As AI rapidly takes over execution tasks such as coding and analysis, an assessment says the human role could shift toward deciding what to do and setting priorities.

On Sept. 26, IT outlet Business Insider reported, based on OpenAI materials on how its internal research organisation uses AI, that human judgement and decision-making skills could become more important in the AI era.

OpenAI said the biggest increase this year in researchers' use of AI agents was in writing research and infrastructure code, technical support, monitoring experiment runs and analysing results. Automation is moving quickly in work where goals and outcomes can be measured relatively clearly, such as writing code or fixing bugs. In mid-August, across the research organisation, AI agents handled work equivalent to 3.1 days for each day of human labour time.

OpenAI said AI still plays a small role in higher-level planning tasks such as deciding what to research, which projects to continue and where to allocate resources. OpenAI explained that people remain responsible for setting research priorities, judging which ideas and results to pursue, and deciding whether to scale, halt or deploy systems.

AI is taking on complex tasks, but human involvement is still needed. In OpenAI’s analysis, more than half of cases in which AI successfully performed tasks estimated to take a person 4 to 8 hours involved at least one instance of human intervention.

The outlet analysed this as showing that work value in the AI era could shift from execution capabilities that simply produce output to the ability to judge what work should be done and make decisions while considering context, risks and trade-offs.