Modular AI Boosts Accuracy, Cuts LLM Confusion

Cognizant's Risto Miikkulainen on Why Multi-Agent Models Beat Single LLMs Michael Novinson (MichaelNovinson) • November 27, 2025

Monolithic artificial intelligence models struggle with complex reasoning and frequent errors. Breaking them into smaller, specialized agents makes problem-solving more accurate and manageable. Modular systems handle long reasoning chains more effectively by assigning fine-grained tasks to expert agents, said Risto Miikkulainen, vice president of AI research at Cognizant AI Lab and professor of computer science at the University of Texas.

Each module performs a targeted function, and the system orchestrates their collaboration to deliver better outcomes. Modular designs also simplify upgrades. "You can take one agent and replace it with something that's new, better, and the rest of the system will know what to do with it if it plays the same role," Miikkulainen said.

"Modularity has been in computer science as one of the fundamental principles. But it works well [with ...