One of the most optimistic claims about artificial intelligence is that it will save academics time. It can draft, summarise, code, format, search, translate and reorganise information at a speed that would have seemed impossible only a few years ago.

But there is a harder institutional question: when AI saves time, who gets that time back? In many academic settings, efficiency gains are absorbed as higher expectations rather than returned as protected time for deeper work.

The acceleration problem

Academic careers already run on visible productivity: publications, grants, student completions, service, teaching innovation, impact and collaboration. When a tool increases the speed of one part of this system, the system rarely responds by lowering workload. It usually raises the pace.

AI will not be an exception. A researcher who uses AI to draft faster may soon find that faster drafting is simply assumed. A supervisor may expect quicker revisions. A research team may expect broader searches. A committee may expect more polished documents in less time.

The bottleneck becomes judgement

If AI absorbs more of the execution, the remaining work becomes harder, not easier. The real bottleneck becomes direction: which question is worth asking, which result matters, which design tests the real issue, and which argument is genuinely yours rather than merely fluent.

These are not mechanical tasks. They require field knowledge, judgement, taste and accountability.

Experts gain leverage; novices may lose development

For an expert, AI can function like a cognitive exoskeleton. The expert already knows the field, the standards and the shape of good work. They can reject fluent nonsense and use AI to scale their thinking.

For a novice, the same tool can become a trap. If AI produces research questions, theory summaries, supervision responses or committee answers that the learner cannot defend, the surface of academic work appears before the underlying development has happened. The danger is premature fluency.

The artefact was never the whole point

A thesis, article, assignment or proposal is an artefact. It matters, but it was always evidence of a process. The deeper purpose is the development of someone who can make decisions, justify them, revise them and stand behind them.

This is why viva-style questioning, supervision conversations and proposal reviews remain important. They ask the researcher to explain why they framed the question this way, what they considered and rejected, and what they would now do differently.

Ask about decisions, not only prompts

Much of the conversation about AI use focuses on prompts. That is useful, but incomplete. The more important level is the decision tier.

  • Why did the student or researcher ask that question at that point?
  • What did they do with the response?
  • What did they verify or reject?
  • What data did they expose, and can they account for where it went?
  • Can they defend the final argument?

Two questions for responsible use

For research training, two questions are useful: is this use of AI building me as a researcher, and can I defend the decisions that shaped this work? If both are satisfied, AI use is more likely to support academic growth. If either fails, the decision deserves reconsideration, even if no formal rule has been broken.