Adult Essential Skills After the Apprenticeship Rule Change

Adult essential skills provision in England has just lost a large part of its volume, and the official statistics attribute most of the fall to a change in how adults reach the qualification. In the 2025/26 academic year to April, 365,530 adults took part in essential skills courses, down 14.8% from 429,180 at the same point a year earlier, according to the provisional further education and skills statistics published by the Department for Education and the Department for Work and Pensions on 16 July 2026. The same release says most of that fall comes from one place: level 2 English and maths taken as part of an apprenticeship.

For a curriculum lead in adult community learning, a further education college or an independent training provider, that number raises a practical question. If fewer adults now reach English and maths through an apprenticeship, who still needs it, what does the funding still pay for, and where could an AI tutor genuinely help a learner aged 19 or over? This guide answers those questions for the adult (19+) provider rather than the school or the 16 to 19 college. More sector posts like this sit in our industries category.

Scope, stated plainly. This post covers the 2025 apprenticeship change, what the participation figures say, the four legal entitlements, where ESOL sits, who the adult learners are, and how to pilot an AI tutor responsibly. It cedes the 16 to 19 condition of funding and Functional Skills initial assessment to our CENTURY Tech alternative guide, which also explains why a college already using an adaptive initial assessment should keep it. Employer-side off-the-job training hours are a separate question we do not cover here, and the wider market map lives in the AI teacher buyer's guide.

This guide is shipped by the team behind Dilr Academy, an AI tutor that builds interactive, multilingual courses from an institution's own curriculum and teaches each learner one to one. Or see DATS, our five-stage method for placing AI inside an organisation and running it.

What changed for adult apprentices and level 2 English and maths in 2025?

Since February 2025, an employer decides whether an apprentice aged 19 or over at the start of their apprenticeship must achieve level 2 English and maths to complete it. The requirement was not abolished. The apprentice is still assessed on the English and maths their occupation needs, and if the employer declines, the apprentice keeps a statutory entitlement to study English and maths up to level 2.

The change was announced by the Department for Education on 11 February 2025, in a press release on shorter and more flexible apprenticeships. In the department's own words, "Businesses will now be able to decide whether adult learners over the age of 19 when they start their apprenticeship course will need to complete a level 2 English and maths qualification (equivalent to GCSE) in order to pass it." The same release estimated that as many as 10,000 more apprentices a year could complete, cut the minimum apprenticeship duration to 8 months, and said apprentices would still be assessed on core English and maths skills relevant to their occupation.

What the release did not do is remove the adult's right to learn. That right sits in the Adult Skills Fund funding and performance management rules for 2026 to 2027, which tell providers they must not fund an apprentice's English and maths from the Adult Skills Fund (ASF), because the apprenticeship is the primary funding route. The exception is where the employer does not agree to the study. In that case, the rules say, "the apprentice may choose to exercise their statutory entitlement to study English and/or maths up to an approved Level 2", with funding available through the ASF. If you are reworking an adult offer around this change, talk to our team.

That sentence matters more to an adult provider than the headline. It means that, where an employer says no, the decision passes to the apprentice, and an individual choice only turns into enrolment when someone explains it, offers a time and place to learn, and makes the course fit around paid work. Our sector view of where AI pays in UK education sets out which institutions feel this most.

What does the drop in essential skills participation actually show?

The essential skills drop shows a change in route, not a measured change in need. The DfE and DWP release attributes most of the 14.8% fall to level 2 English and maths inside apprenticeships, where English participation fell 51.2% and maths fell 55.8% in a year. The release measures participation, not need, so the fall reflects the apprenticeship ceasing to be a default path into the qualification.

The figures come from the provisional August to April data in the 2025/26 release. Level 2 English taken as a component of an apprenticeship fell from 67,600 learners to 33,020, and level 2 maths fell from 81,150 to 35,840. The release links the reduction directly to the February 2025 change for apprentices aged 19 or over, and it is careful to label the figures as provisional in-year data drawn from the Individualised Learner Record.

