Lifelong Learning

Individual Learning Accounts: the entitlement model

The most discussed structural innovation in adult learning is the Individual Learning Account — a government-funded training credit available to every working-age adult, which they can spend on approved education and training throughout their career.

The closest existing model is Singapore's SkillsFuture Credit, introduced in 2016. Every Singaporean aged 25 and over receives a baseline credit (starting at SGD 500, supplemented with top-ups for older workers). The programme is politically popular and has generated significant training activity. The evidence on whether it changes actual skill levels or employment outcomes is more mixed: a large portion of spending has gone to personal interest courses rather than workforce-relevant upskilling, and take-up has been lower among low-income and older workers — exactly the groups most at risk.

France's Compte Personnel de Formation takes a different approach: it accrues based on hours worked and must be used for certified training aligned with labour market needs. It has higher administrative complexity but tighter linkage to employment outcomes.

The design choices that matter most are: who is eligible (all adults, or just those in employment?), how much the credit is worth, what counts as an approved use, and whether there are targeting mechanisms to direct higher subsidies toward more at-risk workers or more economically valuable training. Each choice reflects a different theory of what adult learning is for and who the state's responsibility extends to.

A NZ Individual Learning Account would require significant ongoing fiscal commitment — at the scale of universal access, potentially hundreds of millions annually. The question is not only whether this is affordable, but whether it is better value than other uses of that investment.

Employer obligations: mandating or incentivising?

By OECD standards, New Zealand's employer training obligations are light. Employers are not required to invest in worker upskilling, and there is no formal levy system (like the UK's Apprenticeship Levy) that requires businesses to contribute to workforce development funds.

The arguments for stronger employer obligations include: employers capture the productivity gains from better-skilled workers, so they should share the cost of producing them; mandated investment creates a floor that prevents a race-to-the-bottom where training-investing firms are undercut by those who free-ride; and in sectors where job tenure is long, employers have clear incentives to develop workers they can retain.

The arguments against include: training mandates increase compliance costs that fall disproportionately on small businesses; employers are better positioned than government to know what skills they need; and levies can create bureaucratic friction that reduces rather than increases effective training.

Tax credits for employer-funded training are a middle ground — government subsidises investment decisions that employers make. The evidence on whether tax credits in this space change behaviour or mainly subsidise training that would have happened anyway is inconclusive. New Zealand has limited history with sector-wide training levies; the Industry Training Organisations that existed before the Te Pūkenga merger had some of this character.

Reaching the hardest to reach

Any honest assessment of adult learning policy has to confront a consistent finding: the workers most at risk of displacement are least likely to self-select into training. This is not a character failing. Workers in physically demanding, low-wage jobs have less time, less money, less flexibility, and often less recent experience with formal education — an experience that may not have been positive. An online learning platform with a government subsidy code is not designed for them.

The approaches most likely to reach this population share some common features: training comes to where people already are (workplaces, community venues, marae, union halls) rather than requiring them to come to it; there is human support, not just digital content; trusted intermediaries — unions, community organisations, employers — are involved in outreach and coordination; and the training is clearly connected to outcomes people actually care about.

Union-led learning models, pioneered in the UK and adapted in parts of Scandinavia, use union networks to reach members who would not approach a polytechnic. Community learning hubs, particularly those embedded in marae or Pacific community organisations, have shown reach into populations that formal providers miss. Workplace-embedded training — where upskilling happens during working hours with employer and government cost-sharing — removes the time barrier that stops many workers from engaging with after-hours programmes.

These approaches are generally more expensive per learner than online delivery. Whether that cost is worth it depends on how you weigh the equity imperative against the efficiency case for lower-cost digital delivery.

AI-powered personalised learning: promise and caution

There is genuine excitement about the potential for AI tutoring systems to deliver personalised learning at national scale. An AI tutor can adapt pace and content to an individual learner, provide immediate feedback, identify gaps, and be available at any hour — capabilities that traditional classroom delivery cannot match at reasonable cost.

The evidence base at scale is thin. Most studies of AI tutoring are either small-scale pilots or evaluations of narrow academic subjects in controlled settings. Whether these results generalise to adult vocational learning, across diverse populations with varied literacy levels and digital confidence, is not established.

The quality control challenge is real. Educational AI systems can be confidently wrong, can reinforce existing misconceptions, and can provide a poor learning experience that damages rather than builds confidence. Accreditation and quality assurance frameworks for AI-delivered learning are immature in most countries, including New Zealand.

This does not mean the potential should be dismissed — it means claims should be tested, pilots should be properly evaluated, and procurement decisions should not outrun the evidence. A government that invests heavily in AI learning platforms on the basis of vendor projections rather than robust evidence is not being bold; it is being credulous.