Upskilling

What upskilling means

Upskilling is training the staff a business already has to use AI tools well in the work they already do. The phrase carries two choices. The first is who: existing people, not new hires. The second is what: the person’s own tasks, not a general tour of what AI can do. Both choices are the reason the term is worth defining, because most AI training an organisation has seen was delivered the other way round, to anyone available, on examples nobody recognised.

The site’s definition adds a comparison. The training article finds that programmes built around real workflows do better than generic content, and better than staff teaching themselves. Upskilling, in the way this site uses the word, is the version of training that earns that finding.

It is also deliberately plain. It does not promise to turn a team into specialists. It means that the people doing the work today can do it better with the tools that now exist, and know where those tools stop being trustworthy.

How upskilling differs from AI Literacy and one-off Training

Three terms get used as if they were one. They are related, and separating them makes a plan easier to write.

AI literacy is the working understanding a person needs to use AI safely in their own job. It is a state: a person has it or does not. Upskilling is the programme that gets people there and then carries them on to routine use. Literacy is what the programme builds first. Use on real tasks is what it builds next.

A one-off session is an event. It can be useful, and it is often where a team learns the rules, but it is not a programme. The training article treats upskilling as ongoing, practised on real tasks, because the tools and the tasks worth applying them to change faster than a single session can account for.

Knowing about AI and using it are also different things. Many teams have had a talk or a demo and forgotten most of it within a fortnight, because nobody went back to a desk and used what they heard. The training plan article puts it as a change of order: people use the tool on their own work from the first day, and the sessions exist to make that use safer and better, not to replace it.

What the Training figures say

The training article is the source of the site’s figures on this subject, and they are worth reading together.

• Organisations with formal AI training programmes achieve 2.3 times faster AI adoption and 67% higher AI ROI than organisations without one.
• 70% of AI success comes from people, process and change management, with the remaining share down to the technology.
• Digital training applied to real workflows is 93.7% more effective than generic, traditional training.
• Adaptive learning produces 58% higher knowledge retention than static content.
• Structured training correlates with a 40% improvement in measured employee productivity.
• Only 19% of workers have been through a formal AI training programme.

The pattern running through them is relevance. The largest gain in the article comes from training applied to the workflows people already do, not from material delivered in the abstract. Staff retain what they practise on their own tasks far better than what they are told in a session unconnected to their day.

The 19% figure sets the context. Most staff using AI today got there through self-directed experimentation, and the article notes that this is often how shadow AI starts. It also reports what the site sees in its own programmes: teams on bespoke training built from their own workflows close the AI adoption gap in months, while teams left to teach themselves are still closing it a year later, often having settled on habits that a structured programme would have corrected in the first session.

Why the person’s own tasks come first

A task a person already resents is the best teaching material there is. They know how long it takes, they know what a right answer looks like, and they can tell within a minute whether the tool did anything. Nobody has to invent a brief for them.

The briefing article describes the opposite case. A room full of people who were told to attend has no task in mind, so the trainer supplies one, and a supplied example is always easier than real work. Everybody leaves having watched something succeed on a problem they do not have. That is the single biggest reason, the article says, that a training budget produces enthusiasm on the day and nothing measurable a month later.

This is why the team briefing comes first. It is a one-hour meeting about a week before the training, run by the manager rather than the trainer. It names the tasks in scope as tasks and not tools, asks each person to bring one real document, and says what is out of bounds, such as client data or anything under an NDA. A person who cannot name a task worth bringing is telling the business something useful before the money is spent.

What an upskilling Programme includes

The training plan article sets out a sequence that works for a team of any size, and the order matters more than the length of any one part.

1. Rules and first tasks. The approved tools, what data never goes in, and a first real task done live. The article calls the rules the part that has to exist on day one.
2. Prompts for real work. One structure for writing a clear request, and a shared prompt library so the second person in the team starts with a request that works instead of a blank box. The training article notes that these habits carry over when the tool changes.
3. Checking output and finding the limits. People who have seen the tool fail once check far more carefully than people who have only been told it can. Where a decision affects a person, a human stays in charge of it.
4. Making it routine. The team agrees which tasks are now done with the tool as the standard way, writes down who checks and where the result goes, and names an owner for the library.

