AI in learning design: How to use AI across the ADDIE Model
AI in learning design speeds up production, but it won't fix poor thinking. See how Hungry Minds uses AI across every stage of the ADDIE model…
Read moreAI and emerging tools resources
AI and emerging tools are changing how learning is built, delivered and personalised. Some of it is transformative. Some of it is vendor hype dressed up as innovation.
This hub separates what actually works from what sounds impressive in a pitch deck. This hub, part of the Hungry Minds learning resources library, covers AI authoring, adaptive learning, personalisation, analytics and the emerging tools reshaping the field. Built for L&D leaders, learning designers and decision-makers evaluating where to invest and what to ignore.
The ethics of AI in learning design
Navigating moral considerations in AI-assisted design As a learning designer, you hold immense power in your hands – the power to revolutionise…
Read moreNeed help with AI and emerging tools?
Let’s talk
Success stories
More case studies Client work: in action
City of Melton
City of Melton
Hungry Minds designed three accessible eLearning modules on gender equality and family violence for Melton City Council. See the results.
Read article
What is AI in learning?
AI in learning refers to the application of artificial intelligence, including generative AI, machine learning and natural language processing, to the design, development, delivery and evaluation of learning programs. It spans everything from AI-assisted content authoring to adaptive learning pathways that respond to individual learner performance in real time.
The hype cycle around AI in L&D is loud. Every platform claims to be “AI-powered.” Every vendor promises personalisation at scale. The reality is more nuanced: AI is genuinely useful for specific tasks (content drafting, data analysis, learner routing, feedback generation) and genuinely limited for others (instructional strategy, emotional intelligence, contextual judgement, quality assurance). Knowing the difference is the skill.
We use AI as a tool within our design process, not as a replacement for it. It accelerates research, drafting and iteration. It does not replace the instructional thinking, stakeholder collaboration and quality standards that make programs work. AI makes good designers faster. It does not make non-designers good.
Where AI is genuinely useful now
Not every AI application in learning is ready for production. These are the areas where the technology is mature enough to deliver real value today, not in a future roadmap.
Content drafting and ideation
Generative AI (ChatGPT, Claude, Gemini) accelerates first-draft writing, brainstorming, scenario generation and question writing. It reduces blank-page time from hours to minutes. But it requires a skilled designer to prompt well, evaluate output critically and edit for accuracy, tone and instructional quality. AI drafts. Humans design.
Content summarisation and restructuring
AI excels at taking dense source material (policies, procedures, research papers, SME transcripts) and restructuring it into learner-friendly formats. It compresses, simplifies and reorganises faster than a human can, freeing designers to focus on sequencing, practice design and assessment.
Assessment and question generation
AI can generate quiz questions, scenario stems and distractor options from source content at speed. The quality varies, and every item still needs human review for accuracy, ambiguity and alignment with learning outcomes. But as a starting point, it cuts assessment development time significantly.
Translation and localisation
AI-powered translation has reached a quality level that makes first-pass localisation viable for many content types. It still requires human review for cultural nuance, technical terminology and tone. But for large-scale multilingual programs, it reduces cost and timeline dramatically.
Learner support and chatbots
AI chatbots can answer learner questions, provide hints, explain concepts in different ways and offer practice feedback within a learning environment. When well-designed (scoped to specific content, with clear boundaries), they extend support beyond what a facilitator can provide at scale.
Learning analytics and pattern detection
Machine learning can identify patterns in learner data that humans miss: drop-off points, common misconceptions, time-on-task anomalies, engagement clusters. This data informs iterative design decisions and helps L&D teams focus improvement effort where it matters most.
Emerging technologies to watch
These technologies are not yet mainstream in most L&D environments, but they are maturing fast and worth understanding for medium-term planning.
Adaptive learning platforms
Adaptive learning uses algorithms to adjust content, sequencing and difficulty based on individual learner performance. If a learner demonstrates mastery, the system skips ahead. If they struggle, it provides additional support. The promise is personalisation at scale. The reality requires significant content architecture and clean data to work well.
AI-generated video and avatars
Tools like Synthesia and HeyGen generate video from text using AI avatars. The quality is improving rapidly. Use cases: rapid localisation, presenter-less explainers, prototype videos for stakeholder review. Limitations: uncanny valley for some audiences, limited emotional range, brand/tone control challenges.
Voice cloning and synthetic narration
AI voice synthesis can produce narration from text without recording a human. Useful for: rapid prototyping, content updates without re-recording, multilingual narration at lower cost. The ethical considerations (consent, disclosure, deepfake risk) are real and unresolved in many organisations.
Immersive learning (VR/AR/XR)
Virtual and augmented reality create immersive practice environments for high-risk, high-consequence or hard-to-access scenarios: surgical procedures, hazardous environments, equipment operation, empathy-building experiences. The hardware and development costs remain high, but ROI is proven for specific use cases where real-world practice is dangerous, expensive or impossible.
AI-powered coaching and feedback
AI systems that observe learner behaviour (writing, speaking, decision-making) and provide personalised feedback in real time. Applications: presentation coaching, writing improvement, negotiation practice, clinical reasoning. Still early for most L&D teams, but the trajectory is clear.
Learning experience platforms (LXP)
LXPs use AI to curate and recommend content based on role, interests, skill gaps and peer behaviour. They sit alongside (or on top of) an LMS and aim to create a Netflix-style discovery experience for learning. Useful for: self-directed development, large content libraries, organisations moving beyond mandatory compliance toward learning culture.
