As artificial intelligence advances into creative writing, software development, and audio production, parents and educators face growing uncertainty about how to prepare the next generation. Author and researcher Daniel Susskind, who studies the societal impact of AI, argues that traditional attempts to "future-proof" children through specific technical skills are failing, requiring a shift toward fundamental literacy and flexible teaching strategies.
The Failure of 'Future-Proof' Skills
Efforts to predict which specific skills will remain safe from automation have consistently proven unreliable. In 2013, the UK government launched a nationwide initiative to make coding a compulsory subject in primary and secondary schools, aiming to equip young people with essential 21st-century workforce skills.
However, state-of-the-art generative tools have quickly transformed software engineering. By early 2026, AI safety firm Anthropic reported that 90% of the code powering its Claude Code assistant was generated by AI itself. Skills that were intended to protect students for decades became largely automated before many had even completed their schooling.
Returning to Foundational Education
Rather than attempting to forecast future tech trends or banning AI tools outright, Susskind advocates for a "no-regrets" educational framework centered on foundational reading, writing, and mathematics. OECD PISA data shows international declines in literacy and numeracy among young people since 2009—a trend that poses significant risks in an AI-driven society.
Because modern AI systems frequently hallucinate and make basic logical errors—a dynamic computer scientist Geoffrey Hinton characterized as behaving like "idiot savants"—students must possess strong baseline knowledge to evaluate AI output critically. Without solid grounding in basic principles, users cannot distinguish between accurate reasoning and plausible falsehoods.
The 'Teach Both, Test Both' Framework
To integrate technology without compromising core abilities, Susskind points to historical precedents in education policy, specifically the UK’s 1982 Cockcroft Report on introducing handheld calculators to mathematics classrooms.
The report recommended a split approach: allocating curriculum time to learning mathematics both with and without electronic assistance, and assessing students on both modes. Applying this "teach both, test both" methodology to AI would allow pupils to engage with advanced generative software for creative and analytical projects, while ensuring they retain the essential, non-automated competencies needed for an uncertain future.