Google DeepMind announced expanded capabilities for Gemini API's Managed Agents, emphasizing production-readiness for enterprise deployments. The update centers on Gemini 3.6 Flash, Google's faster, more efficient model, which now powers agent workflows designed to handle complex multi-step tasks without developer infrastructure overhead. While specific pricing deltas remain undisclosed, the managed service model—where Google handles orchestration, memory, and tool integration—targets companies seeking to reduce operational complexity. Example workflows include customer service agents that autonomously process refunds, route escalations to human teams, and log interactions, all within compliance guardrails. The hooks architecture allows developers to inject custom logic at decision points, giving enterprises control without building from scratch.
Complementing the agent push, Google is weaving third-party integrations directly into Search's AI Mode, letting users connect apps like Gmail, Google Workspace, and—critically—external services through OAuth-style secure connection protocols. Technical implementation relies on standardized APIs and token-based authentication, meaning users grant Search read/write permissions without sharing credentials directly. Google has not yet publicly confirmed which enterprise SaaS platforms ship live integrations, though the architecture suggests Slack, Stripe, and other REST API-first services could follow. The pitch is friction-free: users planning a dinner party or booking concert tickets can now pull real-time inventory and purchase directly from Search results, powered by Gemini's reasoning.
This dual expansion—managed agents for B2B developers, app connections for consumer Search—reflects Google's strategy to embed Gemini across its product ecosystem. The Oracle partnership announcement, framing OCI as enterprise AI's neutral ground, signals competitive pressure from other cloud vendors and reinforces Google's need to offer developers escape hatches and flexible deployment options. Success hinges on adoption velocity and transparent pricing; without clarity on managed agent costs versus Gemini API tokens, enterprises may hesitate to commit workloads. Google's move is defensive and offensive simultaneously: defending Search against AI-native competitors while attacking the agent workflow space where Claude and OpenAI have gained traction.