If AI reduces entry-level work, what are universities certifying?

Written by Sondes Gharbi

After attending IATEFL, one idea kept resurfacing across several sessions. AI is expected to reduce entry-level white-collar tasks, many of which have traditionally been carried out by interns or junior employees. According to the World Economic Forum, entry-level roles in the US have already declined by around 35% in some sectors, reflecting early shifts in how organisations structure junior work. (link below)If that is the direction of travel, universities face a more immediate question: what exactly are we certifying when a student graduates, and how do we know it still holds value?If output can be produced without ownership of thinking, then grades risk reflecting performance rather than competence. That creates uncertainty about what a degree actually proves.AI has made it significantly easier to produce strong output: well-organised arguments, clear structure, accurate language. On the surface, the quality is there, but the link between output and understanding is no longer guaranteed. This creates a structural problem for universities. It is not marginal as it affects the reliability of assessment decisions across modules, programmes, and awards.Assessment has traditionally relied on what students produce: essays, reports, presentations. These have been treated as evidence of thinking, not just language or structure. That assumption is now unstable because students can produce without fully owning the decisions behind what they submit.This is not simply about misuse of AI. It exposes a gap that was already there: many academic tasks were designed around completion rather than visible thinking. Students follow the structure, meet the criteria, and submit the work. As long as the output looks right, the process often remains hidden. AI has made this gap impossible to ignore.This is not only a classroom issue since it challenges the validity of assessment as a whole. If output can be generated without ownership of thinking, then the link between grades and actual competence becomes unstable.The response so far has focused on control: restricting tools, using detection software, and asking students to declare AI use. These approaches may slow things down, but they do not address the core issue. They still assume that output can be trusted as evidence.The needed change is not about removing AI from the process, but about redesigning the process itself. In an AI-mediated context, learning must be evidenced through decision-making and justification, not output alone. This is a shift in how learning is evidenced, observed, and evaluated across programmes, not a simple pedagogical adjustment at the level of individual lessons.If students can generate answers, the value lies in the process: how they arrive at them, how they adapt them, and whether they can stand behind them. This requires fewer, higher-stakes tasks where thinking is made visible through decision points, justification, and adaptation.In this kind of environment, AI does not replace thinking, but it makes it visible. It becomes clear whether a student understands what they are doing, or is simply producing something that works on the surface.The question for universities is whether current task and assessment design can still capture learning in a context where output is easy to generate. If it cannot, then strong work is no longer reliable evidence of competence. This has implications far beyond the classroom.If evidence is unreliable, consistency in marking, external moderation, and the credibility of awards are all put under pressure. If universities cannot demonstrate that student work reflects genuine understanding, the signalling value of a degree is weakened because the evidence used to award it is no longer reliably tied to demonstrated competence.This raises a broader question for universities: if thinking must be visible, how is it captured consistently across modules, assessed reliably across markers, and validated at programme level? Without this, improvements at task level remain isolated and the underlying issue of evidence does not change.In an AI-mediated environment, the focus shifts. Not just on what students produce, but on whether they can explain it, adapt it, and take ownership of it. The ability to produce an answer is becoming common. The ability to stand behind it is not.If this change is required, universities will need to redesign how evidence of learning is generated, assessed, and validated at scale so that claims of competence are grounded in observable, defensible thinking.

• How AI is changing the nature of entry level work

When learning stops being supervised: why AI calls for a new kind of trust in ESL

Written by Sondes Gharbi

The myth of measurable learning

For decades, policymakers have treated learning like a checklist; observable, assessable, and reportable. If it can’t be measured, it doesn’t count. That belief shaped how teachers planned, how students behaved, and how institutions justified progress.But as AI enters the classroom, that equation collapses. What once defined “supervised learning” (a teacher marking a task against a standard answer) no longer captures what learning is. AI doesn’t obey our marking schemes. It opens questions we can’t anticipate. It rewards curiosity more than compliance. That is precisely why measurable learning can no longer be our only measure of education.

Supervision without control

Before AI, supervision meant observation: formative feedback, summative marks, and the silent pressure to finish on time. It was order, predictability, and safety. Now, when a learner types a question into ChatGPT or Grammarly, the teacher isn’t supervising anymore, the learner is.Supervision is shifting from an external system of control to an internal system of reflection. The student becomes their own supervisor: questioning, checking, interpreting. AI becomes a mirror of their reasoning, not a substitute for it. This doesn’t mean abandoning structure. It means shifting it. The unit objectives, criteria, and milestones still matter, but they exist to keep learners oriented, not restricted. The real supervision now happens when a learner pauses and asks: How did AI’s answer shape my understanding? That moment (not the exam mark) is the new evidence of progress.