Two other numbers in the same release are worth reading alongside it. Three quarters (75.1%) of adult essential skills learners are on regulated qualifications so far in 2025/26, against 87.8% at the same point in 2020/21, which the release describes as a continuing decline. And 84,150 adults in tailored learning are participating with the purpose of improving essential skills, including English, ESOL, maths and digital. The release is careful about the first of these: it says the falling regulated share since 2024/25 is largely accounted for by the inclusion of additional learning aims, all of them non-regulated, so it partly reflects a counting change. What the figures do show clearly is that apprenticeships account for most of the year's fall in formal level 2 volume. That is a workflow question before it is a technology question, which is where our approach to AI always starts.

For a provider, the implication is about outreach, not curriculum. Adult apprentices whose employers no longer require level 2 English and maths have not lost the entitlement to study it, so a provider that wants to reach them has to make the offer directly, which is the kind of operational change our AI placement diagnostic is designed to test before anyone buys a tool.

What does the Adult Skills Fund still fund for English, maths and digital?

The DWP funded Adult Skills Fund supports four legal entitlements to full funding, set in the Apprenticeships, Skills, Children and Learning Act 2009. For adult essential skills, the relevant two are English and maths up to level 2 for learners aged 19 and over without a grade 4, and essential digital skills up to level 1. Both cover learners in non-devolved areas of England, plus some continuing learners.

The four entitlements, as the ASF rules for 2026 to 2027 list them, are: English and maths up to and including level 2 for individuals aged 19 and over who have not achieved a GCSE at grade 4 or above, or who have been assessed as having a skill level below grade 4 even if they hold the qualification; a first full level 2 qualification for those aged 19 to 23; a first full level 3 qualification for those aged 19 to 23; and essential digital skills or digital functional skills qualifications up to level 1 for those aged 19 and over assessed below level 1.

The second half of the English and maths entitlement is easy to miss. An adult who passed a GCSE years ago but now works at a lower level can still be fully funded, provided an assessment shows it. The rules also say prior learning must not be used to reduce funding for English and maths qualifications up to level 2, and that a learner who only wants to resit the exam will not be funded for the resit alone. The assessment itself is a provider judgement, and our CENTURY guide explains why an established initial assessment tool should stay in place for it.

The scope matters as much as the content. The rules state that the ASF supports adult learners in non-devolved areas and does not apply to individuals resident in a devolved authority area, other than continuing learners. A provider in a mayoral combined authority area works to that authority's own adult skills rules, so the details above are a reference point rather than the rulebook. That distinction is exactly the kind of local constraint an AI operating model has to carry, because a tool configured for one funding regime can quietly misdescribe another.

There is also flexibility outside the qualifications. Grant funded providers with a tailored learning allocation are fully funded for non-regulated English, maths and digital. That includes learners assessed at pre-entry level with significant learning difficulties or disabilities, as part of a personalised learning programme, where assessment shows the learner cannot yet undertake the entitlements. Tailored learning is non-formula funded and, in the rules' words, gives providers the flexibility to support learners furthest from the workplace. That is where much of the 84,150 sits, and it is where a provider has the most room to design new kinds of support.

Where does ESOL sit, if it is not a legal entitlement?

ESOL, English for speakers of other languages, is not one of the four legal entitlements in the Adult Skills Fund rules. ESOL qualifications sit in the ASF offer for learning at level 2 and below, which is fully funded for learners who meet the earnings threshold criteria and co-funded for those who do not. Tailored learning can also carry non-qualification ESOL support.

The ASF rules set that threshold at earnings below 26,800 pounds, raised from 25,750 pounds in the June 2026 changes, and it includes people who are unemployed or economically inactive. They also allow a provider to deliver additional learning above the qualification rate where an individual ESOL learner needs it, through the separate funding rates and formula guidance.

The practical effect is that ESOL eligibility depends on a learner's earnings in a way that English and maths eligibility does not. A provider explaining the offer to an employed adult needs to get that right first time, because a wrong answer about cost is hard to correct once a learner has acted on it. This is a good example of where an AI tool should support a member of staff rather than answer the learner directly: the rule is clear, but applying it depends on evidence a person has to check. Deciding which staff hold that judgement is the kind of role design an AI operating model sets out.

The DfE's own statistics treat ESOL as part of the same family. The release describes essential skills provision as supporting adults to improve English and maths up to level 2, ESOL where English is not their first language, and digital skills for those with little experience of devices. Because ESOL sits in a different part of the funding rules from the English and maths entitlement, a provider designing support for one adult may need to plan across more than one strand.

Who are the adults essential skills provision now has to reach?

