An AI company promises to slash the cost of producing apprenticeship teaching videos by 95 per cent. Another offers a virtual Ofsted inspector that can observe lessons and produce report cards. Others have developed systems to predict which learners might withdraw before completing their courses – or trigger funding clawbacks.
Welcome to the expanding market for artificial intelligence in apprenticeship training, where almost every aspect of delivering a qualification is becoming an opportunity for automation.
What apprentices may lose in the great AI efficiency race
For financially stretched training providers, the attraction is obvious. Technology can take over time-consuming administration, help staff to manage more learners and potentially transform the economics of delivery.
For the companies selling these products, it represents a lucrative new market.
Of the 15 largest privately owned training companies by apprenticeship levy funding in 2024-25, FE Week identified evidence that all but two were using AI in at least one aspect of teaching, assessment, marking, feedback or predicting learner performance.
But the benefits are not necessarily being shared equally.
While some providers say savings are being reinvested into more individual support for apprentices, automation is also enabling larger caseloads and higher profit margins. So, who ultimately benefits from the efficiency revolution?

Higher revenues
The commercial advantages of adopting AI to improve productivity can be particularly attractive for companies.
As AI platform Sainaptic says in its website’s sales pitch to training providers: “By relieving assessors of repetitive marking, you scale capacity without ballooning overheads, leading to higher revenues.”
There is little clear evidence yet of widespread redundancies among AI-adopting training providers caused by its implementation, although at several of them, turnover has increased considerably faster than employee numbers.
In 2025, Cambridge Spark, which developed its own AI feedback and assessment platform, increased turnover by nearly a third and operating profit by two-thirds, paying out £2.5 million in dividends, while expanding its workforce by just 15 per cent.
At Kaplan, average employee numbers fell by 3.4 per cent to 979 in 2025, despite turnover rising 4.4 per cent to £100.7 million. Operating profit also increased by 19.8 per cent.
Kaplan’s director of education Lisa Nelson describes its TutorBot as a “study partner that provides the support of a teaching assistant around the clock”.
She said coaching and tutor caseloads had remained static since 2023 and the company’s “priority remains delivering high-quality, human-first apprenticeship training”.
At Multiverse, overall headcount fell from 822 to 813 in the year to 2024-25, while revenue rose by more than a third to £80 million.
One prospective Multiverse coach told FE Week they had been informed they would be expected to support approximately 100 learners, compared with around 40 in a previous coaching role.
Kaplan’s Lisa Nelson
15 minutes of one-to-one coaching
Each apprentice would receive just 15 minutes of one-to-one coaching monthly, according to the prospective employee.
Multiverse said its coaches’ sentiment on workload and tooling had improved over the previous six months, and pointed out that as well as the monthly group coaching, learners get monthly workshop sessions and can request as many 1:1 sessions as they like.
The account illustrates the potential tension between improving productivity and preserving meaningful human contact, although it does not establish that AI was responsible for the caseload expectations described.
Nick Heard-Jones, who is undertaking doctoral research into private equity involvement in the training sector, believes the primary objective of providers should be to enhance the learner and employer experience.
“Shareholders need an appropriate return on their investment, but we also need to see a step up in quality,” he said.
He is particularly concerned about the implications for the relationship between apprentices and the people supporting them.
“One of the key features of an apprenticeship is an inexperienced apprentice training alongside a subject expert, receiving individual and personalised feedback within the occupational context.
“If coach/assessor caseloads increase as a consequence of AI utilisation, the risk is this will be lost and the apprentices’ experience will diminish.”
While the financial benefits of AI are relatively easy to quantify, changes in the quality of individual learner support are much harder to measure.
Ofsted judgments, Heard-Jones points out, do not reveal what proportion of apprenticeship funding is spent directly on teaching and supporting learners.
Jamie Hibbert of Cambridge Spark
Where do the hours go?
For teaching staff, the benefits of using AI can be considerable. A Jisc pilot of TeacherMatic found FE teachers saved “just over two hours” each week.
One independent training provider representative said tutors who had previously been working 60-hour weeks were “happy” that AI had enabled them to cut their hours to the 40 they were supposed to be working in the first place.
Jamie Hibbert, a digital skills coach at Cambridge Spark, developed a Copilot agent that analyses transcripts of meetings with apprentices and their line managers, then drafts progress reviews against the relevant knowledge, skills and behaviours.
