Microsoft Copilot agents run 4x faster and use 30% fewer tokens: the bottleneck is knowledge, not the model

Microsoft has achieved a technical breakthrough for its workplace AI agents: without upgrading the underlying large model, response speed rises to 4 times the previous level and token consumption falls by about 30%. The bottleneck is knowledge, not the model. Project information is scattered across emails, Teams meeting transcripts and technical documents, and agents have until now had to search these fragments of raw data one by one, which is slow and expensive. The new solution is the Work IQ knowledge compilation engine. Built on SharePoint, it automatically converts information scattered across an organization into a structured knowledge base that agents can retrieve directly and efficiently. It covers scenarios such as cross-document Q&A, meeting summaries, project status roundups and customer material preparation, and is now rolling out to Frontier Preview tenants. The next bottleneck for agents may not be bigger models, but better organization of knowledge. > Source: [read the original](https://finance.sina.cn/stock/jdts/2026-10-15/detail-inivhtui7364667.d.html). This text is a summary for reference only.