Google launches the Gemini agent, one universal agent for enterprise work
Google Cloud used its Gemini at Work event, held before 700+ customers at the historic NASA Hangar One, to launch the Gemini agent. Sundar Pichai described it as 'a single, universal agent for work that has all of your business context.' It answers questions, does knowledge work, generates images and media, and writes and runs code, all from one prompt box. The agent connects to personal workflows, systems of record and enterprise controls.
The architectural bet is unification. Rather than shipping separate assistants for docs, data, code and media, as most rivals have, Google is collapsing them into one agent that routes internally. Pichai listed 'Unified Agent' as the first of several core architectural principles, with governance and enterprise context integration as the others. For IT buyers, the pitch is one surface to secure and govern instead of a sprawl of copilots, a real pain point as companies discover dozens of shadow agents.
The competitive framing is unavoidable. A day earlier Satya Nadella pitched 'Copilot as a new OS for work, with a governed infinite SaaS factory built in,' and said agentic systems of record 'can reset business SaaS.' OpenAI is pushing ChatGPT Work into companies like Oracle, while AWS is shipping agent runtimes and managed agents built with OpenAI. Reuters noted the launch lands as the 'work AI race heats up.' The fight is now over who owns the default agent employees open each morning.
The caveats: the announcement is long on principles and short on pricing, benchmarks and availability details in the sources. A universal agent also means a large blast radius if permissions are misconfigured; Google's own Gemini for Mac is reportedly testing a full-access mode that can act without approval, which has already raised security worries. Watch for pricing per seat versus per task, how the agent handles conflicting data permissions across connected systems, and early customer deployments beyond keynote demos.