AI Is Reshaping Graduate Hiring: How Assessment Centres Must Adapt in South Africa
Graduate hiring is entering a new phase.
Artificial intelligence is not simply adding another tool to recruitment.
It is changing how candidates present themselves, how employers screen applications, how skills are interpreted, and how confidence can be mistaken for competence.
For organisations running graduate programmes, this raises an immediate question: if artificial intelligence can help candidates write stronger applications, prepare polished answers, and optimise their online presence, how should employers redesign assessment centres so they still identify genuine potential?
The answer is not to abandon assessment centres. It is to make them more rigorous, more behavioural, more work-relevant and more defensible.
That matters globally, and it matters even more in South Africa, where graduate hiring carries both commercial and social significance.
The Department of Higher Education and Training reports that unemployment among people with tertiary qualifications remains materially lower than the national average, while Microsoft’s South African skills initiatives underline how quickly artificial intelligence capability is becoming part of the employability agenda.
AI Is Reshaping Graduate Hiring
The global signals are clear. The World Economic Forum reports that employers expect 39% of workers’ core skills to change by 2030, while LinkedIn says that by 2030, 70% of the skills used in most jobs will change, with artificial intelligence acting as a major catalyst. In the same World Economic Forum research, two-thirds of employers say they plan to hire talent with specific artificial intelligence skills. This means graduate hiring is no longer just about academic pedigree and baseline potential. It is increasingly about adaptability, judgement, technological literacy and the ability to work effectively with intelligent tools.
At the same time, the market for graduate talent is becoming tighter and more competitive rather than disappearing. NACE found that nearly 90% of employers expected to increase or maintain hiring levels for the Class of 2025, even though overall projected growth moderated sharply to 0.6%. In the United Kingdom, the Institute of Student Employers reported that graduate recruitment fell by 8% and forecast a further 7% decline overall, with employers still targeting occupations such as information technology, artificial intelligence and digital. Read together, these signals suggest a more selective market, not a collapsed one. Employers are still hiring graduates, but they are becoming more exacting about what “job-ready” actually means.
That shift is already changing the logic of screening. NACE’s 2026 research shows that 70% of employers now use skills-based hiring, up from 65% the year before, and that only 42% still screen candidates by grade point average, down from 73% in 2019. Its 2025 data also shows the attributes employers most want to see on graduate resumes: problem-solving, teamwork, written communication, initiative, strong work ethic and technical skills. That is highly significant for assessment centre design. If the market is moving away from academic proxies and toward observable capability, then assessment centres should become the primary environment in which those capabilities are tested in action rather than inferred from a polished application.
This is where artificial intelligence creates both the problem and the opportunity. On the problem side, candidates can now use generative tools to improve CVs, draft application answers, research employers, prepare likely interview responses and simulate psychometric-style practice. The Institute of Student Employers notes directly that candidates in virtual assessment contexts may use external support such as artificial intelligence to answer questions. On the opportunity side, this same reality can force employers to improve the quality of their assessment design. If a candidate can succeed largely through rehearsed answers or generic presentation polish, the exercise is not measuring enough. An assessment centre that remains robust in the artificial intelligence era will be one that tests live reasoning, ethical judgement, collaboration, listening, prioritisation, learning agility and the ability to apply knowledge under changing conditions.
The first adaptation is to move from recall to application. Traditional graduate assessments often over-reward preparation quality: rehearsed competency examples, predictable interview responses, and presentation decks that look impressive because a candidate has received outside help. In an artificial intelligence-shaped market, employers need to focus more heavily on what a candidate can do in the room. Case discussions should involve incomplete information. Group exercises should require trade-offs rather than consensus theatre. Role plays should test how candidates respond when assumptions change halfway through the exercise. Written tasks should ask candidates to diagnose, decide and justify, not merely summarise. This kind of design makes it much harder for artificial intelligence-generated preparation to carry a weak candidate through the process. It also makes the assessment much closer to real work. The CIPD notes that employers should use a range of methods and support interview insight with other evidence such as psychometric testing or task simulation activities.
The second adaptation is to assess how candidates use artificial intelligence, not just whether they use it. In many graduate roles, the question is no longer whether artificial intelligence will be present in the workflow. It is whether the candidate can use it responsibly, critically and productively. Assessment centres should therefore incorporate exercises where candidates must evaluate artificial intelligence-generated output, spot weak reasoning, identify hallucinations, recognise privacy or bias risks, and decide when human judgement should override machine assistance. A graduate who can prompt a tool is useful; a graduate who can interrogate a tool is far more valuable. As employers increase their demand for artificial intelligence-related capability, assessment centres should test for artificial intelligence literacy as applied judgement rather than as technical jargon.
The third adaptation is to tighten structure and scoring. The artificial intelligence era puts a premium on defensibility. If employers are using technology anywhere in the selection chain, they need stronger governance, not weaker oversight. The Equal Employment Opportunity Commission states that while artificial intelligence may offer benefits, it also has the potential to violate anti-discrimination laws when used in employment decisions. The United Kingdom Information Commissioner’s Office similarly warns that artificial intelligence can bring benefits to recruitment but also introduces risks to privacy and fairness. That means assessment centres should rely on predefined competencies, anchored scoring guides, multiple assessors, evidence capture, calibration discussions and auditable decision records. The point is not only better selection. It is also legal, ethical and reputational protection.
