AI to Support Neurodivergent Learners in Vocational Education and Training

2. Potential of AI and other advanced technologies to support neurodivergent VET learners

AI and other advanced technologies (defined in Box 2.1) have the potential to make VET more adaptive, accessible, and inclusive for neurodivergent learners. AI-enabled adaptivity allows instruction and feedback to be tailored to diverse needs, learning styles and abilities, helping bridge gaps in participation and performance. Extended Reality (XR) technologies enable immersive, practice‑oriented learning in environments that accommodate individual needs and facilitate repetition and rehearsal. Tools such as text-to-speech, speech-to-text and generative AI can enable learners to engage with materials in their preferred modes, enhancing the accessibility of VET. Other tools seek to directly address difficulties that neurodivergent learners may face, such as with literacy and mathematics, executive function, communication and stress management. Interviewees spoke of how AI and other advanced technologies could help build independence and confidence among neurodivergent learners, while preparing them for a workplace where AI literacy and digital skills are increasingly important. AI and advanced technologies have the potential to make VET delivery more efficient and responsive by automating administrative processes and improving management systems, which may have advantages for neurodivergent learners.

Box 2.1. What are AI and other advanced technologies?

The study was centred around a simplified definition of AI:1 Artificial intelligence – or AI in short – is what enables smart computer programmes and machines to carry out tasks that would typically require human intelligence. Other advanced technologies refer to technologies with similar features as technologies embedded with AI, as well as technologies that could be easily enhanced with AI’s capabilities in the near future.

1. The OECD AI expert group (OECD, 2024[39]) defines an AI system as: A machine‑based system that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments. Different AI systems vary in their levels of autonomy and adaptiveness after deployment.

AI can support adaptivity in VET, allowing diverse needs to be met

AI-enabled adaptivity in VET learning allows for teachers (including trainers) to cater to diverse needs, learning styles and abilities. Part of this stems from AI’s ability to quickly generate content, e.g. enabling teachers to generate multiple versions of an exercise sheet. The other part stems from AI’s ability to process data, identify patterns and generate recommendations, e.g. enabling learning platforms to tailor instruction according to the individual learner’s needs. The main advantage of adaptive learning, highlighted by interviewees as well as in previous OECD work (2023[31]; 2021[40]), is that learners get instruction and feedback more tailored to their needs across classroom, workshop and workplace environments, bridging the education and training gaps that exist, and creating more inclusive education and training systems that meet needs of learners with and without a specific diagnosis. Interviewees also mentioned time savings for teachers and better learner engagement as other advantages. AI can also be used to adapt and tailor simulations and virtual environments, as discussed in the next section.

Educators shared their own experiences using generative AI to adapt materials for different learner profiles. Veronika Kaska described how teachers in the Astangu Vocational Rehabilitation Centre in Estonia use ChatGPT to tailor exercises according to learners’ linguistic and other needs, for a pre‑vocational course aimed at non-native Estonian speakers with a disability. She saw the main advantage of this practice as time savings for teachers. Eleni Damianidou described a similar practice in the lower secondary school (general rather than vocational) where she is head teacher. Before this practice started in April 2024, she says that teachers would not have been capable of adapting exercises to this extent, so this represents a substantial advance in terms of the school’s capacity to accommodate diverse needs. However, both interviewees stressed the importance for teachers to check the quality of AI outputs, referring to the potential for misinformation discussed in Chapter 4. Eleni Damianidou estimated that AI does 70% of the work, the teacher does 10% and that she (drawing on her PhD in special education) does the final 20%. In another example, Ann Kennedy, a further education tutor, spoke of using generative AI a few weeks into a course to readjust existing course materials according to the pace of the class.

“In the case of students with ASD, we try to avoid wording that is generalised or very abstract and give more realistic wording and examples in the questions. Artificial intelligence is good at doing that” Eleni Damianidou, head teacher of a gymnasium, Cyprus

Thinking about the future, some interviewees shared their vision for how AI could revolutionise VET by providing instruction personalised to each learner. According to the CEO of Ultranauts, an engineering consultancy which provides corporate coaching guided by a Design for Neurodiversity approach, learning paths for workplace training need to be modular, flexible, and contextually relevant, tailored to individual processing styles (e.g. tactile, visual, auditory). Although not yet applied in practice, Rajesh Ananda saw clear potential in using generative AI to automatically transform content into various formats that meet individual learner needs. Thomas Köhler, Chair of Educational Technology, at the TU Dresden Institute for Vocational Education and Vocational Didactics, described various innovations that this new data dimension could bring to VET. He gave an example of a project tested five years previously among VET teacher-training students whereby the student could upload an essay and get immediate and detailed feedback from an AI agent trained on course materials, which they could then use to improve the essay. Moving beyond physical environments shared by all learners, the possibility emerges to optimise digital learning and work environments for each learner. Mobile devices in workshops or workplaces open up the possibility for interactions with virtual personas and conversational agents, providing personalised instruction for each learner based on observed behaviours. Similarly, Colm McNamee described how AI could in the future act as a virtual tutor, presenting the curriculum in the modality that best suits each learner’s learning style. With AI providing personalised instruction at scale, he imagined that the role of the teacher would transform from traditional pedagogy to facilitating learning.