When AI is wrong, learning is right

Mistakes used to be the teacher’s domain. Now AI makes them too. The instinct is to panic; “What if students learn misinformation?” But the deeper question is: what if that error becomes a catalyst for deeper thinking?Students learn best when they wrestle with uncertainty. Discovering that AI got something wrong, and then tracing why, builds exactly the kind of cognitive resilience language education has always aimed for. As I often remind teachers: We don’t learn from smooth seas. We learn to sail through rough waters.When students correct AI, they’re not just learning English, they’re learning discernment. They’re training themselves to think, not just to trust.

When unsupervised becomes unsafe

There are limits. Unsupervised learning becomes counterproductive when learners disconnect completely, when curiosity turns into distraction, or when a student stops engaging meaningfully with ideas. At that point, supervision doesn’t need to return to control; it needs to extend to care. Other professionals, mentors, or systems may need to step in. AI can’t replace that human layer of empathy and presence and it shouldn’t try to.

A small shift teachers can make tomorrow

Moving from supervision to co-supervision doesn’t require an overhaul. It begins with transparency:- Make the assessment criteria visible from day one.
- Define the end task, timeline, and checkpoints clearly.
- Let students know how, when, and where to ask for support.What skill did I gain from this task that I’ll use beyond this classroom?This question alone reframes learning from compliance to connection.

Teaching students to supervise AI

Ethically, this is where the teacher’s role deepens, not diminishes. We’re not teaching them how to use AI, we’re teaching them how to supervise it. Fact-checking, questioning, and self-reflection become literacy skills in their own right. Responsibility replaces obedience.AI mirrors the learner’s thought process: their biases, gaps, and curiosity. When a student reads an AI-generated response, they’re effectively reading a reflection of their own question quality. Helping them see that mirror clearly, without judgment, is where the true supervision happens.

English as a medium, not a subject

In this approach, English stops being the topic of learning and becomes the tool for it. When students create, present, or record their projects in English, the language isn’t studied; it’s lived. It becomes the vehicle for inquiry, creation, and collaboration, the bridge between curiosity and clarity. That’s where long-term retention happens: not through memorization for exams, but through authentic use in meaningful contexts.

From fear to trust to design

AI unsettles us because it exposes how little of learning was ever about thinking. Most of it was about control. But when teachers begin to design learning (not just deliver it) supervision evolves into something more powerful: shared responsibility.Policymakers must now abandon the idea that learning only counts if it’s measurable because the most valuable learning (the kind that turns a student into an autonomous thinker) can’t always be counted. It can only be witnessed through trust, curiosity, and design.

This article was developed through an AI-supported writing process, reflecting the same principles discussed throughout the piece: using AI not as a replacement for thinking, but as a tool for reflection, refinement, and idea development.

Education Cannot Stop at
Knowledge Consumption

Written by Sondes Gharbi

From Knowledge Consumption to Passive Participation

There is a growing contradiction at the heart of modern education. Many learners today can explain global problems in remarkable detail. They can discuss climate change, inequality, misinformation, artificial intelligence, mental health, and economic instability using increasingly sophisticated language and terminology. They can summarize articles, repeat theories, and reference concepts they have studied in classrooms for years.Yet far fewer have been trained to actively respond to those problems. Far fewer have been asked to design solutions, collaborate across disciplines, communicate ideas publicly, defend decisions, test assumptions, or build something meaningful from what they know.In many educational systems, particularly in secondary and tertiary education, learners are still primarily rewarded for consuming, reproducing, and organizing knowledge rather than applying it in visible, practical, and socially meaningful ways.

