This week’s People Management report on Boston Consulting Group’s 2026 AI at Work study lands on a finding that should make every L&D and HR team sit up straight.
67% of regular AI users say the technology has made work more enjoyable. Frontline use has jumped from 51% to 74% in a single year. And in the UK, 70% of frontline workers now use AI regularly, well ahead of the 57% global average.
And yet, the bit that should worry anyone responsible for workforce capability, only 36% of workers feel adequately trained to use AI effectively. That number has not moved since 2025.
The numbers, on one page
67% say AI has made their work more enjoyable.
36% feel adequately trained to use it (unchanged since 2025).
86% expect to need significant upskilling within the next five years.
42% of global frontline workers who use AI regularly say they save at least one working day per week. For managers it rises to 56%.
70% of UK frontline workers use AI regularly, vs 57% global average.
49% of UK frontline workers fear AI could replace their role entirely, vs 36% global.
Source: BCG, AI at Work 2026, c.12,000 workers across 14 countries. Reported by Mahalia Mayne in People Management, 3 June 2026.
What the data actually says
BCG surveyed nearly 12,000 workers across 14 countries. The headlines are striking. AI now handles routine tasks for 67% of all respondents, freeing them for more complex work. Among UK frontline workers, 59% believe AI agents will be capable of doing at least half their job within three years. For managers, it is 66%.
That fear is the other half of the story. BCG calls it the “joy paradox”: 67% of people say AI has made their job more enjoyable, but 41% report increased mental strain at the same time.
The training gap is now the productivity gap
Here is the maths every CFO and CPO should be running. 66% of the workers who said AI saved them time also said they had received little or no guidance on how to use it. Half said their organisation has no clear rules at all for how human and AI teams should work together.
That is not just a learning problem. It is a governance, risk and equity problem. Anjali Malik at Bellevue Law put it bluntly in the article: unequal access to AI tools or training could create indirect discrimination risks. If half your workforce is getting an extra working day per week from AI and the other half is not even getting a usage policy, that gap will surface as a performance gap, a pay gap, and eventually a tribunal risk.
The CIPD-backed reporting on this is consistent: enthusiasm without structure burns people out and exposes the business. The fix is not more enthusiasm. The fix is structured, accredited, hands-on training that gives every department the same baseline.
What “structured training” actually looks like
The BCG report quotes Daria Riznyk of Intellias: “Organisations that succeed protect learning time as firmly as they would protect a client meeting.” Translation, ad hoc lunchtime webinars and a Copilot licence do not count. What works is a programme that:
- Covers the full spectrum, prompting, automation, data, governance, ethics, change management.
- Sits at a recognised qualification level, so it stands up to internal scrutiny and external audit.
- Is genuinely applied, workplace projects, real datasets, real workflows, not theoretical case studies.
- Reaches every department, HR, finance, operations, marketing, customer service. Not just IT.
- Is funded, because asking budget holders to find new training money in 2026 is not a winning play.
What happens if you do nothing?
Doing nothing is also a choice. Here is what BCG's data, the People Management coverage, and our own client conversations across UK SMEs suggest the next 12 to 18 months will look like for organisations that leave the gap open.
What "do something" actually looks like
Not every approach to closing the gap is equal. Here is the honest side-by-side of the three options most UK employers actually consider.
A quick ROI sketch
If a regular AI user saves one working day per week, and your average loaded cost per employee is £55,000 per year (UK SME average), then:
Quick fit check: is the L4 right for your team?
15-minute call. Tell us the team size and function. We will tell you which of the funded routes (Level 4, AU0009/10/11 leadership units, or a hybrid) closes the BCG-style gap fastest for your business. No pressure to enrol.
Book a 15-minute fit call →How the AI & Automation Practitioner Level 4 apprenticeship answers this
Our AI & Automation Practitioner Level 4 apprenticeship was built specifically for the gap BCG just identified, the workers who want to use AI well but have never been shown how, in any structured way.
What the programme actually does
• Fully levy funded. £18,000 funding band, no new budget needed.
• 11 to 15 months. Palatable duration, real workplace projects, no coding required.
• Built for non-technical functions. HR, finance, operations, marketing, customer service. Not just IT.