The time he spends writing up each review has fallen from as much as 45 minutes to 20, so he can now concentrate fully on the conversation instead of dividing his attention between coaching and note-taking.
“I don’t see it as doing my job for me. I see it as helping me get better at what actually matters: my ability to support and coach apprentices.”
At Hawk Training, operations director Vanessa Jones said in an Aptem case study that the company would be “significantly struggling in terms of capacity” without AI.
“The timesaving is where we can invest in human-to-human education,” she added.
But the question is whether time saved through automation consistently translates into better learner support and staff satisfaction.
At Multiverse, instructors told The Guardian that AI systems monitoring their teaching had caused them “horrendous stress”.
The technology assigns risk statuses and confidence scores, with staff complaining that legitimate teaching decisions could be flagged as undesirable digressions.
Multiverse said human managers retained responsibility for performance reviews and the technology helped to identify areas requiring support.
Multiverse founder Euan Blair
The money behind the machines
The Department for Education’s assessment of England’s edtech market identified assessment technology as its fastest-growing segment, generating £470 million in turnover.
The research, which focused primarily on schools, also identified weaknesses in the evidence for the effectiveness of edtech and how institutions evaluate it.
For apprenticeship technology companies, the financial results illustrate the commercial momentum.
Between 2024 and 2025, turnover rose from £4.7 million to £6.3 million at Bud Systems, £9.9 million to £10.8 million at OneFile (part of Canadian software group Harris Computer), and £7 million to £9.1 million at Aptem.
These businesses provide platforms that have become central to how many apprenticeship providers manage learners, assessment, progress and compliance, and have developed AI-enabled functions.
Once a provider’s learner records, assessment workflows and staff routines are embedded within a platform, changing supplier can become difficult and expensive.
The report warned about vendor lock-in, while one training provider source fears that AI tools are being offered cheaply to encourage adoption before “all of a sudden a significant cost will start to come in”.
The concern is that providers could exchange one substantial expense – human labour – for a growing dependence on technology companies whose long-term charges they cannot easily control.
The companies are also developing close relationships with organisations representing the sector.
Bud, OneFile and Aptem are all patrons of the Association of Employment and Learning Providers (AELP), with patronage costing £10,000 plus VAT annually.
The association represents training providers while maintaining commercial relationships with businesses selling them software and automation.
There is no evidence that those arrangements have improperly influenced its policies.
But they illustrate how closely connected the suppliers, customers and representative organisations have become as technology takes on a larger role in apprenticeship delivery.
AELP chief executive Ben Rowland said: “Many of our patrons are not suppliers but regular AELP members, who are going above and beyond with their contribution. We know all our patrons are proud to support AELP and the sector, and we are proud to have them all as patrons.”
A screenshot from Synthesia’s AI-generated video resembling me, based on a photograph I uploaded onto its platform
When teachers become avatars
In some cases, providers are using the latest technology to produce AI-generated lessons, replacing filmed human presenters with avatars.
Paragon Skills, part of Knovia, has been using AI video platform Synthesia to produce teaching materials for apprenticeship e-learning modules.
Instead of commissioning conventional video production, staff write scripts, select synthetic presenters and generate lessons or workplace scenarios featuring AI avatars.
The company’s programmes include adult care, early years and professional services.
Synthetic presenters allow courses to be updated quickly when legislation changes, and new scenarios can be created without the expense of filming additional material.
Oxbridge Pathways, operated by Liverpool Colleges Ltd, markets AI-generated lessons and a “talking tutor video with lip-sync” across 84 UK qualifications.
It advertises enrolment at £250, plus £20 per unit, excluding final assignment or assessment and awarding-body registration fees.
Oak’s ‘observer at the back of every online classroom’
The automated inspector
A newer market is emerging around quality assurance and inspection preparation.
Apprenticeship software company Oaks advertises an AI inspector called Alana, “who you can talk to”, which conducts spoken mock inspections, examines evidence and produces practice Ofsted report cards.
Another service is described as being an “observer at the back of every online lesson”.
Alana joins online lessons as a muted guest, identifies speakers and drafts teaching observations, which are “sent to the tutor to read and respond, never shown to learners”.
The company says the audio is “deleted once the notes are written”.
