Preparing Learners for AI: Why Foundations Should Come First

Preparing learners for an AI future starts with what comes before the AI tools. Dr Rashmi Mantri, founder of British Youth International College (BYITC) and Supermaths, makes the case in a new Mantri framework for AI literacy in FE: FE providers should anchor AI literacy in the core foundations adult and post-16 learners are still missing.

The data backs her up. Skills England’s 2026 Annual Skills Report finds UK employers still rate transferable capabilities such as adaptability, problem solving, and communication as the skills they cannot hire for. The same report says more than a quarter of UK vacancies are hard to fill due to skills shortages, and the Post-16 Education and Skills White Paper published in October 2025 puts 8.5 million UK adults below proficiency in English, maths, or both.

The Hiring Data Says the Foundation Is Missing

UK hiring data shows what employers actually want is the harder-to-measure stuff. Skills shortage vacancies made up over a quarter of all UK vacancies in 2024, up from 22% in 2017, according to the post-16 education and skills white paper.

• 27% of UK vacancies were hard to fill due to skills shortages in 2024
• 8.5 million UK adults have low proficiency in English, maths, or both
• 7.3 million UK adults lack essential digital skills for the workplace
• Nearly 1 million UK 16 to 24-year-olds are not in education, employment, or training

The shortfall sits in priority sectors first. Skills England’s research, summarised in a sector-by-sector breakdown of skills shortage vacancies, found construction with 52% of jobs unfilled, manufacturing with 42%, healthcare with 40%, and the tech sector with 43%. These are the unfilled posts the Skills England study tied to 743 employers across ten priority sectors.

What employers say they want inside those posts is the harder-to-measure stuff. Mantri quotes the consensus: adaptability, problem-solving, initiative, and the ability to learn continuously. The Post-16 White Paper echoes the list, calling out communication, problem solving, digital literacy, and adaptability as the transferable skills employers value. Skills England describes these capabilities as the ones that amplify technical expertise. Mantri proposes that order: foundations first, tools second.

Tools Without Foundations Produce Passive Users

Mantri’s complaint about parts of the FE sector is the focus on immediate employability. Many FE programmes, she writes, teach learners how to use specific tools or platforms before they understand the principles underneath. The result, she argues, is functional fluency without transformational understanding, what she calls passive use of AI. The pattern mirrors what hiring managers report in finance: tech adoption outpaces training.

Mantri frames the risk in two short sentences. AI use will happen regardless, she writes. The problem is using it without understanding it. Her essay lands on a phrase that summarises her whole case: AI literacy needs the same structured treatment as core subjects. UK further education cannot keep teaching Python or specific AI platforms in isolation, she argues. Foundations in logic, data handling, and pattern recognition are what allow learners to question AI outputs and check them against what they know.

The risk is not that learners will use AI. It is that they will use it without understanding it.

The critique comes from Mantri’s FE News essay on preparing learners for AI, where she is identified as the founder of British Youth International College and Supermaths. The critique aligns with what 2024 surveys in financial services found. EY’s December 2024 European Financial Services AI Survey found 78% of finance leaders say their workforce has only some, limited, or no experience with the newest generative AI tools. Only 25% of firms had set up GenAI training programmes.

The Phased Approach Mantri Outlines

Mantri’s prescription is a phased curriculum, not a tool catalogue. She frames it as building computational thinking through structured exposure to coding, layered on top of mathematics and English. The point is to develop the kind of problem-solving capability employers list when they talk about adaptability and reasoning.

The first phase is core. Mantri writes that learners need confidence in mathematics and English before they touch any AI tool. Mathematics develops logic, structure, and analytical thinking; English underpins communication, comprehension, and clarity of thought. The Post-16 White Paper makes the same point in different language, saying 8.5 million UK adults sit below proficiency in English, maths, or both. White Paper authors describe this as the missing layer under the workforce the country needs to build 1.5 million homes, deliver clean power by 2030, and grow NHS capacity.

Coding sits in the middle of the curriculum as the bridge, not the destination. Mantri is explicit that coding is not about turning every learner into a software engineer. The point is to break problems down, test solutions, iterate, and persist when outcomes are uncertain. These habits transfer into any workplace, from advanced manufacturing to healthcare, finance, and education.

AI applications enter last. By the time learners reach this stage, Mantri writes, they understand how AI systems work, what their limitations are, and how to question the outputs. The distinction matters in any post-16 and workforce context where AI already shapes recruitment, productivity tools, and decision-making systems. Without foundations, learners stay passive users of AI.

Mantri pulls the argument back to a single line about jobs. The real issue, she writes, is whether people will be ready for them. Right now, too many are not. Her essay ends with the framing AI should be taught like reading, writing, and arithmetic: a foundational skill, not a vocational one. The advantage, she concludes, will lie with those who can think computationally, adapt quickly, and apply knowledge in new contexts.

1. Core foundations: build confidence in maths and English before any tool is opened.
2. Computational thinking: introduce coding in accessible languages such as Python, progressing into logic, data handling, and pattern recognition.
3. Real-world AI applications: engage learners with AI systems they can interrogate, only after they understand how the systems work and where they fail.

Adult Learners Face the Highest Stakes

For adult learners and reskillers, the order matters most. Mantri points to those moving through further education, reskilling, or re-entering the workforce as the group her argument is built for. Many of them are navigating labour market shifts in real time.

The Post-16 White Paper gives the scale. Beyond the 8.5 million adults below proficiency in English or maths, 7.3 million UK adults lack the essential digital skills required for the workplace. The white paper also counts nearly one million 16 to 24-year-olds not in education, employment, or training. For Mantri, these figures make the foundation-first case economic, not just educational: a learner who leaves FE without maths and English confidence is locked out of the higher technical pathways the Skills England report says the economy needs. Skills England estimates the UK will need 900,000 more skilled workers in its priority sectors to 2030.

