\n\n
Key developments this week show AI agents moving from experimental tools into the backbone of enterprise operations. Funding of $90 million for Nous Research’s Hermes platform signals a push to bring open‑source agents to Slack, Telegram, WhatsApp and other business channels, while Oracle’s Fusion Claw embeds orchestration directly into its ERP suite, letting firms codify policies, risk limits and decision rights for agents. Anthropic’s internal report highlights the growing safety concerns around agents that can accidentally run commands or submit sensitive forms, underscoring the need for tighter governance as agents become more capable. At the model level, Anthropic’s Claude Haiku 5.5 offers a cheap, 1‑million‑token context window for high‑volume classification and sub‑agent work, making it easier to deploy agents at scale. Consumer‑focused moves, such as Meta’s Muse iPad app and TikTok’s Shopping Assistant with one‑click checkout, illustrate how agents are being packaged for everyday user experiences across devices. Together, these trends point to a future where AI agents are woven into core workflows, requiring robust security, governance and cost‑effective models to unlock their full potential. [[1]]