Universities Urged to Build Students’ AI Evaluation Skills

Generative AI is moving from experimental projects into routine business operations, but model performance can shift with data, context and time. Companies therefore need more than workers who can build models or write prompts: they need people who can test whether AI systems are accurate, reliable and fit for a specific task. Such “eval” skills are becoming essential as businesses deploy AI in real-world workflows.

The latest report urges universities to give students hands-on experience designing tests around practical requirements, setting performance benchmarks, examining failures and refining models from the results. The approach would prepare graduates for roles involving continuous validation and tuning after an AI system is deployed. The report did not identify a specific university, funding amount or implementation date, presenting the proposal as a broader curriculum priority.