What is something you ask AI now that, two years ago, you would have worked out for yourself? Perhaps it is what to cook with the ingredients in your fridge. Perhaps it is how to word a difficult email, test a theological idea, plan a project, write a piece of code or make sense of a decision you have been circling for days. None of that feels dramatic. Convenience rarely announces itself as dependence. It enters quietly, one useful answer at a time. Then, somewhere along the way, you realise the tool has started framing the question, structuring the answer and shaping what you consider reasonable. The issue is not whether you use AI. It is whether you are still mentally present while you use it. THE CENTRAL QUESTIONAre you using AI to extend your thinking, or are you slowly outsourcing it? In this episode of Off Topic, On Point, Odette de Beer, Dr Janine Do Cabo and Christo Pretorius explore how AI is changing the way we think, work and make decisions. The conversation moves through cognitive offloading, critical thinking, organisational responsibility, originality, dependence, identity and the human abilities worth protecting. A fool with a tool Years ago, Odette attended training for a new process-mapping tool. During a discussion with the trainer, he offered a line she did not appreciate at the time: ‘A fool with a tool is still a fool.’ The phrase sounded dismissive. With distance, its point became clearer. A tool can increase speed, range and output. It cannot supply judgement that the user has never developed. It can make a capable person more capable, and it can help a poorly directed person travel faster in the wrong direction. That story gives us a useful way into the AI conversation. The technology is impressive. The person using it still matters. AI magnifies what is already present Christo compares AI with money: both tend to magnify what is already there. Curiosity becomes more questions. Avoidance becomes faster shortcuts. A habit of testing information can become sharper critical thinking. A habit of seeking easy approval can become a stream of plausible answers that are never properly examined. This is why access alone will not create a durable advantage. Every organisation can buy similar tools. Most people can open the same chatbot. The difference lies in what they bring to it: lived experience, a point of view, ethical boundaries, contextual judgement and a reputation people recognise. Janine argues that businesses will become recognisable through originality and trust, rather than through the sheer volume of content they can produce. If your logo disappeared, would people still know the work was yours? AI makes that question more important because it makes generic competence so easy to reproduce. What is cognitive offloading? Cognitive offloading means using an external aid to reduce the mental work a task requires. We have always done it: diaries hold dates, calculators hold arithmetic and maps hold routes. Generative AI extends that pattern because it can also draft arguments, synthesise information, propose decisions and generate ideas. The important distinction is how the thinking is allocated. Dependent offloading: AI performs the core reasoning, structure or decision, and the user largely accepts the result. Autonomous offloading: AI provides a scaffold—perspectives, questions, critique or options—while the user evaluates the material and retains final judgement. A 2026 three-wave study of 589 university students and early-career knowledge workers distinguished these two modes. Dependent offloading was associated with greater transfer of cognitive agency, lower intrinsic motivation and poorer perceived outcomes in creativity, deep processing and independent judgement. Autonomous offloading was associated with stronger intrinsic motivation and more favourable perceived outcomes. The study relied on participants’ perceptions and associations across time; it does not prove that one style directly caused a change in ability. Even so, the distinction is useful. The number of prompts matters less than the role you assign the tool. Mental presence changes the outcome AI use does not automatically weaken critical thinking. Another 2026 study, this time involving 480 medical students, found that the use of AI-based educational technology was positively associated with critical thinking and self-regulated learning. Stronger self-regulation appeared to buffer some of the negative relationship between cognitive load and critical thinking. This was a correlational study, so it cannot establish cause and effect. Its practical contribution is still valuable: how actively people plan, monitor and evaluate their own learning may shape what happens when AI enters the process. Mental presence looks ordinary. You know why you opened the tool. You notice when an answer sounds confident without being well supported. You ask for counterarguments. You check references. You can explain the final position in your own words, and you are prepared to take responsibility for it. Odette often begins with a prompt that changes the role of the exchange: ‘Be my thinking partner. Challenge these assumptions. Test this idea. Show me where the gaps are.’ Janine calls it sparring. The aim is to create friction that improves the thought, then leave the conversation ready to act. “Promptcrastination is building ideas, systems and procedures through AI with no intention of implementing them.”— Dr Janine Do Cabo Originality becomes more valuable when sameness is cheap AI is exceptionally good at finding and reproducing patterns. That ability can save hours. It can also flood a market with work that is polished, familiar and difficult to distinguish from everything around it. There is a technical warning here too. Researchers have shown that models trained indiscriminately on recursively generated model output can experience ‘model collapse’, gradually losing contact with the underlying human data distribution. This finding concerns machine-learning systems; it is not proof that human beings collapse in the same way. It does, however, underline the value of original human material. The Nature study on model collapse strengthens a practical business lesson: recycled output cannot replace direct observation, lived experience, thoughtful disagreement and contact with the people you serve. The competitive advantage shifts from having the tool to bringing something worth magnifying. Your questions. Your standards. Your experience. Your willingness to say, ‘That answer does not fit this person, this context or this decision.’ What must remain yours? Christo returns to purpose. Before using any tool, understand what you want it to do and why. Intelligence alone is rarely what people need from one another. They also need empathy, attention, relationship and the feeling that somebody is genuinely present. That makes situational judgement a capability worth protecting. Two situations can look identical in a dataset and require different human responses. Context, history, responsibility and consequence still have to be weighed by somebody who is willing to own the decision. The conversation also moves into identity. Romans 12:2 calls people to be transformed by the renewing of the mind rather than absorbed into the patterns around them. Janine connects that renewal to knowing who you are and whose you are. Odette is equally direct: she has worked hard to recover her capacity to think and her sense of identity, and she is unwilling to hand either away. There is no need to reject the efficiency AI offers. Odette is delighted to let it handle technical tasks she can do but would rather not spend her best energy on. The boundary is clearer when you know what the saved time is for: reading, learning, rumination, relationships and work that requires your full presence. The four checks that keep your thinking yours A useful way to stay present is to move through four checks whenever AI is involved in meaningful work. Purpose — What am I asking AI to help me do, and why is this the right use of the tool? Presence — Am I questioning, comparing and evaluating, or waiting for an answer I can accept? Proof — Which claims, references and assumptions must I verify before I rely on this output? Ownership — Can I explain the final judgement in my own words and take responsibility for the consequence? These checks are simple enough for an individual prompt and strong enough to shape an organisation’s AI policy. They keep speed connected to purpose and output connected to accountability. What can you do with this? Name the task before opening AI. Decide whether you need information, options, critique, execution or a decision. Do not hand over more of the task than the purpose requires. Write your first position for high-stakes work. Capture what you currently think, what you are unsure about and what evidence would change your mind. Ask for challenge, not agreement. Invite AI to test assumptions, present the strongest counterargument and expose missing perspectives. Verify what could cause harm or mislead. Open the sources, check the references and confirm important claims with reliable primary material or a qualified person. Reclaim the final judgement. Rewrite the conclusion in your own words. If you cannot explain why it is sound, you are not ready to use it. Quick FAQ’s from this episode: Does AI make people less intelligent? Current research does not support a simple yes-or-no conclusion. Outcomes are associated with the way AI is used. Delegating core reasoning and accepting the output can reduce cognitive agency, while using AI to generate perspectives, critique assumptions and support self-regulated learning can strengthen the thinking process. What is cognitive offloading? Cognitive offloading is the use of an external resource to reduce internal mental work. With generative AI, this can include delega
