Google Cloud says AI agents require new ways of working

At AI Day in Stockholm, Google Cloud urged companies to move beyond AI experiments and change how work gets done. Its message was that deploying more capable models alone will not deliver results; companies also need to rethink processes and automation.

James Chalk, Google Cloud’s country manager for Sweden, said the focus should shift from what models can answer to the difficult problems AI agents can help solve. Google describes an agent-driven company as one where AI handles tasks across multiple steps. The event came days after Google Cloud announced it would add OpenAI models and the Codex coding tool to its offerings. The models are also set to become available in Google Antigravity and Gemini Agent.

Ericsson’s Razvan Teslaru said the company uses Google Cloud services for operations and product security. Ericsson’s On Demand service lets operators deploy services in 5G core networks in minutes, while real-time data helps forecast traffic and optimize energy use.

SEB’s Elizabeth Suzuki described an AI agent that helps advisers identify business opportunities and prepare client contacts. Specialized agents analyze market events, competitors, and company finances. SEB moved from idea to finished service in three months, she said. The system uses data in BigQuery, Gemini Enterprise, and a semantic layer in Looker.

Google Cloud also demonstrated Gemini coordinating six specialized agents to prepare a sales forecast and presentation using BigQuery, email, and Excel. The company said tools including Agent Sandbox and Agent Gateway are intended to limit access and check incoming instructions.