A new class of AI agent can hold a task open for days or weeks, moving across systems and channels before returning control to a human. That handback moment — when and how the agent yields — turns out to be the hardest part of the whole design. Read more
Perplexity has handed an AI agent end-to-end access to live systems — writing code, sending messages, and monitoring production with far fewer human check-ins than before. That's a meaningful governance threshold. Here's what crossing it actually requires. Read more
Salesforce has previewed two new offerings — an Enterprise AI Harness and an AI Control Plane — that tackle not just how agents are built, but how they are managed, audited, and constrained once deployed. The distinction matters more than it might first appear. Read more
What happens when one AI agent knows something another one doesn't? In enterprise deployments, that gap isn't just an inconvenience — it's a governance failure waiting to surface. Writer's Enterprise Brain points toward a more coherent architecture. Read more