The adults essential skills provision has to reach are mostly older, and a substantial minority did not learn English first. The DfE's Survey of Adult Skills low skills report finds around 20% of adults in England aged 16 to 65, an estimated 8.5 million people, had low literacy, numeracy or both in 2023, with over half of them aged 45 and above.

The Survey of Adult Skills 2023 low skills thematic report, published by the DfE on 24 March 2026, compares that 8.5 million with an estimated 9 million in 2012. Over half of low performers (56% for literacy, 54% for numeracy) were aged 45 and above. Roughly one third (33% to 35% depending on the domain) of adults whose first language was not English were also low-skilled in 2023, which the report notes was lower than in Italy, the United States, France and Germany.

Three findings from the report shape provision design directly. First, numeracy has improved since 2012 while literacy has shown no marked improvement. Second, rising qualification levels have not protected adults against low skills, which is exactly the group the "assessed below grade 4" clause of the entitlement exists for. Third, the report names targeting interventions by age and prioritising digital skills for low-skilled adults in work as areas of focus. For the wider question of how to judge an AI tutor at all, see the AI teacher buyer's guide.

Put those together and the learner profile a provider should design for looks like this: an adult, often over 45, who may hold a qualification on paper and may speak another language at home. The survey covers ages 16 to 65, so a 19+ provider should read it as a population picture rather than a profile of its own cohort. The tutoring evidence discussed in our Third Space Learning guide comes from school settings, so a provider should be careful about borrowing it for adults, whose timetables and starting points differ.

Where can an AI tutor support essential skills learners, and where can it not?

An AI tutor can support essential skills learners between taught sessions: practising maths and digital skills at a pace that suits shift work, explaining a concept again in the learner's first language, and checking understanding before moving on. It cannot replace the regulated assessment, the funding eligibility check, or the tutor who builds the confidence that keeps an adult enrolled.

In our view, the clearest case is numeracy for adults who speak another language at home. Dilr Academy, for example, rebuilds a whole course in the learner's language, including right-to-left scripts, so an adult working towards level 2 maths can have a fraction or a percentage explained in the language they think in. That is different from ESOL itself. An ESOL learner is learning English, so explaining everything in their first language would work against the goal, and a provider should use first-language support there only as a scaffold that its ESOL tutors choose.

The second case is structure. Dilr Academy builds a course from an institution's own curriculum as modules, topics and stated outcomes, then teaches it one to one. It asks a question before it explains, and its knowledge tracing estimates what the learner has mastered before setting the next step. For an adult who left school with a bad experience of maths, a tutor that asks rather than lectures can feel very different, though no provider should assume that without evidence. Our explainer on how a Socratic AI tutor works sets out the method and its limits.

The limits are just as clear:

• Qualifications. Functional Skills, GCSEs and ESOL Skills for Life are awarded by regulated awarding organisations. An AI tutor does not award them, and Dilr Academy's live page names no accreditation.
• Eligibility. Whether a learner is fully funded, co-funded or outside the ASF entirely is a judgement on evidence, made by the provider.
• Pre-entry and SEND learners. Learners with significant learning difficulties or disabilities on a personalised programme need a teacher who knows them.
• Confidence and retention. Adults leave courses for reasons no software sees: a shift change, a caring duty, a bad week.

There is also a supplier-side test worth borrowing. The DfE's generative AI product safety standards are mainly intended for edtech developers and suppliers to schools and colleges, so they are not addressed to an adult community learning provider. They still make a sensible checklist: they say suppliers should not exaggerate the impact or capabilities of their tools, and that products should report to teachers on usage and on how often learners ask the tool to do the thinking for them. Our AI tutor safeguarding checklist draws on the same DfE standards for schools and trusts.

The same evidence-first thinking underpins our AI execution office, the embedded team that runs an AI deployment after the pilot, so that the tool, the staff and the evidence stay joined up.

How should a provider pilot an AI tutor for adult essential skills?

A provider should pilot an AI tutor for adult essential skills on one cohort, one subject and one outcome it already measures, such as a level 1 maths group or an essential digital skills class. Check eligibility and funding first, build the course from the provider's own scheme of work, and compare progress against the assessment the provider already runs.

Confirm the route. For each learner, decide whether they sit under the English and maths entitlement, the digital entitlement, the level 2 and below offer for ESOL, or tailored learning. If the provider is in a devolved area, use that authority's rules. This step needs a person, not a tool, and it is the first thing a placement diagnostic maps before anything is bought.