Two things hold across all four. The rules and the checking habits are the same for everyone, while the tasks are specific to each role. And the programme has a named owner for the whole of it, because the article says a plan with no owner after the first month is a plan that ends.

Core Course or Bespoke Training

The site runs upskilling in two ways, and the choice depends on how alike the work is. A core AI course suits a whole team that needs the same foundations. Bespoke training suits one function whose data and workflows are specific, because the sessions are built from the team’s own documents and tools.

The banning article makes a point that applies to both. A rule people understand gets followed in cases it did not anticipate, and that is where most real decisions happen. Training does more than a memo at that point, because staff need enough AI literacy to judge a case the page never covered.

HR is the usual first example. The HR article says HR is generally the function expected to lead AI upskilling for the rest of the organisation, and that doing so credibly depends on its own literacy coming first. It also says a single all-staff session does not fit HR the way it fits a department using AI for drafting, because HR’s own data is the sensitive material the rest of the organisation is only asked to handle carefully.

How to tell whether upskilling worked

Whether it worked is a separate question from whether it happened, and the two get mixed up more often than either gets checked. Attendance and a certificate say nothing about whether a decision made six weeks later applies what the session covered.

The training article says knowledge retention is the figure worth tracking, and that a programme is judged over time and not by how many people finished it. The HR article adds a behavioural test: can the people who attended explain, in their own words and without notes, why a given AI-assisted decision is or is not defensible, and name the specific safeguard that makes it so. A programme that cannot survive that question asked cold was not really training, whatever the attendance sheet says.

The training plan article gives four measures to take before and after, none needing special software:

• Use. How many people used an approved tool in the past week, and for how many tasks.
• Time. One chosen task, timed the old way and the new way, with the checking included in the new figure. The saving-time article explains why: a saving only exists if the total across checking, fixing, re-prompting and handoff is smaller than before.
• Rework. How often output needed correcting or redoing. If use rises and rework rises with it, the training taught speed without checking.
• Confidence. A simple one-to-five score asked again at the end. A rise in confidence without a fall in rework is a warning and not a success.

Upskilling and Shadow AI

When a business does not provide a trained, approved route, people find their own. That is the shadow AI pattern, and the training article’s 19% figure is the background to it: most staff using AI are doing so without any structured programme.

This changes how the question should be put. A business that says its staff have not started with AI may be wrong, because use of a kind is already under way in personal accounts and browser tabs. The question is whether that use is planned, owned and measured. Upskilling is the way a business moves it from the first state to the third, and the banning article explains why saying no to everything does not do it: a ban moves the same use out of sight.

When people do not want to take part

Resistance is common and is better treated as information. The training plan article sorts it into four worries: that the tool will replace the person, that they will look foolish, that they tried it once and it was poor, and that it is not allowed. Each has a different fix. Private practice and a manager who goes first deal with the second. A colleague whose requests work, used on the resistant person’s own task, deals with the third. A plain written statement of what the training is for deals with the first.

The article also recommends a champion rather than a mandate: one or two curious colleagues, not necessarily senior, given time to answer questions. The internal champion page describes the same role on a project. Peers are believed in a way that managers sometimes are not.

What to do first

Start with the hour, not the course. Run a team briefing a week ahead and see how many real tasks the team can name. Then name an owner, put the rules on one page, choose three tasks per role and take a baseline. If a whole team needs the same foundations, a core course gives them. If one function has data and workflows of its own, build the sessions from those.

Whatever the route, treat the first month as the start. The evidence on this page points the same way each time: training attached to real work, repeated, owned and measured, is what moves the figures. The week-by-week version of that is in how to train your team to use AI.