Model intelligence is already at expert level — deployment is the only bottleneck

Even in a basic synchronous chat interaction — not a full autonomous agent loop — you can give these frontier models the hardest, meatiest problems in domains like management consulting and receive genuinely expert-level answers. The model intelligence is clearly there. The models are also smart enough to coherently execute across multiple turns with lots of tools and context. So the real question is no longer 'are the models smart enough?' It's purely about how quickly we can deploy agents into every role across every industry.