How to change assessment in the age of AI

By Clare Russell, Principal and CEO at Runshaw College in Leyland, Lancashire

I still remember my secondary school maths teacher’s hesitation when she said: “You may now take out your calculators.” In the late 1980s, calculators felt like cheating, yet they liberated us to explore real mathematics. My grandparents had slide rules, my cousins were the first to use calculators in exams, I went on to study maths at university. The calculator didn’t ruin me – it changed what mattered. We moved from arithmetic to proof, reasoning and modelling.

That history matters now, because education is facing its next “calculator moment” with artificial intelligence (AI). The same anxieties echo through staffrooms and policy debates: will students stop thinking, writing, researching for themselves? After 25 years in further education, I recognise the pattern. We’ve been here before, and we already know what works: systemic change, led by those who control assessment.

When calculators appeared in classrooms, the sector split between traditionalists and reformers. Yet the outcome was clear. Once exam boards allowed calculators, assessment changed. Tests stopped rewarding forty-step arithmetic and started valuing reasoning, interpretation and application. The bar didn’t lower – it moved to more meaningful ground. Crucially, that shift was coordinated at system level, not left to individual teachers.

AI now challenges literacy and research as calculators once challenged numeracy. It produces competent drafts, summaries and translations. Students are already using it, often covertly, while institutions waste energy policing it. The result is a culture of suspicion rather than capability. Meanwhile, every workplace our students will enter expects AI literacy. We cannot prepare learners for an AI-augmented world while pretending that world doesn’t exist.

Individual teachers can innovate – redesigning lessons, focusing on critical use of AI – but we cannot alter the exams our students must sit. Assessment dictates curriculum. Until awarding organisations and regulators adapt, teachers are trapped preparing students for qualifications that measure skills the future no longer values.

The calculator precedent shows the way forward: don’t ban the tool – transform the test. Assessment must evolve to measure what only humans can do: judgment, creativity, insight and critical evaluation.

• In mathematics, future assessment should focus on problem formulation, proof, verification and creative problem-posing. AI can compute flawlessly; humans must decide which problems to model, why a method works and when an answer makes sense.
• In essay-based subjects, AI’s ability to generate text should push us to assess thinking rather than writing. Students could compare interpretations, critique evidence and construct arguments orally or through guided research. Verbal presentations, debates and live questioning would reveal understanding far better than unseen essays.
• In the sciences, assessment could emphasise experimental design, data interpretation and critical evaluation of conflicting studies – skills that define scientific literacy.
• In vocational and technical areas, project-based and practical assessments already align well with the AI era, but these too should make the process visible: how students use AI tools, make decisions and demonstrate professional judgment.
• In modern languages, the focus should move from translation accuracy to communication, cultural understanding and creative expression – areas where AI still falters.

These shifts would make education more authentic and rigorous, but they cannot happen through isolated classroom practice. Only awarding bodies, Ofqual and the Department for Education can redesign qualification frameworks at scale. Teachers can adapt once the system moves together.

Predictably, objections to this shift mirror those raised in the 1970s. “Students need foundations.” Yes, but foundations must serve understanding, not speed or memorisation. Calculators didn’t end numeracy; they allowed deeper mathematical thinking. AI won’t end literacy; it will end certain mechanical forms of writing.

Another concern is inequality of access, but banning AI entrenches inequality, as advantaged students will continue using it privately. Fairness comes from equitable access and explicit teaching of AI literacy, just as standardised calculators once ensured a level playing field.

Embracing AI doesn't lower standards. Calculators made some tasks easier but enabled more advanced mathematics. AI will do the same for analysis, synthesis and creative reasoning. It will demand more, not less, from learners.

Designing new assessments will be complex and costly, but so was the calculator transition. We already spend scarce resources on ineffective AI detection software and malpractice investigations. Redirecting those resources towards thoughtful reform would strengthen, not weaken, educational integrity.

So what should happen now? Policymakers should commission an urgent national review of how AI will be integrated into qualifications. Awarding organisations should begin pilot programmes, experimenting with AI-integrated models, new marking criteria and standardised AI tools within exams. Teachers must be involved, but they need regulatory clarity and permission to innovate.

The UK’s global competitiveness depends on producing AI-literate graduates. Our current qualifications, designed for a pre-AI world, risk becoming irrelevant. We have a brief window to align assessment with the realities of modern work and learning.

After a quarter century in FE, I know how much teachers and students can achieve – but I also know the limits of local innovation. Systemic reform once transformed maths education; it can now transform learning across all disciplines. The calculator moment proved that embracing technology can raise standards and enrich understanding when assessment evolves alongside it.

We face the same choice again, only faster. AI is already reshaping learning and employment. The question is whether our qualification system will evolve to match, or cling to an outdated model.

The next generation of students could be the most capable yet – if we design assessments that measure the skills an AI world truly requires. They need awarding organisations and regulators with the courage to change course, just as their predecessors did half a century ago.

We figured it out then. We can do it again, if we start now.