concentrated in shape while becoming far more contestable in behaviour. A caution, though: the internet also lowered the cost of entry and still produced strong concentration, because data and network effects became new fixed costs that rebuilt the old moats higher. AI may do the same, as the cost of training the models becomes the new barrier. So the claim is narrow: AI makes a particular band of markets, those whose barrier was the cost of assembling human cognition, contestable for a period, and the question is who moves while the window is open. Cyprus’s window This is where the argument comes home. Cyprus has no place in the concentrated layers, where the residual costs are physical, regulatory, or measured in billions of dollars of computation and energy. We will not train the models, and should not try. Our opening is in the markets AI has just made contestable: the knowledge-intensive niches where the old barrier was the expense of assembling talent at scale, and where a small, sharp, judgement-dense team can now challenge incumbents who were until recently protected by their size. The strategy is precise: find the markets AI has just prised open, and enter them in the interval before the incumbents finish reorganising, because that interval is the opportunity and it will not stay open indefinitely. The warning travels with the opportunity. Contestability cuts both ways: the same forces that let a Cypriot firm contest an incumbent’s market let the next entrant contest the Cypriot firm’s. A cost advantage in a technology available to all is no advantage for long. The one position that endures is what a newcomer with the same tools cannot instantly copy: a genuine relationship with customers, accumulated trust, a specialist reputation earned over years in a niche, as with Israel in security or Estonia in digital government. If the firm is being pulled towards its smallest viable size in some markets and defended at scale in others, and the scarce ingredient in either case is no longer capital or scale but judgement, then everything turns on who exercises that judgement. That is the founder, and it is where this column will turn next. Demetrios Zoppos is a Partner at 33East. The views expressed are exclusively his own. Subscribe to our Newsletter Latest News Holidaymakers beware: What the EU’s new biometric borders mean for Cyprus travelers Two more cloudy days until skies clear on Friday Who curates the curators? EU parliament to vote on condemning 1974 Cyprus sexual violence by Turkish army The World Cup, Trump’s newfound woes, and the ridiculous reality show The (Un)Housing market in Cyprus They are human beings Follow en.philenews on Google News and be the first to know all the news about Cyprus and the world.