
Illustration by Mariano Pascual
As companies invest in AI, they often gravitate toward the most powerful models and broadest platforms available, without considering whether those models and platforms are the best fit for the tasks they need to accomplish. But, unsurprisingly, most businesses aren’t aiming to solve arbitrarily complex problems. They have discrete tasks that need to get done: catching fraudulent transactions, summarizing meetings, routing service requests, predicting equipment failures, diagnosing problems, or writing code. While many companies have contracted with frontier AI labs to do these tasks—OpenAI, Anthropic, and the other brand names of the AI boom—there may be better ways to meet specific needs without triggering the many risks and costs associated with frontier AI.
