Almost all of them use it. HEPI's 2026 survey of 1,054 UK undergraduates found 95% use AI and 94% use it for assessed work, but only 48% say their lecturers are helping them build AI skills for their careers. Frequent use is not the same as proficiency, and preparation for AI at work varies. Your organisation's rules on data, checking, disclosure and accountability are yours to teach.
Figures checked on 18 October 2026 against HEPI Report 199 and NACE's July 2026 release. Prompted by Petra Recter's commentary for University World News, 17 October 2026.
Key takeaways
- Use is near universal. 95% of UK undergraduates use AI and 94% use it for assessed work, according to HEPI's 2026 survey of 1,054 students. 12% now put AI-generated text directly into assessed work, up from 3% in 2024.
- Teaching is not. 68% say AI skills are essential, but only 48% say their lecturers are helping them build those skills for work, and only 38% say their university provides AI tools.
- Employers are already seeing it. In the Institute of Student Employers' 2026 survey, a record 92% of UK graduate employers suspected or confirmed cheating in selection, most often undeclared AI in interviews (66%). In the US, half of graduating seniors say they are not building AI skills at all.
- A different rulebook. University AI rules are about academic integrity. Workplace AI rules are about data, clients, accuracy and accountability, and some of them are specific to your organisation, so only you can teach those.
- A funded route. A graduate aged under 25 at the start can take the Level 4 AI and Automation Practitioner with training fully funded up to the band maximum, and a degree does not rule them out. Older graduates can take it too, on the usual rules for apprentices aged 25 and over. A non-levy employer hiring one new can add a £2,000 hiring payment.
- And a clock. If the reported plan for a 25% employer contribution on apprentices aged 25 and over goes ahead, the years before a graduate turns 25 become the cheapest time to train them.
Two stories from the same day this week, read together, describe your next graduate intake. FE Week reported that ministers want to make training anyone aged 25 and over dearer (our breakdown). University World News carried a commentary from Oxford University Press on how universities are trying to teach students to use AI properly. The graduates you hire in 2027 will be young enough for a fully funded apprenticeship, though not all, and most will arrive with AI habits formed mainly at university rather than at work.
What the 2026 numbers say about students and AI
95%
of UK undergraduates use AI in at least one way
HEPI, 1,054 students, surveyed December 2025
48%
say their lecturers are helping them build AI skills for their careers
HEPI, same survey
50.5%
of US graduating seniors say they are not building AI skills
NACE 2026 Student Survey
The Higher Education Policy Institute's third annual survey of UK undergraduates, run by Savanta, describes AI use as near universal. 94% use generative AI to help with assessed work, and the share pasting AI-generated text directly into assessed work has risen from 3% in 2024 to 8% in 2025 and 12% in 2026. Nearly two-thirds (65%) say assessment at their university has changed significantly because of AI.
The gap is in the teaching. 68% of students say AI skills are essential to thrive, but only 48% feel their teaching staff are helping them develop those skills for their careers, and only 38% say their university provides them with AI tools. HEPI's own conclusion is that institutional adoption is lagging behind student behaviour.
"AI literacy and capability must be embedded across the curriculum."
Charlotte Armstrong, co-author and Policy Manager, HEPI, March 2026The American picture, from the National Association of Colleges and Employers, shows the same gap from the other side. Employers in NACE's spring 2026 survey said more than a third of entry-level jobs now require AI skills, nearly triple the share six months earlier. Yet half of graduating seniors (50.5%) say they are not building AI skills, and 31% think those skills will be of little or no importance to their career. It is US data, but UK employers recruiting the same generation should expect a version of it.
UK graduate employers are already seeing it from the recruiting side. In the Institute of Student Employers' 2026 survey of large graduate employers, a record 92% suspected or confirmed cheating in their selection process, up from 85% a year earlier and 78% the year before, and the most common form was undeclared AI use in interviews, reported by 66%. Employers are answering with rules rather than bans: 82% now give candidates guidance on acceptable AI use, up from 55%, and 64% accept AI somewhere in hiring, mainly for CVs and cover letters, while drawing the line at online assessments, interviews and assessment centres. ISE's chief executive, Stephen Isherwood, warned that candidates who misrepresent themselves risk ending up in the wrong job.
92%
of UK graduate employers suspected or confirmed cheating in selection
ISE 2026, a record; 85% in 2025
66%
saw undeclared AI use in interviews, the most common form
ISE 2026
82%
now give candidates guidance on acceptable AI use
ISE 2026; 55% a year earlier
Use is not literacy: the rulebook changes on day one
Students have been taught, or have worked out, the rules of AI at university. Those rules exist mainly to protect the fairness of assessment. The rules at work protect something different, and how much of them a graduate has met varies: some will have had good teaching or a placement, many will not, and none will know your organisation's own rules.
| The question | At university | At work |
|---|---|---|
| What the rules protect | Fair assessment and academic integrity | Customer and personal data, accuracy, and the organisation's name |
| Who gets hurt if it goes wrong | Mostly the student | Customers, colleagues and the business |
| What gets you into trouble | Passing off AI work as your own | Pasting data into the wrong tool, sending an unchecked figure, letting an agent act beyond its brief |
| Disclosure | Declare AI use in assessed work | Say where AI helped, so a reviewer knows what to check |
| Which tools | Whatever is free, or whatever the university provides | Only the tools and accounts your organisation has approved |
A graduate who is skilled at using AI for essays may never have been asked whether a client spreadsheet can go into a chatbot on their personal phone. That is not a criticism of graduates; it may simply never have come up. It does mean the first week of a graduate job is where your rulebook has to be taught, because your organisation's rules cannot have been taught anywhere else.
