AI-first or Human-first startups?

The generative AI-focused startup landscape is experiencing a thrilling surge, with startups mushrooming from all corners. Using 3rd-party LLM API, startups are (re)inventing many things these days. I have been investigating generative AI for two years now. The transformer-based LLMs are fantastic at getting you started in the right direction or keeping you compliant with the norms of existing knowledge. They are good at reducing chores, but they could be more innovative. As AI shifts at an arms-race speed, we will soon see new architecture and more creative AI. My approach is to deploy AI in a supportive role, where AI agents take on tasks that are mundane or overly burdensome for human operators. This not only frees up valuable human resources but also enhances overall productivity. So when I pitch my to-be-launched product, ShowNotes, I tell friends that the AI agents are there to help with security, SEO, classifications, organization, and visibility throughout the system. They don't write for you. They may recommend disclaimers or suggest alternative approaches, but they are secondary in these roles. Human-first systems that focus on human users to be creative will remain very important, and the AI-generated content phase will become a thing that we do not look forward to consuming but is there to fill some gap. For example, I loved AI art initially and spent hours playing with art generation; now, I cannot stand it. Human imperfections, subtleties, and nuances are so much more valuable nowadays.