AI-DRIVEN TRANSFORMATION

The shift already underway

AI has moved out of the technology sector. Radiologists work with models that flag anomalies. Lawyers use systems that surface relevant case law. Farmers receive yield predictions from satellite data. Materials scientists screen candidate compounds computationally before touching a lab bench.

In almost none of these cases has the profession disappeared. What has changed is the distribution of effort. The routine, pattern-matching portion of expert work is increasingly automated, which shifts the human contribution towards judgement, framing and verification — deciding which questions are worth asking and recognising when an answer is wrong.

What becomes more valuable

• Problem framing — models answer the question they are given. Working out which question matters remains stubbornly human.
• Verification — AI systems fail confidently and without warning. Knowing a field well enough to catch a plausible-sounding error is now a core skill.
• Cross-domain thinking — the valuable work increasingly sits between fields, where models have thin training data and few examples to imitate.
• Physical and experimental skill — someone still has to build the apparatus, run the experiment and interpret what actually happened.
• Ethical judgement — deciding what should be built is not a technical question and cannot be delegated to a model.

How we prepare students

Preparing students for this does not mean teaching them to use AI tools. Those change every few months. It means building the underlying capabilities that keep their value.

• Teaching AI as mechanism, so students can reason about where a system will fail rather than trusting it by default
• Prioritising experimental work, where results have to be obtained and defended rather than generated
• Structuring sessions around open-ended problems with no single correct answer
• Deliberately crossing domains — the AI-and-materials or AI-and-biology intersections, where the interesting work is
• Treating ethics as part of the engineering, not a separate module