Physix Frontier badgePhysix Frontier
PublishChinese | EN

Physix Frontier · News Briefing Card (IT Home · Oct 15, 2026)

Microsoft Copilot agents get 4x faster, cut token use by 30%

KEY FACTS

  • Microsoft has introduced a new technology for AI agents such as Copilot, boosting response speed by up to 4x.
  • The technology reduces token consumption by about 30% without upgrading the underlying large model.
  • It relies on the SharePoint and Work IQ knowledge compilation engine, turning scattered information into a structured knowledge base.
  • The feature has rolled out to Frontier Preview tenants, limited to pay-as-you-go users.
  • Microsoft plans to extend it to custom agents in Copilot Studio and develop automated maintenance capabilities.

KEY DATA

4xResponse speed increase
about 30%Token consumption reduction

PHYSIX OBSERVATION

Instead of swapping models, Microsoft is boosting efficiency through knowledge compilation, finding a new cost-side lever. The biggest pain points for enterprise AI adoption are latency and token bills, and this move directly cuts about 30% of the expense. The signal is clear: agent competition is shifting from model prowess to data pipelines. Whoever turns fragmented enterprise information into structured data first will win the office scenario.

Source: IT Home report