Overview
- NVIDIA’s Jensen Huang and a16z’s Marc Andreessen characterize generative AI as an infrastructure-scale shift, with Huang noting some AI-native firms report token generation gross margins near 90%.
- A new Google report urges companies to build an AI-ready workforce and integrate agents under human control, moving beyond tool purchases to organization-wide learning and adoption plans.
- Enterprise case studies highlight that agents often fail without clear ownership and human-in-the-loop guardrails, with Gartner finding only 15% of IT application leaders pursuing fully autonomous agents and surveys showing users still want oversight.
- Security leaders warn that autonomous, machine-speed behaviors outpace periodic audits, citing Anthropic research on automated attack steps and calling for continuous posture management, visibility into models and agents, and strict management of non‑human credentials.
- An AWS re:Invent panel advises shifting to behavioral monitoring, short‑lived and tightly scoped access, explicit bounds on agent autonomy, and the use of automated reasoning and new tooling such as AWS’s Security Agent to embed controls during development.