Happy Monday!
You have probably sat in the meeting where someone says it. The budget review is running long, the L&D line item comes up, and a voice from finance or ops asks: "Why would we spend money training people in AI if they might leave in six months?"
It is a reasonable-sounding question. It misses the point.
A recent piece in HC Magazine reported that employers now have an obligation to build AI skills across their workforce, even when employees leave. That framing should pull your attention, because it names what I have been watching shift for the past eighteen months. AI upskilling has moved from a perk you offer to attract talent into something closer to a duty of care. It's a hygiene skill like being able to use e-mail or utilise the web. I'm just about old enough to still remember managers who would say 'I don't do emails' and that would be an acceptable turn of phrase. Could you imagine someone saying that today?
The logic behind that shift changes how leaders need to think about training budgets, retention risk, and what "good employer" actually means in 2026.
The old calculus was simple. Training costs money. People leave. So training people who leave is wasted money. Tidy and wrong. Cue the sighs.
Why it was wrong and short-sighted was because it treated skills as private assets that walk out the door with the employee. Some do. But AI literacy is different from, say, learning a proprietary CRM. When your workforce understands how to use AI tools responsibly, the benefit shows up in how they make decisions, handle data, flag risks, and collaborate while they are still with you. The value is extracted in real time.
There is a second reason the old logic fails. Regulators and industry bodies are starting to treat AI competence the way they treat health and safety training. You would not skip fire drill inductions because new hires might resign in Q3. The obligation exists because the risk exists right now, for every person in the building. The same principle applies to AI as a hygiene skill. If your people are using generative AI tools (and they are, whether you have sanctioned it or not), then the ethical and legal expectation is that you have equipped them to use those tools properly.
This is where I see a lot of leaders get stuck. They want to invest in AI skills. They can see the case for it. But they are blocked by a belief that feels solid: "We cannot justify the spend if we cannot guarantee retention." That belief treats training as a cost centre. It frames every pound spent on development as a bet on whether the individual stays long enough to generate a return.
The frame is backwards. The only question worth asking is: what happens to your organisation if nobody here understands AI properly?
Because that is the actual risk. Untrained people staying and making decisions they are not equipped to make.
I spent the first few years of generative AI adoption pleading with companies to start AI literacy programmes for exactly this reason, and the cost was concrete. It showed up in poor procurement decisions, in teams adopting tools without understanding bias or data privacy, in middle managers unable to evaluate vendor claims. The absence of AI skills is already expensive and creates a two-tier workforce between the cans and the can-nots. Most organisations just have not learned to measure it yet.
The HC Magazine piece frames this as an employer obligation, and I think that language will stick, because it tracks with how employment law tends to evolve. Duties around digital skills, data handling, and workplace safety have all followed the same pattern: voluntary best practice becomes expected standard becomes enforceable requirement.
My prediction: easy one, within two years, AI literacy training will appear as a named requirement in UK public sector procurement frameworks and in regulated industry compliance checklists. The signal to watch out for is the first major tribunal or regulatory finding where an employer's failure to provide AI training is cited as a contributing factor in a data or decision-making failure. When that case lands, which is pretty much inevitable now, the voluntary window closes fast. I would change my mind if the current government deprioritises AI regulation in favour of growth-first policy, but the direction of travel from the EU AI Act and existing UK sector regulators makes that unlikely. Article 4 of the EU AI Act means that your organisation, if it operates or serves people in the EU, should already be undertaking AI literacy training. You might already not be compliant.
So if you are a people leader or L&D director reading this and wondering whether to push that AI training budget through, treat it as infrastructure. The people who leave will leave with skills they can use elsewhere, and that is fine. The people who stay will be the ones keeping your organisation literate and able to make good decisions with tools that are only getting more powerful.
One thing to try this week: audit how many of your employees are already using generative AI tools without any formal guidance. That number is your starting point, and probably your best argument for building real AI literacy across your organisation.
Frequently Asked Questions
Do employers have a legal obligation to provide AI training?
The obligation is still emerging in most jurisdictions. Employers face growing expectations around AI competence similar to existing health and safety duties, and regulators are moving towards treating AI literacy as a workplace standard. The trend points towards formal requirements, particularly in regulated industries and public sector contracts, following the pattern set by data protection and digital skills duties.
Why should companies invest in AI training if employees might leave?
Because the value of AI training is realised while the employee is still with you. Trained staff make better decisions, handle data more carefully, and flag risks in real time. The bigger cost is untrained people remaining in their roles and making poor choices with AI tools they do not understand.
What happens when organisations delay AI upskilling?
Delays result in unmanaged AI tool adoption, where employees use generative AI without understanding bias, data privacy, or accuracy risks. This leads to poor procurement decisions, vendor dependence on tools nobody internally can evaluate, and data handling errors that carry real regulatory exposure. Most organisations have not yet learned to measure these costs, but they are already accumulating.
How should L&D leaders position AI training in budget discussions?
Position it as infrastructure, the same category as compliance training or cybersecurity awareness. Frame the conversation around organisational risk: what decisions are being made right now by people who lack AI literacy? That reframes the discussion from "will they stay?" to "what breaks if nobody here understands this?"
What is a good first step for building AI literacy in an organisation?
Start by auditing how many employees are already using generative AI tools without formal guidance or policy. That number reveals your current exposure and gives you a concrete, data-backed case for investment. From there, focus on practical, role-relevant training, so the learning connects to decisions people are already making.