Disclosure is about to matter more, too: text from Claude models released since August now carries an invisible watermark.
Expect two kinds of new starter
Most graduates will sit somewhere between these two, but the ends of the range need different handling.
The confident, unsupervised user. Fast, confident and productive, with habits formed where the main risk was a misconduct panel rather than a data breach. They need rules and the habit of checking, not encouragement.
The sceptic. NACE's 31%. Some object on accuracy, some on environmental or ethical grounds, and some simply prefer to do the work themselves. They can become your best checkers. They need a clear reason, approved tools, and permission to say when AI is the wrong answer, not pressure.
Give the candidate a one-page AI-written draft with two planted problems: a figure that is wrong and a source that does not exist. Add one question: could the client list behind this have gone into an AI app on your phone? Ask them to review the draft and talk you through what they would change. You will learn more about their AI judgement in ten minutes than from any line on a CV about being proficient in AI.
Universities are catching up, slowly
The commentary that prompted this piece was written by Petra Recter, global director of content and digital strategy for higher education at Oxford University Press, and it is partly about OUP's own product. OUP added an AI literacy module to its Oxford Insight courseware from August 2025 that lecturers can assign to students. It covers six areas: what generative AI is; AI and academic integrity; advantages and disadvantages, including made-up answers; critical thinking and checking outputs; prompting; and a course-specific activity written by an expert in the subject.
HEPI's recommendations point the same way: structured AI induction for every student, AI taught explicitly in the curriculum, clear assessment guidance, access to tools for all, and training and time for staff. All of that is welcome, and most of it transfers to work. Two things do not: what may and may not go into an AI tool at your organisation, and who is accountable for what it produces. And most of the 2027 intake studied before any of this was standard.
What to teach graduates in year one
The first fortnight sets the habit. If nobody tells a graduate what not to paste into an AI tool, they will find out the hard way, and so will you.
Lisa O'Reilly, Director, TESS Group
Your AI rules, on day one
Do: put approved tools, approved accounts and the data that never goes in into induction, with a named person to ask.
Why: the habit a graduate forms in their first fortnight is the one you will be managing for years. Start with how to write an AI policy.
Checking
Do: teach them to check every figure against its source, ask the tool where its answer came from, and ask what it has assumed.
Why: university punishes copying; work punishes confident mistakes. They have usually practised avoiding the first more than the second.
Disclosure
Do: make it normal to say where AI helped.
Why: a reviewer who knows a section was AI-drafted knows what to check. Concealment is the problem, not use.
Briefing
Do: teach prompting as briefing: role, instruction, context, format.
Why: essay prompting and work prompting are different crafts. Our plain-English prompt engineering guide is a good first hour.
Judgement
Do: be explicit about what is never handed to AI, and who signs off what it produces.
Why: as tools start acting on their own, that line matters more. See the four things every AI agent needs.
Building
Do: give the ones with aptitude a route from using AI to redesigning how work gets done.
Why: that is now a job in its own right: the AI automation specialist.
0 of 9 done
University taught them when AI is allowed. Work has to teach them when it is right: is the number checked, is the data allowed to go in, and would you put your name to it?
Rod Doyle, Director, TESS GroupThe funded route for graduate hires
If your graduate is aged 16 to 24 when they start, training and assessment on the Level 4 AI and Automation Practitioner apprenticeship (ST1512) is fully funded up to the funding band maximum, levy payer or not, for starts from 1 August 2026. Not every graduate is under 25, though: mature and postgraduate entrants aged 25 and over can take it too, with the usual employer contribution of 5% at a non-levy employer, or 25% once a levy payer's funds run out. At TESS it runs for 12 months of live online training plus a three-month end-point assessment window, needs no coding, and covers the workplace rulebook above as well as automation and agents.
A degree does not rule them out. The GOV.UK funding policy allows an apprenticeship at a higher, equal or lower level than a qualification someone already holds, as long as it gives them significant new skills and the content is materially different from what they have already studied. For most graduates, AI and automation at work is materially different from their degree; the provider confirms it at initial assessment.
A non-levy employer hiring a new apprentice aged 16 to 24 can also claim a hiring payment of up to £2,000, provided they have not employed that person for more than 90 days before the start and the practical period begins on or after 1 October 2026. The full incentive stack sets out what else can be combined.
And there is a clock. If the reported plan for a 25% employer contribution on apprentices aged 25 and over goes ahead, the years before a graduate turns 25 become by some distance the cheapest time to train them.
Where the training fits
Your graduate intake, and the people who will manage them.