Developers of existing training and education platforms shared ambitions for AI to deliver personalised instruction at scale, although many acknowledged that the full potential had not yet been realised at the time of interview. An OECD report (2023[31]) on future‑ready VET also found that personalised learning with AI was rare in surveyed VET institutions in Estonia, Norway and Scotland. Chief Operating Officer Brad Tombling spoke about how Bud Systems, a training management platform for apprenticeships and skills delivery, is currently building a product that provides AI-generated (and human-verified) qualitative feedback to users’ PDF or Word submissions. He identified timely and regular feedback on written exercises as an important factor in keeping VET learners engaged. With both a marketing and a teaching background, the Hoja AI CEO and Co-Founder, Pavan Konanur, is exploring whether the same innovations that drive hypertargeted marketing can be applied in the education sector. Currently the platform tailors content to learners’ interests and goals, learning styles and pace but he saw the potential for “hyperpersonalisation”, in which content and communications are updated in real time for each user. For vocational topics such as welding, plumbing and electricity, he imagined that personalised video content could be useful.

While few current examples emerged, interviewees also spoke of the potential to improve existing assistive technologies through adaptive approaches. Janus Asko, Co-Founder of EyeJustRead, described how their reading tool combines eye‑tracking and voice data to deliver insights for reading professionals (common to Danish schools) on challenges and behaviours, and on which interventions are most beneficial for each individual learner. The company is currently exploring how machine learning can better identify reading patterns and offer more personalised feedback. Nwanneka Udeka, a speech-language pathologist, spoke of AI’s potential to enable alternative and augmentative communication (AAC) and cognitive assistance apps to provide real-time feedback to learners, thus empowering them to define and meet their own goals, ultimately enhancing their engagement with learning. Robert McLaren, Director of Policy at Policy Connect, described AI’s potential to customise support offered by wearable devices. Currently, smartphone or smartwatch apps can offer task reminders and instructional videos for vocational tasks, but AI integration remains limited. Robert McLaren explained that developers (e.g. of the tool AssistiV) were beginning to explore how AI could customise these supports according to an individual’s usual behaviours or provide extra support for situations that the individual finds difficult.

XR and other advanced technologies can improve VET learning environments

Extended Reality (XR) and other advanced technologies can improve VET learning environments by offering innovative and practice‑oriented methods that may be particularly appreciated by learners who have struggled in traditional learning environments. Interviewees highlighted the potential of XR to facilitate immersive and situated learning in VET, creating adaptable environments that accommodate individual needs and allow learners to safely practise complex or hazardous tasks virtually before entering real workplaces.

There are natural synergies between XR and VET, according to Kevin Gonyop Kim, a professor of spatial computing and 3D Technologies whose PhD focussed on applying digital technologies in VET. Straddling digital and physical worlds, XR can provide immersive, context-aware learning environments that are particularly suited to VET (as discussed in Box 2.2), where situated and experience‑based learning prepares learners for the workplace. XR is an umbrella term that encompasses technologies blending physical and virtual environments, including Virtual Reality (VR), Augmented Reality (AR), and Mixed Reality (MR). XR can include both non-immersive (where one is aware of their physical surroundings, e.g. watching a display and using a controller) and immersive technologies (e.g. wearing a headset and providing input through gloves or handheld devices).

Box 2.2. XR can provide immersive, context-aware learning environments that are particularly suited to VET

Kevin Gonyop Kim described how XR works especially well in vocational fields involving physical spaces or objects (e.g. gardening, floristry, fashion design) where digital representations can be easily created and reused as learning materials. He provided the example of apprentice gardeners capturing real gardens through photos or drone imagery, reconstructing them in 3D using machine learning, and using the virtual garden in classroom settings to bridge theoretical and practical learning. With a virtual garden, one can simulate the changing of seasons without having to wait until the next semester. When used in carpentry, XR can help apprentices “visualise the invisible” by overlaying force propagation, explained Pierre Dillenbourg. While some interviewees saw cost of XR as a barrier, there are some situations where it can be cost efficient. Veronika Kaska spoke of one vocational school in Estonia that uses VR to simulate tractor driving and operation because tractor repairs were becoming too expensive. The XR Action Plan, an initiative by the Flemish Department of Education and Training, allowed VET institutions to borrow VR equipment (headsets, iPads, 360° cameras) and software free of charge, according to Kevin Polley of Knowledge Centre Digisprong. Available software covered welding and carpentry, for example, and teacher training was also provided. The XR Action Plan was being evaluated at the time of interview for possible continuation.