The Problem Is Not Knowledge- It Is Educational Imbalance

This is not an argument against knowledge itself as foundational and abstract knowledge remain essential, especially in early education. Primary education plays a critical role in helping learners develop literacy, numeracy, conceptual understanding, and the ability to think abstractly. The issue is not that theory exists. The issue is what happens when education stops there.As learners move into adolescence and adulthood, the balance arguably needs to change. Knowledge alone becomes insufficient if students are rarely asked to use it in contexts that resemble real life. A learner may understand the theory behind communication, leadership, sustainability, or entrepreneurship while having little experience actually negotiating with others, presenting ideas, solving ambiguous problems, or working through uncertainty collaboratively.This creates a gap between academic performance and real-world capability where many students become highly informed but professionally passive. A learner may spend years being rewarded for finding the correct answer, following predefined structures, and meeting assessment requirements, yet rarely be asked to initiate ideas, navigate uncertainty, defend decisions, or build solutions independently. Over time, these habits can extend beyond education itself. Learners may become accustomed to consuming, responding, and complying rather than actively contributing, experimenting, or creating. In very subtle ways, education can unintentionally train learners to become observers rather than participants in the world around them.This matters not only economically, but socially. Societies increasingly need people who can collaborate, adapt, communicate across differences, evaluate information critically, and contribute to solving complex problems. These abilities rarely develop through memorizing information alone. They develop when learners are asked to negotiate ideas, test solutions, communicate under pressure, respond to feedback, defend decisions, and work through unpredictable situations with others. This is partly why approaches such as project-based learning, interdisciplinary learning, and experiential education continue to gain attention internationally. These models attempt to move learning beyond isolated content acquisition toward application, communication, and visible thinking.Most importantly, this does not mean abandoning academic rigor. In fact, applying knowledge meaningfully often demands deeper understanding than reproducing it for an exam. Designing a solution, pitching an idea, conducting research for a real audience, creating a campaign, building a prototype, or solving a community-based problem requires learners to transfer knowledge rather than simply recall it.

Why Visible Thinking Matters

The rise of artificial intelligence makes this conversation even more urgent. AI increasingly exposes a structural weakness within many traditional assessment systems: if educational models mainly reward polished output, and AI tools can now generate polished output rapidly, then output alone becomes weaker evidence of learning. This does not mean learning disappears. It means the indicators of learning may need to evolve. In real-world environments, employers and institutions rarely care only about the final answer. They care about how people reached it: how they communicated, adapted, justified decisions, collaborated under pressure, and improved their thinking through feedback and reflection. The focus may increasingly move toward: decision-making, reasoning, collaboration, reflectivejustification, communication, processes, adaptability, and the ability to apply knowledge meaningfully in unpredictable situations.In other words, the challenge is no longer simply whether students can produce answers. It is whether educational systems can make learner thinking, participation, and capability more visible. This visibility matters because real-world environments increasingly depend on how people think, contribute, adapt, communicate, and solve problems with others under uncertainty. When thinking remains invisible, education risks rewarding students for reaching the finish line rather than showing how they reason, collaborate, respond to feedback, and navigate challenges along the way.

Language as a Medium for Participation and Problem Solving

This shift also changes how subjects such as English can be approached.  Rather than existing only as isolated linguistic study, language can become a medium for solving problems, developing ideas, collaborating with others, and participating in real-world communication. Learners can use English to create podcasts, conduct interviews, pitch projects, debate solutions, design campaigns, or present proposals connected to authentic issues and audiences. The language remains important, but it becomes part of a larger process of participation and creation.

A Different Educational Direction

This perspective also informs the work behind initiatives such as EDreframe, which explores how project-based English programmes can make learner thinking, communication, collaboration, and decision-making more visible through real-world tasks and reflective processes.Rather than treating English primarily as isolated language performance, the approach positions language as a medium for participation, problem-solving, and capability development within increasingly AI-mediated educational environments.Ultimately, the question may no longer be whether students can reproduce knowledge. The deeper question is whether education is helping learners become people who can meaningfully apply knowledge, participate in society, adapt to changing professional environments, and contribute to solving the problems around them.

This article was developed through an AI-supported writing process, reflecting the same principles discussed throughout the piece: using AI not as a replacement for thinking, but as a tool for reflection, refinement, and idea development.