• Up to 5 qualifications included. All Level 3 and above, including Microsoft AB-730.
• Free CMI leadership bolt-on. A no-cost wrap for closed and hybrid cohorts where leadership matters.
• Ofsted Good, 4.9/5 across Google and Trustpilot. Independently verified quality bar.
The programme covers AI ethics, governance and GDPR, no-code automation, change management, data-driven decision making, and Microsoft Copilot in practice. Exactly the structured curriculum BCG’s report says is missing in most organisations today. See the full guide to the Level 4 standard for module-by-module detail, or the cohort delivery options if you are thinking about how to roll it out.
Map your workforce against the AI training gap, in one 30-minute call
Tell us how many people, what they do, and where AI is already saving (or not saving) time. We will tell you which funded route closes the gap, and how it sits in your levy pot.
The HR takeaway
If you are an HR or L&D leader reading this, the BCG numbers give you a simple, defensible business case to take to your exec. Two thirds of your people would enjoy their jobs more with AI. Half of them think their role is at risk. Almost none of them have been properly trained.
You can let that gap widen quietly, and watch it become a retention problem, an inclusion problem and a productivity problem at the same time. Or you can put a funded, structured programme in front of it now, while the levy still pays for it.
That is the conversation worth having this quarter.
Wider context
The BCG finding lands inside a wider 2026 evidence base. Skills England’s Annual Skills Report 2026, published 1 June, named AI as one of five system-level challenges and explicitly highlighted the Level 4 AI & Automation Practitioner apprenticeship as a flagship rapid-response training product. The three AI leadership apprenticeship units (AU0009 / AU0010 / AU0011) close the leadership-layer gap. The levy mechanics explainer shows exactly how to fund it.
Sources & further reading
- Mahalia Mayne, “Two thirds of workers using AI say it has made work more enjoyable, survey finds,” People Management, 3 June 2026.
- Boston Consulting Group, AI at Work 2026, c.12,000 workers, 14 countries.
- The complete guide to the AI & Automation Practitioner Level 4 apprenticeship.
- Skills England 2026 priority sectors, mapped to funded routes.
- Open, closed or hybrid cohort: choosing the right delivery model.
Frequently asked questions
What is the AI Joy Paradox?
The term comes from Boston Consulting Group’s 2026 AI at Work study. It captures the contradiction at the heart of current workforce data: 67% of workers using AI regularly say the technology has made work more enjoyable and is freeing up significant time, while 41% simultaneously report increased mental strain. The paradox is enthusiasm without structure: people love what AI does for them but feel unprepared to use it well.
Why has the “36% adequately trained” number not moved since 2025?
Because most organisations have rolled out AI tools (Copilot, ChatGPT, Gemini) faster than they have rolled out structured training to use them. Licences are cheap and quick. Training is harder, slower and requires committed time. The BCG report makes the case that ad hoc upskilling does not close the gap, a recognised, applied, funded programme does.
What is the “indirect discrimination risk” the article references?
Quoted from Anjali Malik at Bellevue Law in the People Management piece. The argument is that unequal access to AI tools or training could surface as a performance gap, a pay gap, and eventually a tribunal risk. If half your workforce is getting an extra working day per week from AI and the other half is not even getting a usage policy, you have an inclusion problem that the law may take an interest in.
Can the Level 4 apprenticeship really reach every department?
Yes. The standard is built for practitioners across HR, finance, operations, marketing, customer service and beyond. No coding background is required. Every learner builds workflows on the systems they actually use in their function. That is the whole point of the standard, it is not an IT apprenticeship.
How much of an organisation’s levy does this use?
Up to £18,000 per learner, drawn from your apprenticeship levy account. For SMEs (annual pay bill under £3m) without a levy pot, government covers 95% and you co-invest 5%. The levy mechanics explainer walks through this for finance teams.
What is the difference between this and the new AU0009/10/11 leadership units?
The Level 4 apprenticeship is for the practitioners doing the work, who need to build AI capability into their day job over 11 to 15 months. The AU0009 / AU0010 / AU0011 units are shorter (30+ guided learning hours each) and target the leaders setting strategy, governing AI, or driving transformation. Most organisations need both layers.