Video and multimedia
Video is the most engaging and the most expensive medium in learning. It excels at demonstration, storytelling, expert credibility and emotional connection. It struggles with dense information, interactivity and rapid iteration. Use it where its strengths matter. Do not use it as a default for content that works better as text or interaction.
What AI cannot do (yet)
The limitations matter as much as the capabilities. Overselling AI leads to poor implementation, wasted budget and learner experiences that feel hollow.
Instructional strategy
AI cannot diagnose a performance gap, determine whether training is the right intervention, choose between modalities based on context, or design a learning architecture that accounts for organisational politics, learner motivation and implementation constraints. Strategy remains a human skill.
Quality assurance
AI generates plausible content that is sometimes wrong, sometimes biased, and sometimes subtly off-target. It cannot reliably verify its own accuracy, assess cultural appropriateness, or judge whether content meets regulatory requirements. Human review is non-negotiable for anything going to production.
Stakeholder collaboration
The design process is relational. It involves navigating SME relationships, managing competing stakeholder priorities, reading political dynamics, and building trust with clients. AI has no role here.
Emotional and cultural intelligence
Content dealing with sensitive topics (trauma, diversity, cultural safety, mental health) requires human judgement that AI does not possess. Tone, framing, language choices and representation decisions in these areas need lived experience and professional sensitivity, not pattern matching.
How to evaluate and adopt
The pressure to "do something with AI" is real. But adopting technology without a clear learning problem to solve is how organisations waste budget and erode trust.
Start with the problem, not the tool
What is the actual constraint? Speed of production? Scale of personalisation? Cost of translation? Quality of analytics? Define the problem first, then evaluate whether AI solves it better than existing approaches. If the answer is “we want to use AI because everyone else is,” that is not a strategy.
Pilot before scaling
Test AI tools on low-stakes projects with clear success criteria before committing to enterprise licenses or workflow changes. Measure actual time savings, quality impact and learner outcomes, not just novelty or team enthusiasm.
Governance and ethics
Establish clear policies: what data can be input to AI tools, what content requires human review before publication, what disclosure is needed when AI-generated content is used, how bias is monitored, and who is accountable for quality. These decisions need to be made once at the organisational level, not improvised per project.
Upskilling your team
AI tools are only as effective as the people using them. Invest in prompt engineering, critical evaluation of AI output, and workflow redesign. The biggest returns come from experienced designers using AI to amplify their expertise, not from replacing junior roles with automation.
How we approach it
We use AI tools daily in our design and production workflow. We are not AI sceptics. We are AI realists. That means: using generative AI for research, drafting, ideation and iteration. Using it to accelerate, not to replace the instructional thinking that makes programs effective.
We also help clients navigate their own AI adoption: evaluating tools, establishing governance, piloting applications, and upskilling L&D teams to use AI effectively and responsibly. The goal is not to chase every new tool. It is to adopt the ones that genuinely improve learning outcomes and team capacity.
Industries & use cases
A government department exploring AI for policy-to-training translation has completely different requirements, risk profiles and governance constraints from a tech company using adaptive learning for onboarding. The technology is the same. The implementation strategy is not.
We have supported AI adoption conversations and implementations across government, health, corporate and education sectors. The approach is always the same: define the problem, evaluate the tool, pilot responsibly, and scale what works.
FAQs
Questions, answered.
The things people ask us most. Got something else? We’re happy to chat.
Ask a questionNo. AI will replace instructional designers who only do what AI can do: basic content assembly, reformatting and templated course builds. Designers who diagnose problems, design strategy, collaborate with stakeholders and apply learning science have more job security than ever. The craft is safe. The commodity work is not.
We use generative AI (ChatGPT, Claude) for research, drafting, ideation and content restructuring. We use AI-powered transcription and translation tools (ElevenLabs) for multimedia production. We evaluate and recommend platform-specific AI features (adaptive learning, analytics, chatbots) based on client needs. We do not use AI to produce final content without human review and instructional design oversight.
It depends on scale and content volume. Adaptive learning delivers the most value when: you have a large learner population with varied starting knowledge, significant content depth to draw from, and clean performance data to inform the algorithm. For small programs or niche audiences, the setup cost often outweighs the personalisation benefit.
Start small: pick one bottleneck (drafting, question writing, summarisation, translation) and pilot an AI tool against it for 4 to 6 weeks. Measure time savings and quality impact. Establish basic governance (what goes in, what gets reviewed, who is accountable). Then expand to the next use case. Do not try to transform everything at once.
The main risks: factual inaccuracy (AI confabulates), bias reproduction (AI reflects training data), quality erosion (AI output accepted without review), data privacy (sensitive content input to third-party tools), and learner trust (undisclosed AI-generated content feels deceptive). All are manageable with governance. None are reasons to avoid AI entirely.
Yes. Even a lightweight policy (what data is allowed, what requires human review, what disclosure is needed) prevents problems that are expensive to fix after the fact. Governance does not need to be complex. It needs to exist before the first tool goes into production use.
But wait, there’s more…
Careers in learning design
Guides professional growth, job pathways and skill development for aspiring and practising learning designers.
Learning campaigns
A learning campaign drives adoption, not just awareness. See how communication, change strategy and manager activation turn rollouts into behaviour change.
Learning design
Shapes content, structure and practice so people actually learn — not just sit through a slideshow.