How Thinking Evolves When Working with AI

Written by Sondes Gharbi

During a recent conversation with a Senior Manager in People Consulting and Learning at EY, we discussed the kinds of capabilities people increasingly need when working alongside artificial intelligence. Although our conversation focused on the workplace, it prompted me to reflect on a broader educational question: How should thinking evolve when AI becomes part of learning?In this article, I use the term “thinking evolves” to describe the way AI shifts where and when learners engage in critical thinking. Instead of replacing thinking, AI changes the cognitive processes involved in framing problems, evaluating responses, refining ideas, and making informed decisions.For the past few years, discussions about AI in education have largely centred on whether students should use AI. Increasingly, however, this feels like the wrong question. AI is already becoming part of many learners’ everyday lives. The more important question is not whether learners use AI, but how they think while using it.When I refer to thinking evolving, I do not mean that critical thinking becomes less important. Quite the opposite. AI changes where and when critical thinking occurs. Traditionally, much of a learner’s cognitive effort was directed towards producing an answer. Today, AI can often generate a competent first draft within seconds. As a result, learners are increasingly required to invest their thinking before, during, and after interacting with AI. The educational challenge therefore shifts from producing information to framing problems, evaluating responses, refining ideas, and making informed decisions.This shift has implications far beyond education. The World Economic Forum has consistently identified analytical thinking, creative thinking, curiosity, lifelong learning, and AI literacy among the capabilities expected to become increasingly important in the workplace. Yet these should not be viewed solely as employability skills. They are equally valuable as life skills that enable individuals to navigate uncertainty, solve problems, evaluate information critically, and continue learning throughout their lives.One consequence of this shift is that learners must become more intentional when communicating with AI. This is often described as prompting, but effective prompting is not simply about learning clever techniques. It begins with clarifying one’s own thinking. Learners who struggle to explain what they need frequently receive vague or unhelpful responses, not because AI has failed, but because their own understanding of the problem remains underdeveloped. In this sense, writing an effective prompt is an exercise in critical thinking rather than a technical skill.

Context also becomes increasingly important. AI has no knowledge of the learner’s classroom, organisation, previous experiences, or objectives unless these are explicitly provided. Learning to identify relevant contextual information requires analysis, judgement, and an awareness of what factors genuinely influence a problem. Rather than expecting AI to infer missing information, learners must first understand their own context well enough to communicate it clearly.Equally important is recognising that the first response generated by AI should rarely be considered the final one. Effective use of AI is inherently iterative. Learners benefit from questioning initial responses, requesting alternative perspectives, asking for greater depth, challenging assumptions, and refining ideas over multiple interactions. This process mirrors many of the habits associated with critical inquiry and reflective practice. The quality of the final outcome often depends less on the initial prompt than on the learner’s willingness to continue thinking throughout the interaction.Depth is another area where human judgement remains indispensable. AI frequently produces plausible and coherent responses, but these responses are necessarily limited by the information available to the model. They rarely capture the nuances of a particular classroom, organisation, research project, or lived experience. Learners therefore need to move beyond accepting surface-level suggestions and instead ask why an idea might work, under what circumstances it could fail, what evidence supports it, and how it should be adapted to their own situation. In many cases, the most valuable contribution comes not from AI but from the learner’s own observations, experiences, and professional judgement.This is particularly evident in academic writing. AI may suggest directions for further reading or identify influential authors, but meaningful scholarship still depends on engaging critically with original sources, evaluating evidence, synthesising competing perspectives, and constructing well-supported arguments. Similarly, when developing proposals, educational programmes, or business ideas, AI can generate possibilities, but it cannot fully understand the practical realities, institutional cultures, or personal experiences that shape successful implementation. These remain fundamentally human contributions.Perhaps the most significant educational implication is that responsibility does not shift to AI. AI can generate suggestions, but learners remain responsible for verifying information, evaluating evidence, exercising judgement, and making decisions. Critical thinking therefore becomes not less important, but more visible throughout the learning process.If education focuses primarily on teaching students how to use AI tools, we risk preparing them for technologies that will inevitably change. If, instead, we help learners understand how thinking itself evolves when working with AI, we prepare them to work thoughtfully with technologies that have yet to be invented. In my view, this represents one of the most important educational challenges (and opportunities) of the coming years.

Featured CPD Workshops

Helping institutions redesign learning and assessment for AI-mediated education.

Learning Design
• Task-Based Learning & 21st-Century Skills in the AI Era
• Using AI as a Differentiation Tool in Language EducationAI & Communication
• Debate, Critical Thinking, and AI in the Classroom
• AI-Assisted Peer Feedback for Language LearningAssessment & AI
• Assessing Writing When Learners Use AI
• Training AI to Better Assess Writing and Speaking
• Designing Assessment Around Your Curriculum

Available for teaching teams, academic departments and institutional CPD programmes; delivery can be adapted to context.