A number of quant firms are among the largest global consumers of commercial LLMs

The only area you absolutely can’t use LLMs is in sub ms latency