OpenAI's Sam Altman outlines 'proactive AI' as the next phase after chat models and agents, citing rising costs and user confusion as current challenges

At an OpenAI enterprise event, CEO Sam Altman presented a three-phase thesis for AI development: chat models (like ChatGPT), agent-based systems (like Codex), and 'proactive AI' that runs constantly in the background without requiring users to initiate requests.

Altman identified two major obstacles to AI adoption: costs have become 'a huge issue' (citing Uber's example of burning through its entire annual AI budget in Q1 alone), and most users struggle to understand what AI can do or how to use it effectively.

OpenAI is developing a super app combining agentic features from Codex with ChatGPT and other tools, and pursuing 'proactive AI' that would run continuously with access to a company's full context—though this shift will require changes to security protocols, compute allocation, and data protection practices.