Our Level 4 AI and Automation Practitioner is built for people in non-technical roles learning to use AI safely on real work, then automating it. Graduates aged under 25 at the start are fully funded up to the band maximum; older graduates follow the usual co-investment rules. The next inductions are Friday 20 November 2026, Monday 14 December 2026 and Monday 25 January 2027, live online. If you want a whole intake briefed in a day first, the AI Prompting Accelerator covers prompting and checking in a single session.
TESS Group has trained UK employers since 2009, is rated Good by Ofsted at its last inspection (June 2025), UKPRN 10031093, and delivers live online to employers across England.
Hiring graduates in 2027? Book 30 minutes with Rod. We will look at your intake, what they need to learn about AI at work, and whether the funded apprenticeship route fits.
Book a 30-minute call with Rod → or use the contact formThe honest summary
- Use: 95% of UK undergraduates use AI, 94% for assessed work.
- Teaching: only 48% say lecturers help them build AI skills for their careers; universities are catching up, slowly.
- The gap: most graduates know the university rulebook, which is about integrity. Preparation for the workplace one, about data, checking, disclosure and accountability, varies, and your organisation's own rules are yours to teach.
- Your job: AI rules on day one, then checking, disclosure, briefing, judgement and, for some, building.
- The money: graduates aged under 25 at the start are fully funded on the Level 4 up to the band maximum, a degree does not rule them out, and non-levy employers can add a £2,000 hiring payment for a new employee. Older graduates follow the usual rules for apprentices aged 25 and over.
Frequently asked questions.
How many UK students use AI?
95% of UK undergraduates report using AI in at least one way, and 94% use generative AI to help with assessed work, according to the Higher Education Policy Institute's Student Generative AI Survey 2026, based on 1,054 full-time undergraduates surveyed by Savanta in December 2025.
Are graduates taught AI skills at university?
Partly. In HEPI's 2026 survey, 68% of students said AI skills are essential, but only 48% felt their teaching staff were helping them develop those skills for their careers, and only 38% said their university provides AI tools. Universities are adding AI literacy teaching, but much of it focuses on academic integrity rather than workplace rules on data, checking and accountability.
What AI skills should employers teach new graduates?
Six things, in order: your organisation's AI rules on day one, including approved tools and the data that never goes in; checking figures and sources; disclosing where AI helped; briefing AI tools clearly; judgement about what is never handed to AI and who signs off its output; and, for those with aptitude, using AI and automation to redesign how work gets done.
Can a graduate do an AI apprenticeship?
Yes. GOV.UK funding policy allows an apprenticeship at a higher, equal or lower level than a qualification someone already holds, provided it gives them significant new skills and the content is materially different from their previous study. The Level 4 AI and Automation Practitioner (ST1512) usually meets that test for graduates; the provider confirms it at initial assessment.
Is there funding to train a graduate in AI?
If the graduate is aged 16 to 24 at the start, the apprenticeship's training and assessment are fully funded up to the funding band maximum, levy payer or not, for starts from 1 August 2026. Graduates aged 25 and over can also take an apprenticeship, with an employer contribution of 5% at a non-levy employer, or 25% at a levy payer whose funds have run out. A non-levy employer hiring a new apprentice of that age can also claim a hiring payment of up to £2,000, provided they have not employed them for more than 90 days before the start and the practical period begins on or after 1 October 2026.
How are UK graduate employers responding to AI in recruitment?
Mostly with rules rather than bans. In the Institute of Student Employers' 2026 survey, 82% of employers gave candidates guidance on acceptable AI use, up from 55% a year earlier, and 64% accepted AI somewhere in hiring, mainly for CVs and cover letters. The line is drawn at online assessments, interviews and assessment centres, where undeclared AI use was the most common form of suspected cheating, reported by 66% of employers.
What is AI literacy?
AI literacy is knowing what AI tools can and cannot do, using them with the right data and the right tools, checking what they produce, being open about where they helped, and judging when not to use them at all. At university it is mostly framed around academic integrity; at work it is about data, accuracy and accountability.
Rod Doyle is a Director of TESS Group, an Ofsted-rated Good apprenticeship provider founded in 2009 that delivers the AI and Automation Practitioner Level 4 (ST1512) and AI leadership units live online across England. He writes the employer guides on this site. LinkedIn.
Sources: HEPI, Student Generative AI Survey 2026 (HEPI Report 199, Rose Stephenson and Charlotte Armstrong, March 2026; 1,054 full-time UK undergraduates, surveyed by Savanta in December 2025); NACE, Ready or Reluctant?, 22 July 2026, for the US employer and graduating-senior figures; the Institute of Student Employers' Student Recruitment Survey 2026, as reported by Resultsense and the Guardian on 14 October 2026, for the UK employer figures and Stephen Isherwood's comments; Petra Recter, How do institutions embed AI literacy in higher education?, University World News, 17 October 2026; Oxford Insight, AI Literacy Module, for the module launch date and contents; GOV.UK, Apprenticeship funding, policy for new starts from 1 August 2026, for funding, prior qualifications and the hiring payment. Checked 18 October 2026. TESS Group delivers AI apprenticeships and has a commercial interest in employers choosing structured training.