Google is undertaking a substantial restructuring of its artificial intelligence division, according to MIT Technology Review. The company is tightening oversight of its DeepMind unit—acquired 12 years ago—following a period of internal challenges, including delays to a flagship AI model and reported morale issues.

As part of this organizational shift, Demis Hassabis will transition from his day-to-day management duties at DeepMind to a broader role as the unit’s chairman and Alphabet’s chief scientist. Leadership of DeepMind will be assumed by CTO Koray Kavukcuoglu, who takes the title of senior vice-president. Simultaneously, veteran chief scientist Jeff Dean is exiting Google after a 27-year tenure to establish a new venture, Discovery Loop, alongside three former colleagues. Discovery Loop aims to automate scientific research processes, with Google participating as an early investor. Financial strain has also impacted the division, as Google AI recorded cash flow negative results in the latest quarter for the first time on record.

AI Leadership and Operational Changes

Google is now consolidating its AI leadership teams within California. Strategically, the company is pivoting its development focus from specialized, narrow applications—such as the protein-folding tool AlphaFold—toward agentic AI systems capable of conducting research with greater autonomy.

Why It Matters

This leadership pivot signals a departure from the experimental silo model that characterized the early rise of deep learning labs. By concentrating authority in California and pushing for autonomous agentic systems, Alphabet is attempting to bridge the gap between academic-style breakthroughs and direct product integration. The transition from specific tools like AlphaFold to autonomous research agents suggests that the industry is entering a phase where the value proposition shifts from 'what the model knows' to 'what the model can autonomously execute' to reduce overhead in long-cycle development workflows.