AI STRATEGY cases, score historical examples and investigate unusual outcomes with dramatically less friction.”
He adds that even the process of understanding a prediction can be augmented:“ Given a prediction, a set of possible decisions and a known level of uncertainty, these systems can help explore the consequences and tradeoffs of the different choices available. And then agents extend that even further because the prediction no longer has to end on a dashboard.”
Juan José notes that even a predicted supply shortage, for example, can trigger an agent to investigate affected orders, alternative suppliers, customers and possible responses.
“ That is enormously powerful,” he says.“ It also raises the stakes. A bad prediction sitting on a dashboard can be ignored. A bad prediction connected to an agent that can change orders, contact customers or move money can propagate very quickly.”
He explains that every ingredient needed to do predictive AI well can now be enhanced – but the importance of validation, permissions, uncertainty management and human oversight increases at exactly the same time.
Uncertainty guaranteed“ The technology is evolving extremely quickly, but not uniformly,” Juan José goes on.“ Different capabilities are
aimagazine. com 75