What Hiring Managers Are Listing First

The skills employers name first are not the ones the headlines suggest. ManpowerGroup’s 2025 Talent Shortage Survey found 76% of UK employers report difficulty recruiting skilled talent, and 77% globally. The top five in-demand skills in the UK in 2025, according to ManpowerGroup and reported by the top in-demand UK transferable skills in 2025, were IT and data, operations and logistics, engineering, sales and marketing, and front office and customer-facing roles.

Inside those categories, the transferable skills employers want most are the ones Mantri names. The 2025 data ranks the most sought-after transferable skills as reliability and self-discipline, resilience and adaptability, critical thinking and analysis, creativity and originality, and reasoning and problem solving.

The pattern echoes across sectors. EY’s December 2024 European Financial Services AI Survey asked finance leaders to rank the top attributes they will seek when recruiting entry-level talent for an AI-enabled workforce. The number one answer, cited by 77% of respondents, was the ability to adapt and flex. The second, at 70%, was having an innovative and experimental mindset. The third was the ability to collaborate across focus areas at 44%. Tech savvy ranked fourth, at 34%, and dropped out of the top three priorities entirely. Finance firms are now putting human judgment above tech savvy when hiring for AI-era entry-level roles.

Computational Thinking Beats Tool Training

Computational thinking is the load-bearing idea in Mantri’s essay. She defines it as breaking problems down, testing solutions, iterating, and persisting when outcomes are uncertain. The skill sits underneath every technical capability employers ask for. A group of more than 150 mathematicians has now echoed the underlying concern in a separate declaration, warning governments not to take AI math claims on press-release evidence alone.

The mathematics case is closer to home than most FE curricula admit. The mathematicians Leiden Declaration makes the same point as Mantri from a different angle. Foundations in mathematics are how learners separate the patterns AI can actually reason about from the ones it only appears to handle. Mantri’s curriculum places Python and computational thinking at the centre of this work, not as an end in itself but as the language in which logic, data handling, and pattern recognition become concrete.

The sector framing makes the same call. Nilesh Patel, global solutions head for education at Tata Consultancy Services, writes in FE News that FE providers must become agile hubs for continuous professional development. The model he describes centres on foundational pathways, modular learning, and stackable credentials.

Who Pays When Foundations Get Skipped

The cost of skipping foundations lands on the entry-level worker. Bank of America’s campus intern bet in June 2026 was framed as a hedge against AI displacement: the bank said it would welcome nearly 4,000 summer interns and full-time campus recruits while continuing to push automation into coding, support, research, and digital banking. The bank is also managing staffing through natural attrition while $13.5 billion a year goes into technology, more than $4 billion of it on new initiatives including AI.

• Logic and structure, from mathematics
• Communication and clarity, from English
• Problem-solving habit, from computational thinking and coding
• Persistence under uncertainty, from iterating on real problems
• The instinct to question AI outputs and check them against what learners know

The entry-level deal has changed in practice. Coding assistance at Bank of America has lifted technology staff efficiency by more than 20%, and the bank’s AI virtual assistant Erica has logged more than 3.4 billion client interactions. New hires now arrive at desks where AI summaries, coding aids, and policy search are part of the working environment from day one. Stanford Digital Economy Lab research found a 16% relative decline in employment for workers aged 22 to 25 in the most AI-exposed occupations after controlling for firm-level shocks. The learner who arrives without the foundation Mantri describes is the one most exposed to software substitution. The same gap shows up in adult reskillers: the Post-16 White Paper records 7.3 million UK adults lacking essential digital skills, the largest single group in the foundation-gap data.

Frequently Asked Questions

What core skills do UK employers value most in 2026?

ManpowerGroup’s 2025 Talent Shortage Survey found the top in-demand transferable skills in the UK were reliability and self-discipline, resilience and adaptability, critical thinking and analysis, creativity and originality, and reasoning and problem solving. Skills England’s 2026 Annual Skills Report calls out communication, problem solving, digital literacy, and adaptability as the transferable capabilities employers say they cannot hire for.

Why does computational thinking matter for AI literacy?

Mantri’s framework treats coding as the bridge between core foundations and AI applications. Computational thinking breaks problems into parts, tests solutions, iterates, and persists when outcomes are uncertain. Those habits transfer into any workplace, from advanced manufacturing to healthcare and finance, and they are what allow learners to question AI outputs and check them against what they know.

How should FE providers structure AI literacy programmes?

Mantri proposes three phases: core foundations in maths and English, computational thinking through structured exposure to coding in languages such as Python, then engagement with real-world AI applications. The Post-16 Education and Skills White Paper published in October 2025 sets a parallel direction, calling for foundational pathways, agile curriculum development, and stackable credentials across FE.

What is the difference between teaching AI tools and teaching AI foundations?

Tool training teaches learners how to use specific platforms. Foundation training teaches them how the systems work, what their limitations are, and how to question outputs. Mantri argues the gap between the two turns learners into passive users of AI. Critical users come from understanding how the systems work.

Should adult learners prioritise AI tools or core skills?

The Post-16 White Paper counts 8.5 million UK adults below proficiency in English, maths, or both, and 7.3 million lacking essential digital skills. Mantri’s argument for adult reskillers is that building confidence in core skills alongside digital understanding is both educational and economic: a learner who leaves FE without maths and English confidence is locked out of the higher technical pathways the economy says it needs.