The jagged frontier is a way of describing the uneven landscape of what AI can and cannot do well. The term comes from a September 2023 Harvard Business School working paper, “Navigating the Jagged Technological Frontier,” by Fabrizio Dell’Acqua, Ethan Mollick, and colleagues, conducted with Boston Consulting Group. The metaphor is that the boundary of AI capability is not a smooth line; it juts in and out unpredictably, so two tasks that look equally hard to a human can sit on opposite sides of it, with the model excelling at one and failing at the other.
The paper grounded the idea in a large field experiment with 758 BCG consultants randomly assigned to work with or without GPT-4 across realistic consulting tasks. On tasks inside the frontier, consultants using AI were dramatically faster and produced higher-quality work. But on a task designed to fall just outside the frontier, where the AI’s confident output was subtly wrong, consultants using AI were more likely to reach the wrong answer than those without it. The danger is precisely that the frontier is invisible: the model sounds equally fluent on both sides, so users cannot tell when they have crossed into territory where it should not be trusted.
This concept matters because it reframes the practical challenge of adopting AI. The hard part is not whether AI is capable, but learning where its capability ends, since the jaggedness means blanket trust and blanket skepticism are both wrong.