What Generative AI Is, and What It Is Not
Share
Generative AI is the term behind almost every tool you have used at work in the last two years. It has a precise meaning worth knowing.
Direct answer
Generative AI produces new text, images, audio or code rather than classifying or predicting from existing data. That is the whole distinction. It matters because producing new content means invention is a feature of the system rather than a fault in it.
Based on the AI and Automation Practitioner standard (ST1512), a Level 4 apprenticeship for people already in work. Funding rules described here are the England position and are set by the funding body.
Potentially free for you
The learner never pays. Employers potentially pay nothing either, and where the apprentice is aged 16 to 24 the government funds the whole cost even if the employer does not pay the levy.
Closing any gap. A levy transfer from another employer covers 100% of the cost. Unused transfer allowance is not carried forward at the end of the tax year, and it must be agreed before the apprentice starts.
From 1 August 2026 the government stopped adding the 10% top-up to new funds entering levy accounts, so there is less in the system than there was.
Part of our guide to AI Explained in Plain English for Work.
The distinction that defines it
Older AI mostly classified or predicted. Is this email spam. Will this customer churn. What is the number next month.
Generative AI produces. Write this email. Draw this picture. Summarise this document. Explain this code.
That changes what correct means. A classifier is right or wrong. A generated draft is better or worse, which is a different kind of judgement.
It also changes testing. You cannot write a test that checks a generated email is correct, which breaks conventional quality assurance.
Why the change happened when it did
The underlying approach existed for years. What changed was scale, both in data and in the computing used to train.
Beyond a certain scale, the systems started producing output good enough to use rather than good enough to demonstrate.
That crossing point is what made this a workplace issue rather than a research one.
It also means capability improvements arrive suddenly rather than gradually, which is why planning around this is difficult.
What it is genuinely good at
Form. Matching a format, a register, a structure. This is where it is most reliably useful.
Transformation. Turning bullets into prose, prose into bullets, technical into plain, long into short.
First drafts. Getting past a blank page, which is a real cost in most jobs and is rarely measured.
Explanation. Restating something in different terms until it makes sense to a particular reader.
There is more on this in our guide to AI explained in plain English.
What it is not
It is not a database. It has no reliable record of facts and no way to tell you which parts it is sure about.
It is not reasoning in the way a person does, whatever the marketing says. It produces text that describes reasoning, which is different.
It is not deterministic. The same input can produce different output, which is a feature and a testing problem.
It is not neutral. It reflects the patterns in its training data, including the ones nobody would have chosen.
What changes about risk
The output is plausible by construction, so a wrong output looks exactly like a right one.
Volume changes things. A person producing forty wrong sentences a day is noticeable. A tool producing them is not.
Attribution gets harder. When generated content is wrong, tracing where the error entered a document takes real effort.
That is why the checking step is the whole of good practice, and why organisations that skip it have incidents.
Where UK rules differ from what you will read online
Most guidance about generative AI online is written for the United States, and two things differ materially here.
The first is data protection. UK GDPR applies to anything you put in, which is a stricter starting position than most US-written advice assumes.
The second is that the UK has taken a regulator-led approach rather than a single AI statute. Your sector regulator's guidance is what applies to you, and it may be more specific than any general rule.
For a UK employer that means checking with your own sector regulator or professional body rather than reading about the EU AI Act and assuming it transfers.
Using it well at work
Use it where form matters and facts come from you. That is the shape of nearly every safe use.
Give it the source material rather than relying on its memory.
Check anything specific. Numbers, names, dates, rules, citations.
Decide who checks before it reaches a customer, and name a person rather than a principle.
Key takeaways
- It produces new content rather than classifying or predicting.
- Correct becomes better or worse, which breaks conventional testing.
- Strongest at form, transformation, first drafts and explanation.
- It is not a database and it is not deterministic.
- A wrong output looks exactly like a right one, by construction.
Frequently asked questions
What is generative AI in simple terms?
Systems that produce new text, images, audio or code rather than classifying or predicting from existing data. That is the whole distinction.
How is it different from older AI?
Older AI mostly classified or predicted. Generative AI produces content, which changes what correct means and breaks conventional testing.
What is it best at?
Form and transformation. Matching a format, turning bullets into prose, making technical writing plain, and producing first drafts.
What is it bad at?
Anything requiring reliable facts. It has no record of what it knows and no way to tell you which parts it is confident about.
Why is the same question answered differently each time?
It is not deterministic. Output varies by design, which is useful for drafting and a genuine problem for testing.
Is generative AI biased?
It reflects the patterns in its training data, including ones nobody would have chosen. That is a design consequence rather than an accident.
What changed to make it useful?
Scale in data and computing. Beyond a certain point the output became good enough to use rather than good enough to demonstrate.
How do I use it safely?
Use it where form matters and the facts come from you, give it the source material, check anything specific, and name who reviews before it reaches a customer.
The funded route, in figures
Figures are the England position. Levy-paying employers usually draw the whole cost from the levy they already contribute, so they pay nothing extra. Employers below the £3 million pay bill threshold pay nothing for an apprentice aged 16 to 24. For an older apprentice the government funds 95 per cent, and a levy transfer from a larger employer can cover the rest. Eligibility and funding rules are set by the funding body and can change.
Where to go next
If you want to build this judgement into a recognised qualification using your own work, the funded route does that.
We help UK employers and employees access government-funded AI training. Training is delivered by third-party providers. We do not guarantee a place, a qualification, a job or any particular outcome. Eligibility and funding rules are set by the funding body and can change. Always check the current position before you commit. See our disclaimer.