As machine learning systems proliferate, new job roles are emerging at the human-AI boundary. Incanters will specialize in prompting models effectively, while process engineers develop quality control systems to catch hallucinations. Statistical engineers will measure and control ML variability, similar to how psychometricians model human behavior. These roles address the unpredictable nature of current AI systems.

Model trainers will combat the growing problem of contaminated training data as internet content quality deteriorates. Companies may hire subject-matter experts to create uncontaminated training materials and catch subtle errors. Meanwhile, meat shields will serve as human accountability points when AI systems fail, taking responsibility in legal and public domains that algorithms cannot handle.

These emerging professions reflect our struggle to integrate AI into workflows while maintaining quality and accountability. As one industry veteran noted, this represents "the largest harvesting of human expertise ever attempted" — with all its benefits and potential exploitation. The demand for haruspices who can interpret model behavior when systems malfunction will only grow as AI becomes more prevalent.