“Today, Gemini became an agent,” Google Cloud CEO Thomas Kurian said Thursday. “You give it objectives, not just instructions.” The pitch at the Gemini at Work 2026 event: Gemini plans the work, spawns temporary job-specific sub-agents, uses custom skills and tools, connects to company systems, and brings back something finished inside the documents, inbox and developer environments people already work in. Crucially for enterprise buyers, each job runs on the model that fits best, orchestrating across Google’s Gemini family and, notably, Anthropic’s Claude models today, with other leading private and open models to follow.
Google wrapped the pitch in the enterprise controls agents have lacked. Every Gemini agent gets its own identity, separate from the user’s, so policies apply to agents the same way they apply to people. Budgets are set before work starts and cover the whole cost of the job, addressing the runaway-spend problem that the Wall Street Journal recently flagged as nearly impossible for businesses to budget for.
Alphabet CEO Sundar Pichai supplied the demand evidence: nearly 80 percent of Google Cloud customers are using the company’s AI products, and nearly 500 enterprise customers have each processed over 1 trillion tokens on the platform in the last year. On model quality, Google says its latest model, Gemini 4 Argon, outperforms Claude Opus 5.5, Claude Fable 5.1 and GPT-6 Astra on the DeepSWE v1.1 coding benchmark at 77.9 percent, and that internally it has cut post-submit code rollbacks 30 percent while increasing agentic code submissions by 35 percent.
Caveats apply. Every one of those numbers is Google’s, and the company’s long history of killing products (a median lifespan of about 4.1 years across 308 discontinued products) makes the “universal agent for work” promise sound bolder than it is. But the enterprise framing is the point: the agent race is being fought as much on cost control and governance as on model quality.