By Prof. Joseph Sekhampu,
When AI can produce a convincing answer in seconds, what does a university qualification actually certify? Prof. Joseph Sekhampu examines why judgement must become central to a graduate’s value.
Universities produce the research, but AI increasingly controls how the public encounters it. What happens when platforms acquire power over knowledge without equivalent authority?Universities could win the argument over what belongs in the curriculum while losing influence over what becomes visible outside it. AI gives the debate about decolonisation a new urgency.Detecting AI in assignments addresses only part of the problem.
The deeper challenge is establishing whether graduates can recognise a flawed answer when it is already in front of them.When answers become abundant, trust becomes the university’s defining responsibility: establishing where knowledge comes from, why it deserves confidence and who answers when it is wrong.For much of their modern history, universities enjoyed an authority that they rarely had to explain. They determined what counted as legitimate knowledge, who was qualified to produce it, how claims should be tested and who had demonstrated sufficient mastery to be awarded a qualification.
South African universities inherited this tradition while confronting an additional question about whose knowledge had been legitimised and whose had been excluded. Debates about transformation and decolonisation were therefore never simply arguments about curriculum. They were also arguments about epistemic authority.Artificial intelligence introduces a different challenge to that authority. A student can now ask an AI system to explain an economic model, critique an argument, write computer code or summarise research within seconds.
The answer may be excellent, mediocre, subtly wrong or entirely fabricated. Either way, access to explanation no longer depends on entering the university’s knowledge system.The internet had already weakened the university’s control over information. AI goes further by making interactive explanations available at negligible marginal cost, while the cost of determining whether an answer deserves confidence does not fall at the same rate. What changes is not the university’s capacity to produce knowledge, but its position between knowledge and those seeking it.
AI makes that separation possible because it can turn scholarship into answers without requiring users to encounter the scholars, evidence or institutions from which those answers derive.
The challenge facing the university is therefore not simply technological. It is whether an institution can retain epistemic authority when others increasingly control the interface between its knowledge and the public. The growing importance of that interface gives AI systems influence over which sources become visible, which ideas are compressed and whether users encounter the original scholarship at all.
A plausible synthesis can remove the need to return to the research from which it was assembled. AI platforms can therefore acquire epistemic power without acquiring equivalent epistemic authority.The connection to decolonisation goes deeper than representation. Decolonisation has always concerned who has the power to produce knowledge, organise it into recognised categories and determine which forms acquire authority.
AI shifts part of that organising power towards systems whose outputs depend on what has been digitised, indexed and made legible at scale. Digitising archives, strengthening African-language resources and making locally produced scholarship discoverable are no longer peripheral technological tasks. They influence whether epistemic transformation travels beyond the university itself. Universities could win the argument over what belongs in the curriculum while losing influence over what becomes visible outside it.The separation between producing a convincing answer and possessing the judgement behind it becomes especially consequential inside the classroom.
Universities have understandably focused on students using AI to produce assignments. Detection software and redesigned assessments address a real problem, but can preserve the wrong question.
The more important question is what intellectual capability a qualification is supposed to certify when producing a competent answer no longer demonstrates that the student can reason towards it.The university’s response cannot be to recreate the scarcity that AI has removed. Universities will have to become more explicit about what they are certifying.
A graduate’s value cannot rest only on producing competent work when machines can increasingly assist in producing it. It must also rest on the ability to interrogate evidence, recognise uncertainty, defend choices and remain accountable for conclusions reached using those models.Building education around those capabilities is harder than declaring them important. Demonstrating judgement requires academic attention that is already scarce in large classes and heavily burdened institutions.
South African universities cannot respond by assuming unlimited lecturer time. They must identify points within a qualification where direct verification matters most. Universities do not need to establish that every sentence in every assignment originated with a student.
They do need credible moments when students demonstrate that the judgement being certified is genuinely theirs. The goal should be to establish not whether students can work without AI, but whether they can recognise when a model is wrong, incomplete or unworthy of reliance.
The challenge is no longer to prove that every answer was produced without AI. It is to establish that the graduate can think when the answer is already available.The institutional response must extend beyond teaching.
Universities that produce knowledge increasingly encountered through AI have an interest in whether that knowledge remains visible, attributable and open to scrutiny. Protecting the discoverability of scholarship and the connection between claims, evidence and their sources becomes part of protecting academic authority itself.
Universities will retain important advantages in research and awarding qualifications, while AI systems become increasingly capable intermediaries between knowledge and its users. The resulting division of labour changes what academic authority must rest upon.
When answers become abundant, authority depends less on possession of the answer and more on the capacity to establish where it came from, why it deserves confidence and who remains accountable when it is wrong. The university’s future role may therefore be defined less by its control over knowledge than by its ability to preserve the conditions under which knowledge can still command justified trust.
Prof. Joseph Sekhampu is Chief director of the NWU Business School.
