AI Literacy

What AI Literacy means

AI literacy is the working understanding of AI that staff need to use it safely in their own jobs. The word working matters. It is not a qualification or a technical skill. It is enough understanding to use a tool well, to know when its answer should not be trusted, and to know what may and may not be put into it.

The site’s own definition adds the part that shapes everything else. It is built around the specific workflows staff already do and not delivered as generic awareness content, because that relevance is where the largest effectiveness gain comes from. That is also why a bespoke programme starts from your own documents.

Understanding, not Technical depth

The HR article gives the clearest description of the level. The literacy that makes a training session credible is closer to knowing what a spreadsheet formula can and cannot be trusted to do than to understanding how one is implemented. A person does not need to know how a model works inside. They need to know how it behaves.

In practice that means recognising a few things. A chatbot’s confidence is not evidence. A prompt’s phrasing changes its output more than most people expect. A tool trained on one organisation’s data will not automatically behave the same way on another’s. The article says a generalist team can reach this level in a single session built around its own work. It is a different claim from making them AI specialists.

Built around the work people already do

The site’s article on what makes AI training stick reports that digital training applied directly to real workflows is 93.7% more effective at moving staff towards organisational expectations than traditional, generic training. The reason it gives is relevance. Staff retain and apply what they practise on their actual tasks far better than what they hear described in the abstract.

The same article reports that adaptive learning, where the content adjusts to how a person is progressing, produces 58% higher knowledge retention than static content delivered identically to everyone. It says retention is the figure worth tracking, since finishing a course and being able to use what it covered are not the same thing. Those numbers are the article’s, and this page repeats them as it states them.

What the Law says

The site’s answer on the home page is that Article 4 of the EU AI Act requires organisations to ensure AI literacy among their staff. That obligation is the Act’s. It is the reason buyers now search for the term, and it is why the site treats the subject as a compliance one.

This page goes no further than that sentence. It does not say what counts as enough, who counts as staff, or what a regulator would look for, because the site does not publish those things. This page is not legal advice, and no audit or course substitutes for advice from a qualified lawyer. The EU AI Act page covers the wider regulation, and EU AI Act and GDPR content is built into every foundational programme and not bolted on at the end.

What it includes in practice

The all-staff briefing shows what the site treats as the core. It covers what generative AI is, what a large language model is, and everyday examples of use. Then it spends the larger part of the hour on risks and rules: hallucinations, data privacy and information sensitivity, approved tools against third-party ones, what can go wrong and how to report an incident.

It also covers the regulatory framework and its core principles, the internal guidelines as a checklist of what staff can and cannot do, and the AI systems register. The foundations course adds skills, such as writing prompts for day-to-day tasks, and its outcomes include explaining what generative AI and a large language model are, and what a hallucination is. Together these are the working understanding the definition describes.

The first habit: checking

The habit the site returns to most is checking. The hallucination page says a person who has checked a hallucinated answer once, on a task they recognise, carries that habit into every tool they use afterwards. A person who has only been warned about the concept in the abstract usually does not.

That is the reason for building literacy around real work. Understanding that a model can be confidently wrong is easy to say and hard to feel. Seeing it happen on your own document is what makes the habit stick.

Who needs it

Everyone who uses AI at work needs some of it, and the site names three groups that need more. Leaders come first. The leadership masterclass is recommended before staff training begins, and it covers the organisation’s own approved tools, data protection, human responsibility and verification, and ethics, bias and oversight.

HR comes next. The HR article says HR is generally the function expected to lead AI upskilling for everyone else, and that doing so credibly depends on understanding what generative AI can and cannot reliably do. It says that is why a curriculum for HR has to run before HR is asked to run one for anybody else. The third group is owners. The ownership article says an owner does not need to be technical, but needs enough AI literacy to ask good questions, and that training in responsible and compliant use gives owners the vocabulary for it.

