Microsoft Copilot Researcher drafts reports then critiques them with rival AIs

Microsoft Copilot Researcher now drafts reports with one AI model and critiques them using rival AIs for accuracy, completeness, and citations. The Council feature compares multiple models. On DRACO benchmark, factual accuracy rises 2.6 points, analysis depth 3.3 points, presentation 3.0 points, and overall score 7 points. Improves reliability for business reports over single-model tools.