It is hard to have a conversation about learning at the moment without AI coming up. Clients are asking whether it can make projects cheaper and faster. Vendors are promising whole courses at the press of a button. And the tools themselves are improving every few months.
We use AI regularly in our own work, including AI presenter platforms such as Synthesia and Colossyan, and generative tools for drafting and imagery. Here is our honest view of where it genuinely helps, and where it doesn’t.
Where AI genuinely helps
Getting from blank page to first draft
AI is very good at producing a starting point: a rough outline, alternative ways to phrase a tricky explanation, a set of draft quiz questions or a first pass at a script. It doesn’t replace the thinking, but it speeds up the moment when there is something on the page to react to.
Presenter-led video at a sensible cost
AI presenters make it practical to produce short explainer and update videos without studios, crews or reshoots. When a policy or process changes, the script can be edited and the video regenerated in hours rather than weeks. For content that changes often, that is a real advantage.
Translation and versioning
Producing a course in several languages, or in versions for different audiences, used to be one of the most expensive parts of a project. AI makes first-draft translation and voiceover much faster, although it still needs checking by a fluent reviewer.
Illustrative imagery
For scenarios, characters and concept visuals, generative imagery can fill the gap between generic stock photography and expensive custom illustration, as long as it is art directed with care.
Where it doesn’t help, yet
Working out what really matters
The hardest and most valuable part of any learning project is deciding what people need to know and do. That depends on conversations with subject experts, an understanding of the audience and judgement about the organisation. AI can summarise a source document, but it can’t tell you which three points will change behaviour on the shop floor.
Accuracy in specialist subjects
AI tools can produce confident, fluent content that is subtly wrong. In technical, clinical and regulated subjects, that is a serious risk. Everything needs expert review, and the review needs to be just as rigorous as it would be for human-written content.
Tone, trust and brand
People notice when communications feel generic. During organisational change in particular, the tone has to feel like it comes from people who understand what colleagues are going through. That takes a human ear.
Questions to ask before using AI on a project
- Who will check the content for accuracy, and are they given time to do it properly?
- Do your organisation’s policies allow source material to be uploaded to AI tools?
- Will an AI presenter suit the audience and the subject, or would a real person be more credible?
- Who owns the generated assets, and can they be updated later?
Our approach
We treat AI as a set of tools within a careful process, not a replacement for one. We use it where it saves time or opens up options that wouldn’t otherwise be affordable, we are open with clients about where and how it has been used, and every piece of content is checked by people before it reaches learners.
Used that way, AI makes good learning faster and more affordable. Used as a shortcut around the thinking, it simply produces more of the wrong thing, more quickly.