Interviewees identified aspects of a virtual environment that would be particularly appreciated by neurodivergent learners, particularly those with ASD and ADHD. The first is that a virtual environment can be adapted to accommodate individual needs in ways that real environments cannot. Kevin Gonyop Kim explained that researchers are currently exploring how they can simplify virtual environments “by blacking out the rest of the world” for learners who have difficulty focussing (e.g. learners with ADHD) and then gradually increase the complexity of the environment as learners progress. The idea of filtering out distracting sounds or unnecessary information by using XR was described by Geena Vabulas (The Oaks Specialist College) as very promising and yet underexplored, with some projects in research phases but not yet in widespread use. Kevin Polley (Knowledge Centre Digisprong) pointed to research (Smet, 2023[41]) showing the possibility of using VR glasses when a learner with ASD is experiencing overstimulation, allowing them to retreat to a virtual space where they can relax without leaving the classroom. Jan Schlueter described how driving simulator SANDI adjusts driving scenarios progressively based on driving performance but also based on eye gaze and physiological data (heart rate and sweat responses) which capture the physical and emotional state of the learner. The driving simulator, often used in vocational rehabilitation, is designed to build confidence and support the path to independence, particularly for those experiencing high anxiety or other cognitive barriers.

“It’s all about giving the individual who is sitting in that driving chair a pathway to independence, meaning it's all about getting a driver's licence” – Jan Schlueter, CEO of Neurodiverse Technologies, developers of SANDI

The second advantage is that a virtual environment can allow neurodivergent learners to repeat and rehearse tasks in a controlled space that provides psychological and physical safety. In Finland’s largest special-needs VET college, Educational Digital Specialist Piia Jokelainen described how 360° workplace scenes built with Thinglink, AR smart glasses and simulators are used so that learners can pre‑visit sites, rehearse task sequences and build a sense of safety before moving to real settings. Access to XR equipment is organised through a central library that teachers can borrow from. For Kevin Polley, the ability to repeat tasks was a strength of the welding and carpentry tools made available through the XR Action Plan (discussed in Box 2.2). Other interviewees highlighted the benefits of repetition and rehearsal for building confidence and managing anxiety and/or overstimulation. This can be useful for developing everyday skills that will help the individual navigate VET and the transition to work. Veronika Kaska was aware of a hospital using VR to simulate the experience of grocery shopping for patients recovering from a stroke. She and her colleagues thought that similar technology could be extremely useful for building independence among young learners with ASD in their pre‑vocational courses, some of whom struggle with anxiety and are socially or verbally restricted. Visiting the store to practise grocery shopping is something that these learners currently do with a teacher. She explained that VR could allow students with very high anxiety to practise this and other everyday skills as many times as needed to desensitise before moving to real-life practice.

Similar tools can be used to practise interview and other workplace interaction skills in a low-pressure environment:

• The VR soft skill training tool, Bodyswaps, was mentioned by two interviewees, Helen Nicholson-Benn (Assistive Technology/AI specialist, Jisc) and Deborah Millar (Executive Director of Digital Transformation, Hull College). Bodyswaps was designed to promote employability skills by simulating workplace interactions such as an interview or interactions with customers. The learner experiences the simulation through a VR headset, playing their assigned role while their body movements and voice are recorded. The learner then swaps roles (e.g. with the interviewer or customer), giving them the opportunity to see their recording along with suggestions for improvement. Helen Nicholson-Benn noted particular interest in the tool among VET institutions while her organisation Jisc evaluated the tool positively in a pilot report (2022[42]). In her college, Deborah Millar said that this tool was mainly used for learners who were non-native English speakers but that she planned to make it available also to learners with special educational needs.

• Auticon (an IT consultancy with a majority-autistic workforce) is developing, with the University of Stuttgart and Fraunhofer Institute, a VR immersive environment tool that will help neurodivergent individuals practise challenging scenarios such as job interviews with real-time emotion feedback. Ursula Schemm, who works in corporate communications at the consultancy, described this area of research as new and growing but said that it was difficult to find market-ready solutions to meet the diverse needs of the Auticon workforce.

