JBS Dev advises prioritizing AI cost and portability over model capability

joe rose, president of strategic technology provider jbs dev, challenged the vendor narrative that requires massive data lakes before adopting generative and agentic ai. he argued that modern large language models can effectively process imperfect data, citing a client example where ai reconciled messy medical billing records involving pdfs, images, and mismatched fields. rose emphasized that while a human in the loop is necessary for guardrails, automation rates can grow incrementally. he predicted that future ai discussions will center on reducing costs and enabling models to run on laptops rather than data centers, citing diminishing returns from training on existing internet data.