Google DeepMind redefines Gemini as AI agent, not chatbot

Image credit: Search Engine Journal

Google DeepMind is evolving its Gemini platform from a conversational chatbot into an artificial intelligence agent capable of performing actions for users, aligning with the company’s broader vision for agentic AI.

This strategic shift represents a significant development in Google‘s AI trajectory, moving beyond question-answering systems to create AI that can actively collaborate and execute tasks, a goal informed by advancements in coding and software engineering.

SVP and Chief AI Architect Koray Kavukcuoglu said Google DeepMind now views Gemini primarily as an AI agent rather than solely a language model or chatbot.

Kavukcuoglu stated that the ultimate objective for Gemini is to develop a system that can take actions for and alongside humans.

He noted that while architectural innovation is ongoing, he declined to provide further specifics regarding these developments.

Google’s emphasis has increasingly moved toward agentic AI, incorporating insights gained from the development cycles of Gemini 3.0, 3.6, and 3.7 into this transition.

Although AI capabilities are experiencing revolutionary changes, Kavukcuoglu explained that the fundamental principles, such as deep learning, pre-training, reinforcement learning, and optimization, largely remain consistent.

The environment in which AI operates has undergone substantial changes, requiring models to infer user intent, manage ambiguity, and effectively collaborate with human users.

Google has made considerable progress in its understanding of agentic actions and complex workflows, particularly through the development of Gemini 3.5, according to Kavukcuoglu.

Kavukcuoglu said the most important improvement for artificial intelligence would be to enhance the intelligence of models, which would result in more intuitive and superior performance across a wide range of tasks.

This evolution underscores Google’s commitment to developing AI systems that are not just reactive but proactive and capable of deeper integration into user workflows.

Source: Search Engine Journal