Why Rules need it

A policy on its own does not do the work. The site’s article on why banning AI backfires says a rule people understand gets followed in cases it did not anticipate, and that is where most real decisions happen. Staff need enough AI literacy to judge a case the page never covered.

The governance article makes the same point about oversight. Basic literacy is part of what makes a checkpoint effective: knowing that a model can produce a confident hallucination in the same tone as a correct answer is what lets a reviewer catch a wrong result and not approve it on confidence alone. A checkpoint is only as real as the person running it. See AI governance for the wider rules.

A worked example

This is an illustration, not a real client. A small company writes a one-page rule on what may go into an AI tool and sends it round. Three weeks later a member of staff pastes a customer complaint, with the customer’s name and address, into an unapproved assistant to draft a reply. The rule listed client names and contact details as off limits, but the person read it as covering contracts only.

The rule was fine. The gap was understanding. The company runs a short session on its own documents. Staff try the approved tool on a redacted complaint, see a confident wrong detail in the reply, and work through why the name and address should never have been entered. The rule has not changed. Everyone now knows what it is for. That is what literacy adds to a policy.

Getting a team ready

The site’s article on briefing a team says people arrive at training with nothing to try it on, and that is a decision somebody made by not making it. The fix costs one meeting, about a week before the session, run by the manager and not the trainer, and it takes an hour.

It covers three things. First, which parts of the team’s work are in scope, named as tasks and not tools. Second, an instruction to bring one real example each. Third, what is not allowed near the tools, such as client data or anything under an NDA. If the team cannot name a single task worth bringing, that is worth knowing before the budget is committed. The briefing article has the detail.

Formal Training against Self-Teaching

The training article reports that organisations with formal AI training programmes achieve 2.3 times faster AI adoption and 67% higher AI ROI than organisations without one. It also reports that only 19% of workers have been through a formal programme, with most staff experimenting on their own.

Self-teaching still builds some skill, but the article says it is slower and less consistent, and that it can settle habits a structured programme would have corrected in the first session. It also says the effort should be ongoing and not a single event, because the tools change faster than a one-off session can account for. These are the article’s figures and the site’s own measurement is not involved.

Where it is taught

The six core AI courses run to a fixed outline and are taught on the tools an organisation already has open and on its own material. The foundations course is the starting point for all staff. The all-staff briefing is built to reach a whole organisation at once, and the leadership masterclass is for leaders before staff begin.

A bespoke programme builds the same understanding from your own processes and documents. Two of the trainers, Lucia and Somangsu, list AI literacy among the programmes they deliver. The rest sit in the wider training catalogue.

Mix-ups worth avoiding

Four confusions come up often enough to name. The first is treating AI literacy as technical training, when the level asked for is about behaviour and not construction. The second is treating it as awareness content, a slide deck or a video, when the site’s whole argument is that generic content is the weaker route.

The third is treating a one-off session as done. The site’s advice is ongoing, and to measure retention and not completion. The fourth is treating literacy as only the concern of the staff who use AI most. Leaders, HR and the owners of a tool each need it for their own part of the job.

Where it goes wrong

One failure to avoid is generic training with a tick at the end. It produces enthusiasm on the day and little a month later, and it leaves the rule and the understanding of the rule separate. Another is a training day with nothing to work on, which is why the briefing comes first.

A third is treating literacy as a substitute for a checkpoint. A trained reviewer still needs a defined point in the process where they check, staffed and used under deadline pressure. The last is treating the training as proof of compliance. This page makes no such claim, and the site says no course substitutes for advice from a qualified lawyer.

Where it appears in an Audit

Literacy shows up in an audit as a question about people and not about software: whether a team can and will use what gets recommended. The article on what an AI audit actually checks treats adoption as part of the recommendation and not as a training problem to solve after the tool is bought.

Once a workflow is built, training is delivered to the people who will run it. That closes the loop the audit opens, and it is why the site places training at the start of the sequence and again at the end.