The explosion of AI has naturally brought much excitement to everyone working in eLearning. A study released last year found 10% of ChatGPT’s 700 million weekly active users were using the platform for their own personal self-learning. It doesn’t take an expert to see how this could translate to workplace learning and development. And with opportunities in personalised content creation, instant automated feedback, and predicting where learners need more support, eLearning creators are desperate to dive in. It’s easy to get carried away with the potential of AI in eLearning design. But in risk‑averse, highly regulated industries L&D needs to tread carefully.
That includes all financial services (banking, insurance, asset management), the legal profession, the public sector, and many more. If you’re an L&D professional working in one of these industries, you face a genuine Catch-22: You need proof AI works safely before you can use it, and you can’t get proof without using it.
It’s like trying to learn to swim without getting in the water, and it’s leaving many frustrated individuals unable to harness the technology.
There is a route forward, though. While you can’t test AI’s potential to the same degree as your L&D peers in tech or retail, there are plenty of safe ways to experiment without diving in headfirst and risking compliance issues.
1. Use AI as a brainstorming copilot
AI shouldn’t do all of the work from top to bottom, and it doesn’t need to. Sometimes you just need a fresh perspective – and AI can be that creative spark to complement what you’ve already done.
You can use it to generate outlines, learning objectives, or alternative ways to chunk content. But – crucially – you must always set clear boundaries in your prompt. For example, you might prompt it with:
‘Give me three metaphors to explain the tension of using AI while staying GDPR-compliant. Do not offer any new ideas beyond this prompt.’
‘Review the following learning outcomes and tell me how well they align with historical psychological frameworks and current best practice. Do not rewrite or edit any of the outcomes.’
This is a safe use case because you’re not feeding it sensitive data. You’re simply asking for structure and ideas, just like bouncing thoughts off a colleague. AI accelerates, but humans approve.
2. Translation and localisation
Here’s where AI really shines for global audiences, but with caveats.
It’s brilliant for first-pass translations or cultural adaptation suggestions, saving time for human linguists later. Different models perform better in different languages, so choose wisely – and make sure you never skip the human review.
AI can also flag idioms or phrases in your English content that won’t translate well, which is another time-saver before translation. AI can speed up localisation, but cultural nuance still needs a human touch to ensure the content is fully understood.
For example, we recently worked on digital anti-racism training for thousands of people across a global law firm, and we worked closely with the firm’s leaders and EDI experts throughout to check that every single word used was appropriate, culturally sensitive, and would resonate with its intended audience in different countries.
3. Accessibility checks
Before AI, I would spend countless hours writing alt text, checking colour contrast, and simplifying dense content into plain English.
Now, LLMs can make this faster, more consistent and easier to iterate. For example, you can take a particularly wordy paragraph and prompt an AI model to suggest a more cognitive-friendly version, helping to keep content accessible for learners who benefit from clearer language. You can also describe an image and its context, or show one (depending on the tool), and generate a solid starting point for alt text.
Even visual design becomes easier: if a project insists on a restrictive colour palette, ask an LLM for accessible ways to apply it across headings, diagrams, and layouts. The suggestions aren’t always perfect (especially when it comes to colour precision) but they help spark ideas when you’re stuck.
On the multimedia side, tools like Adobe Premiere Pro let you auto‑generate transcripts and closed captions with surprising accuracy. You can also take an LLM-driven approach to file conversion that saves time when shifting between caption formats (like SRT and VTT).
With these approaches, you’re improving inclusivity without exposing sensitive data. If your organisation has a secure LLM, even IP-protected content can be checked safely. In short, AI can help make your content more inclusive without compromising security.
4. Fictional world-building
Creating scenarios for soft skills training can be creatively rewarding, but it can also be a big time sink, as most eLearning designers will know.
AI can help here by generating example scenarios, graphic design briefs, or photo ideas based on your audience, learning outcomes and tone.
It won’t replace expert designers, but if you’re a small team or working with tight budgets, it’s a great way to level up your content with speed and ease.
AI in eLearning design: Humans always approve
Using AI in eLearning design feels a bit like having a robot junior assistant: fast, eager, and full of ideas, but you wouldn’t forward on a client proposal without checking it yourself first.
The same principle applies here: AI accelerates, but humans approve. By starting small, even highly regulated sectors like financial services can begin to see the benefits.
Your next read: The AI-powered workforce: Where are you on the ‘three-loop journey’?