Pick one cohort. A small numeracy or digital group is a better first test than ESOL, because the role of first-language support is clearer. Working adults on evening or weekend timetables are a natural first group, because practice between taught sessions is where a one-to-one tutor can add most.

Build from your scheme of work. The course should follow the provider's own scheme of work and outcomes, not a generic library. With Dilr Academy, the institution brings the curriculum and the languages its learners need, and access runs through the licensing provider rather than individual sign-ups. Our guide to adaptive learning AI and the UK curriculum explains how knowledge tracing maps onto a syllabus.

Measure what you already assess. Use the initial and diagnostic assessments and the progress reviews the provider already records. Compare the pilot cohort with a similar group, and report honestly if the difference is small. The DfE standards' warning about exaggerated claims applies to a provider's own governors as much as to a supplier.

If the provider wants an outside view before committing, the DATS method starts with exactly this kind of placement question: where in the learner journey does AI pay, and where does it add risk. Our approach to placing AI inside institutions explains why we begin with the workflow rather than the product.

What is the best AI tutor for adult essential skills learners in 2026?

There is no single best AI tutor for adult essential skills learners in 2026. The best choice depends on whether the provider needs initial assessment and a ready-made Functional Skills library, school-age practice, or courses built from its own scheme of work in several languages. Dilr Academy fits the last case, and its live page describes no standalone initial assessment.

Judged on those criteria, the options look different for different providers. A college that needs an adaptive initial assessment for Functional Skills is better served by an established platform such as CENTURY, which describes itself as an initial assessment, teaching and learning platform for FE colleges, as our CENTURY comparison explains. A provider whose learners are mostly school-age will find school-focused tools such as Sparx and Third Space Learning closer to its needs. Khanmigo tutors over Khan Academy's own library rather than a provider's curriculum; our Khanmigo alternative guide covers the detail.

Dilr Academy is built for the case where a provider owns the curriculum and needs it taught one to one, in the languages its learners speak, by a tutor that asks before it tells. Its live page shows 12 courses built so far, presented as examples rather than a catalogue, and it is licensed to the institution rather than sold to individuals. Our full comparison of AI education tools, linked below, sets out the criteria side by side, and our UK AI teacher platform landscape maps the wider market.

Whatever the forum, the question behind a search for the best AI tutor comes down to this: will it teach, or will it just give answers. For adults returning to English and maths, that is the right question, and it is the one a pilot should answer with evidence.

Frequently asked questions

Does Dilr Academy count towards Functional Skills or ESOL qualifications?

No. Dilr Academy is an AI tutor, not an awarding organisation, and its live page names no accreditation for Functional Skills, GCSE or ESOL qualifications. A provider can use it to teach and practise towards a qualification that a regulated awarding organisation assesses, and the funding claim still follows the provider's own enrolment on an approved qualification or on tailored learning.

That keeps the roles clear. The provider decides the learner's route, enrols them, and owns the assessment and achievement record. Dilr Academy teaches the course the provider has defined, tracks mastery inside it, and keeps each institution's learners and progress separate from every other institution's. Our overview of what Dilr Academy is explains the course model in more depth, and our guide to AI governance for education institutions covers the oversight a provider's board should expect.

Can an apprentice still study level 2 English and maths if the employer says no?

Yes. Under the Adult Skills Fund rules for 2026 to 2027, an apprentice aged 19 or over at the start whose employer does not agree to English and maths study may choose to use their statutory entitlement to study English and maths up to an approved level 2, funded through the Adult Skills Fund in non-devolved areas of England.

A provider can make that route visible. An apprentice cannot use an entitlement they have not heard of, so a short, accurate explanation from the provider, followed by a course that fits around work, gives an entitlement a real chance of becoming an enrolment. Talk to us through our contact page if you want to test that offer with a real cohort, or read more about DILR.AI and the AI solutions we build for institutions.

Want to see this on your own scheme of work? Explore Dilr Academy for training providers, read our AI tutor buyer's guide, see our DATS methodology, or read about our approach to placing AI inside institutions.

Written by the Dilr.ai engineering team, practitioners who ship enterprise AI in production. Follow us on LinkedIn for shipping notes, or subscribe via the RSS feed.