What aspects of graduate jobs are most exposed to AI, and how should organisations respond? Golo Henseke, Associate Professor at UCL, explains.
Generative Artificial Intelligence (GenAI) has entered the workplace at a breathtaking pace. By the middle of 2024, one in four UK workers reported using AI-powered tools in their jobs - a rate of diffusion that rivals, and arguably surpasses, the introduction of personal computers in the 1980s.
Among graduate jobs - positions typically requiring university-level training or equivalent expertise, such as professional, managerial, executive, and technical roles - AI adoption has been even faster: about two in five reported using AI in their job in the first half of 2024.
For graduate employers, this raises an urgent question: what aspects of graduate jobs are most exposed to AI, and how should organisations respond?
A task-based view of graduate work
Jobs are not monoliths. Economists now understand them as bundles of tasks, such as reading, writing, analysing, persuading, or repairing, some of which are susceptible to new technologies while others remain stubbornly human.
Unlike earlier waves of automation that targeted routine manual and cognitive work, GenAI reaches deep into knowledge flows. It is strongest where information must be read, synthesised, and communicated, placing white-collar and graduate jobs squarely in the spotlight.
To measure this for the UK economy, colleagues and I developed the Generative AI Susceptibility Index (GAISI), which maps the 44 general work activities from the Skills and Employment Survey (SES) to the capabilities of large language models.
Where a chatbot such as Claude or ChatGPT can reduce task completion time by at least a quarter, we count that task as ‘exposed’. By weighting task exposure with how important workers say those activities are in their jobs, GAISI provides a job-level measure of susceptibility.
The results are striking: in 2023/24, almost all UK jobs included at least some AI-susceptible tasks.
On average, 40% of activities in a typical job could be accelerated by GenAI, either directly through chatbot interactions or indirectly via the integration of GenAI into local digital infrastructure. However, only around one in eight jobs were highly exposed (with more than half of their tasks affected).
What AI really does at work
Evidence from millions of human–AI conversations confirms which parts of work are already being reshaped. When we mapped Anthropic’s dataset of Claude chatbot interactions to SES tasks, five activities dominated:
- Using specialist knowledge (24% of conversations) – people consult AI to retrieve, explain, or translate domain-specific information and adapt that expertise to their personal or technical problems.
- Analysing complex problems (14%) – breaking down and interpreting complex cases, data, texts, or code to extract patterns, diagnose issues, and generate insights or solutions.
- Thinking of solutions (12%) – brainstorming fixes or creative options and turning them into implementable steps, often in the form of code, configurations, or practical how-to instructions.
- Writing long documents (7%) – composing and refining reports, essays, legal or business documents, manuals and specifications, long-form stories, and translated technical content, ensuring correct structure, style, and terminology.
- Instructing, training, or teaching (6%) – explaining concepts through step-by-step walkthroughs, tailored lessons, and purpose-built learning materials.
Taken together, these five activities form a distinctive bundle of cognitive work. This bundle is essential in around three out of ten graduate jobs (compared with just one in ten non-graduate roles), according to our analysis of the Skills and Employment Survey 2024.
They mirror the cognitive heart of graduate work. GenAI will not eliminate these tasks, but it will shift the time balance: less on first drafts or information retrieval, more on judgment, communication, and spotting problems in AI outputs.
The graduate job profile under AI
Many graduate roles are heavy on exactly the tasks GenAI excels at: writing with accuracy, analysing problems, planning, and communicating professionally.
According to GAISI, about 30% of graduate job tasks can already be accelerated directly by a chatbot, with another third dependent on the integration of GenAI into existing digital tools such as CRM systems, project management software, client databases, or coding environments.
Not every aspect of graduate jobs, however, lends itself to AI augmentation or displacement.
Collaboration, interpersonal tasks, numerical calculations, fault-spotting, and plan implementation remain largely insulated. Motivating staff, handling emotions, or making judgments and decisions are beyond the reach of current AI solutions, while numerical calculations were already automated long before GenAI.
This duality means that most graduate jobs are likely to be reshaped rather than replaced: knowledge-intensive tasks may be pared back, while the human-facing and decision-making elements of graduate work could become more central.
The most and least exposed pockets of graduate jobs
Exposure varies sharply by occupation. Using SES-linked GAISI scores, the most exposed graduate job groups are:
- Science, Research, Engineering and Technology Professionals (93% of jobs heavily AI-exposed).
- Business, Media and Public Service Professionals (61%)
- Business and Public Service Associate Professionals (57%)
These roles combine intensive knowledge work with tasks GenAI supports directly: writing code, conducting research, and producing professional communications.
By contrast, the least susceptible graduate roles are:
- Health professionals (0% of jobs heavily AI-exposed).
- Health and social care associate professionals (0%).
- Teaching and educational professionals (0.5%).
These jobs involve decision-making, personal interactions, and some physical elements where GenAI has little to offer.
The crucial point is that exposure rises with skill requirements – a reversal of the automation story of the 1990s and 2000s. Graduates, especially those in analytical or technical fields, will face the sharpest disruptions.
Implications for graduate employers
What does this mean for recruitment and early-career development? Three points stand out:
- Exposure is not the same as replacement. Most graduate jobs combine exposed and resilient tasks. GenAI is unlikely to eliminate roles, but it will alter how time is spent. Routine drafting and background research may shrink, while oversight, creativity, and personal interactions take on a larger role.
- Early-career workers may be especially vulnerable. Field studies show that GenAI tools disproportionately raise the productivity of less-experienced workers. For employers, this could translate into fewer entry-level recruits delivering the same pre-GPT level of outputs at similar quality.
- Skills portfolios will shift. Technical knowledge remains essential, but its value lies increasingly in knowing how to use AI tools well. Complementary human skills – critical thinking, ethical judgment, collaboration, and leadership – are likely to gain in relative importance.
Preparing graduates for an AI-shaped future
GenAI is not a distant prospect: it is already embedded in graduate work. Just under half of the tasks in a typical graduate job are susceptible to acceleration by GenAI, either directly via ChatGPT or Claude, or indirectly through the integration of these technologies into existing productivity tools. The highest exposure is in IT, R&D, and scientific professions.
Real-world use confirms that AI is being applied to exactly the tasks graduates know how to do well: knowledge engagement, analysis, writing, and teaching.
For graduate employers, the challenge is not whether to adapt, but how. Recruitment and training will need to emphasise skills that complement GenAI, including critical judgment, communication, collaboration, and the ethical use of digital tools.
Subject knowledge remains vital, but it is no longer sufficient on its own. Preparing graduates to work with AI will determine which organisations and young people thrive in the years ahead.