In a recent reflection on the Life Sciences Jobs Plan, I asked where professional judgement now forms as AI changes early-career work. A later conversation with Cambridge Cognition made that question more concrete.
The example that stayed with us came from its PhD internship programme. The projects are real but not client-critical: substantial enough to give someone a genuine sense of the work they would do if they joined, while keeping appropriate boundaries around client risk. At the end of the placement, interns present what they have found.
In a recent cohort, interns were able to use AI as part of their work. What followed offered a useful glimpse of how early-career researchers begin to develop professional judgement around a new tool. AI could accelerate parts of the work, but its outputs still needed to be questioned, tested against existing knowledge and discussed with more experienced colleagues. Some answers looked plausible. That did not necessarily make them right.
This is where the design of the internship becomes important. The interns are working on meaningful but bounded projects, with support from experienced professionals. It gives them room to use AI in a genuine working context, encounter its limitations and develop the professional curiosity needed to decide when an output can be relied upon and when it needs to be challenged.
The question, then, is not simply whether young people can use AI. It is where they get the opportunity to develop the judgement needed to use it well.
This is a small observation from one employer conversation, but it captures a question we have been hearing more often in discussions about education, work experience and early-career development.
Much of the current discussion about AI focuses, understandably, on tool use: how people prompt, work faster and produce better outputs. Those things matter. But the internship example points to something sitting underneath prompting. To use AI well, a learner needs enough knowledge to question what is produced: to recognise when an answer is too neat, too confident or missing something.
Professional curiosity is one of Cambridge Cognition's values. Becca Day, its Head of People, described it as asking why, rather than accepting something at face value, and suggested the habit should begin well before people reach work. As AI becomes more embedded in everyday work, that habit matters more.
Cambridge brings together specialist judgement built over years of scientific, clinical and organisational practice. How that judgement begins to form, before people are expected to rely on it professionally, is the harder problem.
The design of the internship itself is therefore significant. Cambridge Cognition gives interns meaningful but bounded work: projects chosen because they matter but are not client-critical, with experienced professionals available to provide context and challenge. An intern might take something to prototype stage, with a more senior team member later deciding whether and how it becomes something the business can use.
That creates something increasingly valuable in an AI-enabled workplace: a protected environment in which professional judgement can begin to form through practice.
That balance matters because judgement does not develop through observation alone. Learners need contact with real problems. They need to see how decisions are made and how evidence is tested. They also need enough protection to learn without being placed in situations where the consequences are disproportionate.
These opportunities may become harder to find if some of the routine work through which people once gained a first foothold is increasingly automated.
AI is also changing what a polished output can tell us. Fluency is becoming easier to produce, including in applications and early evidence of work. That makes the reasoning behind the output more important. Can someone explain how they reached a conclusion? Can they say where they are unsure? Most usefully: can they show where AI helped, where it may have misled, and how they checked?
Becca Day's point was that behaviours and values are the hardest things to assess. A CV and an interview show what someone knows, and people become practised at describing it. How someone responds in a real situation is harder to see, and much of it only becomes clear after they join. She connected this back to professional curiosity: the difference between someone who asks why something is done a particular way, and someone who carries on because it has always been done that way.
The practical implication for those of us designing learning experiences is that access to AI needs to sit alongside opportunities to develop judgement in context. Young people need repeated chances to use AI on meaningful problems, test its outputs against evidence, explain their reasoning and learn from people with greater experience.
That is part of what CognateUK is beginning to explore through its employer-linked learning work. The approach is still developing, and conversations with employers are helping to test the assumptions behind it.
The conversation did not settle the question, but it made it sharper. If AI changes some of the early tasks through which people used to build experience, educators and employers may need to think more deliberately about where judgement now forms.
Editorial note: This article reflects CognateUK's educational perspective, informed by employer conversations and programme design work. It is intended to contribute to wider discussion about AI, early-career capability and learning design.
Image: Mikhail Nilov / Pexels. Stock photograph used for illustration.
