OpenAI announced a levels to intelligence approach they have to classifying their AI systems. In short, the Levels range from 1 to 5, starting at chatbot, then reasoned (GPT-4 reportedly falls somewhere between the first two levels), then notably, agent, innovator, then AIs capable of doing the work of organizations. Admittedly it's not an explicit road map, but that OpenAI views agency as a potential next step as opposed to remaining in a more tool AI orientated space seems concerning.