Its advertised forthcoming products include a “revenue rescue engine” to identify apprentices at risk of triggering funding clawbacks, and a “predictive flight risk” tool which estimates the likelihood of learners withdrawing and generates “micro-rescue plans”.
A further product, the “learner journey time machine”, promises to turn an apprentice’s progress into an “Ofsted-ready narrative”.
Such systems might enable earlier interventions and better management. But they also raise questions about whether technology is being used primarily to improve education or to optimise funding and inspection outcomes.
Aptem Enhance is ‘proven to help providers to cut admin and focus on learner outcomes’
How much do apprentices know?
Despite the rapid adoption of AI, there is considerable variation in how providers explain its use to learners.
Training providers are not subject to a universal requirement to notify learners when AI assists with marking or feedback.
However, UK data protection law requires providers to be fair and transparent about how they process personal data, and the Information Commissioner’s Office (ICO) has warned that failing to explain AI-assisted decisions could breach these requirements.
Separately, Ofqual has highlighted a strong consensus among assessment experts that students should be informed when AI scoring systems are used – and that those systems should be explainable.
Of the 13 largest providers where FE Week identified evidence of AI use, Kaplan was the only one found to have a clear, publicly accessible statement explaining how it uses AI in relation to learners (although Multiverse told FE Week it informs apprentices when feedback is AI-generated, and says it has made it as easy as possible for apprentices to ask for a human to review.)
Kaplan promises that learners will be told when they receive AI-supported feedback, that consequential decisions will retain safeguards against over-reliance on AI, and that policies will help learners to avoid plagiarism.
And of the 15 biggest providers, FE Week identified publicly accessible, specifically AI-related policies or guidance on learners’ use of AI at just one –Train’d Up Railway Resourcing Limited.
It’s document defines AI misuse as using it to review, rewrite, summarise, paraphrase, improve or adjust the word count of assignments, or failing to follow guidance where AI use is permitted.
Other providers may give learners guidance through internal systems or course materials that are not publicly accessible.
But the disparity raises questions about how clearly apprentices are being told what they can and cannot do with AI – and how much they know about providers’ own use of the technology.
Edtech investor and expert Richard Taylor
Who is watching the machines?
The pace of commercial development is testing regulators’ capacity to keep up.
The Department for Education’s generative AI product safety standards, updated in January 2026, seek to establish safeguards for AI products used in education, but edtech specialist Richard Taylor described them as a “belated acknowledgement” of inadequate safeguards.
He estimated that 95 per cent of current UK edtech AI claims fell outside the standards.
A senior source involved in AI safety work with the DfE believes that more than half of the edtech products being used in FE fall short, and technology suppliers themselves complain that regulators are not providing sufficiently timely guidance.
The government is launching a new edtech certification scheme, but its remit is focused on schools and colleges and the Department for Work and Pensions says it is not currently framed as applying to apprenticeship providers.
The DfE says providers should be transparent about automated decision-making and ensure that learners understand how their personal data is processed.
But the ICO recently identified wider compliance weaknesses in audits of 28 education technology providers, including confusion over responsibilities for handling personal data.
The risks of connecting AI systems to sensitive learner and employer information are becoming more acute.
A recent government cybersecurity survey found that 82 per cent of FE colleges surveyed had adopted some AI tools, but only 66 per cent of those using or considering AI had specific cybersecurity processes to manage the associated risks.
Meanwhile, Google’s Threat Intelligence Group has warned that autonomous AI agents are being used in cyberattacks to access sensitive databases at speeds previously impossible for human attackers.
That raises questions about the security of interconnected learner and assessment records, particularly as AI products gain access to more institutional systems.
Some providers are also discovering that a useful AI product can be unacceptable to employers whose information it processes.
Decoded, a provider specialising in AI training, developed a chatbot to answer apprentices’ questions about end-point assessment.
Senior learner success coach David Parry said it successfully saved coaches time, but employers concerned about data protection objected to apprentices using an external AI system and asked the provider to switch it off.
For training providers, the financial attraction of AI is clear. Automation offers the possibility of increasing learner numbers, reducing administrative costs and improving productivity without a corresponding expansion in the workforce.
But the uncertainty over whether money saved through automation is being reinvested into teaching matters in a sector funded by public money.
Ultimately, the measure of success should not simply be how much money or staff time AI saves, but whether apprentices receive a better education as a result.