The fourth adaptation is to improve fairness and inclusion by design. Technology can widen access, but it can also quietly introduce exclusion. The CIPD notes that technology use in recruitment and onboarding has increased sharply, with 78% of organisations increasing technology use in the last 12 months and 31% using some form of artificial intelligence or machine learning, up from 16% in 2022. Yet it also stresses that methods must be tested for fairness and candidate experience, and that reasonable adjustments should be proactively offered. The Institute of Student Employers adds that virtual assessment centres can reduce geographical bias and improve accessibility, but also highlights risks linked to digital poverty, inexperience with platforms and disadvantage to certain groups if adjustments are not made. In South Africa, where inequality in connectivity, devices and quiet workspace remains real, this is not a marginal design concern. It is central to whether an assessment centre identifies talent or merely filters for privilege.
The fifth adaptation is to redesign what “graduate potential” means. Too many graduate assessment centres still reward polish, confidence and familiarity with corporate language. Yet NACE’s data on the skills employers want most points in a different direction: problem-solving, teamwork, written communication, initiative and adaptability. The World Economic Forum likewise highlights analytical thinking, resilience, flexibility, leadership and social influence among the most important core skills. These are not traits that should be inferred from a strong application form. They should be observed in behaviour. Strong assessment centres therefore need fewer superficial brand questions and more carefully constructed moments that reveal how a candidate thinks, collaborates, challenges, recovers and learns.
The sixth adaptation is to treat the assessment centre as part of employer branding, not just candidate filtering. Early-careers talent is increasingly evaluating employers for growth, development and future relevance. If the process feels outdated, opaque or punitive, high-potential candidates will notice. If it feels rigorous, fair, work-relevant and intelligently designed, the organisation signals seriousness. The CIPD stresses that candidates should be told what to expect and that processes should not be unnecessarily long. That matters more now because the first rung of the career ladder is under pressure. Recent CIPD commentary warns of a potential talent crisis if organisations allow entry-level pathways to erode while artificial intelligence reshapes junior work. Assessment centres should therefore communicate not only standards, but also opportunity: this is how we work, this is how we learn, and this is how we expect future leaders to think.
For South African organisations, the stakes are even higher. Graduate hiring is not just a staffing process. It is a capability pipeline. It feeds leadership benches, transformation objectives, succession depth and long-term competitiveness. The local labour market already puts pressure on employers to convert education into employability more effectively, while the acceleration of artificial intelligence skills investment in the country makes digital and judgement capability more important by the month. If graduate programmes continue to rely on outdated screening logic, organisations will either miss strong candidates or hire for surface polish rather than real promise. Neither outcome is affordable.
So what should a modern assessment centre include? At minimum, a blend of structured interviews, live case analysis, group problem-solving, written reasoning tasks, behavioural observation, and a measured assessment of how candidates engage with technology and artificial intelligence-related ambiguity. Scoring should be based on clear criteria and backed by trained assessors. Virtual delivery should be used where it improves scale and access, but not as an excuse for weaker evidence. And every exercise should answer one hard question: does this reveal the candidate’s likely performance in real work, or merely their ability to prepare well for recruitment theatre? The organisations that can answer that question honestly will make better hires.
Artificial intelligence is reshaping graduate hiring, but it does not make assessment centres obsolete. It makes them more necessary. In a world of optimised applications and machine-assisted preparation, the assessment centre becomes the place where organisations can still observe judgement, integrity, collaboration, learning agility and human effectiveness under pressure. The future belongs to employers that redesign their assessment processes around evidence rather than impression. For Duja Consulting clients, that creates a clear agenda: re-engineer graduate assessment now, before outdated methods become a hidden strategic risk.
If your organisation is reviewing its graduate programme, learnership pipeline or assessment centre methodology, Duja Consulting can help you redesign the process so it is fairer, more work-relevant, and better suited to an artificial intelligence-shaped talent market.
Source List:
- Duja Consulting SEO Specification, March 2026
- World Economic Forum, The Future of Jobs Report 2025
- LinkedIn Economic Graph, Work Change Report: AI Is Coming to Work, January 2025
- National Association of Colleges and Employers, Job Outlook 2025
- National Association of Colleges and Employers, Employers Expect Hiring to Level Off for the College Class of 2025, April 2025
- National Association of Colleges and Employers, Employer Use of Skills-Based Hiring Practices Grows, January 2026
- Institute of Student Employers, Student Recruitment Survey 2025
- Institute of Student Employers, How to Get the Most Out of Virtual Assessment Centres, December 2024
- Chartered Institute of Personnel and Development, Selection Methods, November 2024
- Chartered Institute of Personnel and Development, Using Assessment Centres Effectively for Candidate Selection and Development, September 2025
- U.S. Equal Employment Opportunity Commission, What Is the EEOC’s Role in AI?, April 2024
- Information Commissioner’s Office, AI Tools in Recruitment, November 2024
- Information Commissioner’s Office, Thinking of Using AI to Assist Recruitment? Our Key Data Protection Considerations, November 2024
- Department of Higher Education and Training, Characteristics of the South African Labour Force, 2025
- South African Government, President Cyril Ramaphosa: Microsoft Investment and AI Skilling Initiative Announcement, March 2025