Artificial intelligence research workflow automation can help search literature, extract claims, organize sources, generate hypotheses, plan experiments, run analyses, and capture evidence. The workflow should preserve provenance so conclusions can be traced to sources, methods, intermediate results, and review decisions.
Automation can accelerate iteration, including research about future AI systems, but it does not turn generated suggestions into verified discoveries. Incomplete runs, weak sources, failed experiments, and negative results must remain visible rather than being optimized out of the reported evidence.
