A handful of startups now claim their artificial intelligence systems can out-predict even the most skilled human forecasters, tackling questions ranging from stock market swings to geopolitical flashpoints like the Strait of Hormuz, and even how a Shakira single might chart. The claim sounds audacious, but the early evidence is prompting serious attention from hedge funds, governments, and corporations alike. By the end of the decade, according to Dan Schwarz, cofounder of the forecasting company FutureSearch, current human-led decision-making will look like "complete Stone Age thinking" by comparison.
The technology rests on a simple but powerful idea: feed an AI model enormous volumes of public and private data, then ask it to assign probabilities to future events rather than make flat predictions. This approach mirrors what "superforecasters" - a small class of individuals with a documented record of unusually accurate predictions - have long done manually. What's changed is the speed and scale at which machines can now replicate that skill. In early September, two AI models took the top spots in the Metaculus Cup, a competition pitting algorithmic and human forecasters against one another, marking what many in the field consider a symbolic turning point. As with any technology handling sensitive data and high-stakes decisions, the question of accountability looms large, which is precisely why audits matter more than promises when startups market these systems to institutional clients. why audits matter more than promises
From Trading Floors to Government Offices
Algorithmic trading has shaped Wall Street for more than a decade, but AI forecasting threatens to push automation further still. Toby Shevlane, cofounder of the startup Mantic, said the technology is "probably already having a small impact" on trading decisions among early-adopter clients. Theo Summer, who cofounded the startup Preseen with Veniamin Veselovsky, reportedly turned a modest stake into a seven-figure sum within months using the prediction platform Kalshi - an example that has caught the attention of both retail traders and institutional players.
Prediction markets such as Kalshi and Polymarket, now used by millions of Americans, are increasingly shaped by AI-driven participants. Veselovsky warned that ordinary retail investors, lacking access to expensive AI tools, risk being squeezed out entirely as these platforms become "dominated by AI forecasting systems." The imbalance raises familiar questions about fairness in markets that were originally designed as accessible, low-barrier spaces for public speculation.
Beyond Finance: Policy, Medicine, and Risk
The reach of these tools extends well past trading screens. Mantic has supplied a major American company with probability assessments of regulatory shifts affecting its industry, and has also worked with the British government on related forecasting projects. Schwarz noted that other clients are exploring applications in medical research, using AI predictions to help prioritize which treatments deserve funding. The most sophisticated systems no longer simply answer questions posed to them - some now generate their own questions about future developments and then estimate the likelihood of various outcomes.
A Technology With Global Stakes
The democratization of forecasting power carries real risks alongside its promise. Veselovsky expressed cautious optimism about the technology's potential benefits while acknowledging deeper anxieties: "I'm also a little bit scared because, you know, maybe Russia or China has better systems." That tension - between genuine analytical value and the unsettling prospect of asymmetric access to predictive power - is likely to define debates over AI forecasting as it moves from niche competitions into boardrooms, trading desks, and government offices worldwide.