Google DeepMind CEO Demis Hassabis said advanced general intelligence is still not here, but he believes current AI systems are becoming more capable through better workflows and greater agency. His remarks came during a conversation that focused on how today’s tools are evolving and what still separates them from human-level intelligence.
Hassabis described AGI as a system that could match or exceed human abilities across many tasks, but he suggested that reaching that point will take time. While he did not give a firm date, his comments indicate that the industry should expect continued progress before anything resembling full general intelligence arrives.
A major theme of the discussion was the role of workflows. Hassabis argued that AI products become far more valuable when they are embedded into practical step-by-step processes rather than treated as isolated chat tools. In his view, the biggest gains for users often come from designing systems that can help with structured tasks, coordinate actions, and move work forward with less manual effort.
He also pointed to agency as an increasingly important feature of modern AI. That term generally refers to systems that can take initiative, carry out multi-step goals, and operate with some level of independence. Hassabis suggested that current tools are moving in that direction, even if they remain far from the kind of broad autonomy associated with AGI.
The DeepMind chief’s comments reflect a common view in the AI sector: today’s models are powerful but still limited. They can generate text, answer questions, and assist with coding or planning, yet they still depend heavily on human direction and oversight. Hassabis’s remarks suggest that future progress may come less from a single dramatic breakthrough and more from combining model improvements with better systems design.
That distinction matters for how companies are deploying AI. Instead of waiting for fully general systems, many developers are focusing on narrow but useful applications that improve productivity now. Hassabis’s emphasis on workflows and agency fits that trend, highlighting how AI can deliver value before it reaches anything close to human-level reasoning.
The conversation also underscored how much attention in AI has shifted from abstract predictions to practical implementation. Businesses and product teams are looking for systems that can reliably perform tasks, integrate with software, and support users across many steps of a process. Hassabis’s comments suggest that this applied approach may define the next phase of AI adoption.
At the same time, the debate over AGI timing remains one of the most closely watched topics in the field. Some industry leaders have predicted rapid progress, while others argue that current systems still have major shortcomings. Hassabis’s stance appears more measured. He sees continued advancement, but not an immediate leap to machines that can fully generalize across domains.
For now, his message was that the path forward is likely to be incremental. AI tools are becoming more capable and more autonomous in limited settings, but the broader goal of AGI remains ahead.