← back
Does GenAI "belong" to data scientists? — Phil Hetzel, Braintrust
Takeaway
Organize agent development around product understanding and evaluation rigor rather than automatically inheriting traditional ML ownership.
Summary
- Traditional enterprises often assign generative AI to existing ML teams, while AI-native companies tend to use smaller cross-functional product and engineering teams.
- Foundation-model providers already perform base-model training, shifting application work toward prompts, context, integration, and product-specific evaluations.
- Natural-language behavior changes let people close to user problems contribute directly to agent development.
- Data scientists bring valuable model-risk awareness and testing discipline, but existing ownership of models alone does not settle who should own agent products.
ai-teamsdata-scienceproduct-engineering
Original description
At most traditional enterprises, GenAI got handed to the ML platform team because it had AI in the name. Phil Hetzel from Braintrust argues that was the wrong move, not because data scientists lack value, but because Anthropic and OpenAI already ran the data pipeline. What is left is prompt and context engineering, distributed systems, human annotation, and functional evaluation across a much broader surface area than precision and recall. The mistake is isolating it to one team. The answer is a diverse one. Speaker info: / philliphetzel