• Jan Schlueter mentioned an interview simulation tool currently being developed by Vanderbilt University to help neurodivergent individuals manage anxieties associated with this socially complex event while also training the interviewer to engage successfully with a diverse range of interviewees. The tool will simulate the full interview experience from sitting in the waiting room to being interviewed to dealing with unexpected situations, tracking eye gaze and physiological data.

AI and other advanced technologies can enhance the accessibility of VET learning materials and instruction

Tools like text-to-speech and speech-to-text enable learners to engage with VET learning materials and instruction in their preferred format, while generative AI allows learners to optimise materials for accessibility. Text-to-speech and speech-to-text tools were seen by many interviewees as foundational for learning as they enable each learner to engage with the curriculum in ways that align with their strengths and sidestep difficulties with, for example, reading, writing or auditory processing. Moreover, by enabling participation on more equal terms and by acknowledging differences in the way that students learn, these tools have the potential to make VET more inclusive. Similarly, an OECD paper on “The Potential Impact Of Artificial Intelligence On Equity And Inclusion In Education (Varsik and Vosberg, 2024[43]) describes the potential for AI to allow students with certain special education needs to participate alongside their peers to a greater extent, contributing to a more diverse and inclusive learning community and enriching the educational experience for all students by fostering an environment of diversity and mutual understanding.

Text-to-speech tools convert written text into spoken words, allowing learners to process information through auditory channels rather than relying solely on reading. This can help learners who struggle with reading fluency, focus or decoding written language. Christopher Patnoe (Head of EMEA Accessibility and Disability Innovation, Google) described these tools as “profoundly impactful” as having text read aloud helps learners with dyslexia to internalise and understand learning materials better than reading could. Clayton Lewis (Professor Emeritus of Computer Science, University of Colorado) spoke of students with dyslexia who had struggled enormously with secondary school because of reading difficulties but later thrived in postsecondary education (and beyond in some cases) after being introduced to text-to-speech tools.

Speech-to-text tools (also known as dictation tools) convert spoken language into written text. In addition to the use of speech-to-text tools for writing (discussed in the next section), interviewees were enthusiastic about their use for live‑captioning classes and meetings. Live captioning provides visual reinforcement of spoken information, helping learners who struggle to follow rapid or complex speech. Reading and listening simultaneously can help learners to absorb information, while removing the pressure to take extensive notes and making it easier to look up complex language or terminology in the moment. Note‑taking tools such as Plaud, Otter.ai and Limitless combine live captioning with LLM (Large Language Model) features that can summarise notes and generate actions and insights.

Some interviewees remarked that tools that end up being useful for neurodivergent learners are often originally designed to serve other groups. Kellie Mote (Programme Lead (Accessibility) at Jisc) and Ann Kennedy (further education tutor) both noted that live captioning was often introduced to accommodate deaf learners, but there was a growing awareness of the benefits for learners with ADHD, ASD and learning disabilities. As a further education tutor, Ann Kennedy applies live captioning to all class recordings as she has noticed that many learners appreciate this feature. Kellie Mote said that it was often the case that innovative technologies are targeted first at learners with lower-incidence but higher-support needs (e.g. deaf learners, learners with vision impairment or mobility difficulties) and later trickle down to neurodivergent learners. A feature in Google ChromeOS’s built-in reading tool, which enables one to highlight a word and have it read aloud, was aimed initially at low-vision users, according to Christopher Patnoe. Later the team realised that the feature was useful for individuals with dyslexia and for individuals learning a second language, and once combined with a distraction-dimming feature, for those with ADHD. Deborah Millar described how many live captioning and text-to-speech features used by neurodivergent learners in Hull College originated as tools to support non-native language speakers. These examples illustrate how thoughtfully designed features can benefit multiple groups as well as learners with co‑occurring neurodivergences.

Although text-to-speech and speech-to-text tools are not new, these tools have become more powerful in recent years due to improvements in LLMs and in cloud computing. Marius Frank of Microlink described advances in cloud computing as one of the most profound changes in the past 10 years, since it allows information to be processed at much faster speed than a laptop or mobile device can support. This is what enables tools such as Microsoft Immersive Reader to provide text-to-speech at speed, in different languages and across different Microsoft software. Recent technological advances in LLMs have improved the accuracy, formatting and contextual awareness of speech-to-text tools and the pronunciation, fluidity and intonation of text-to-speech tools, according to interviewees.

“If we can start embedding some of these basic universal tools and the use of it to unlock learning for individuals at an early stage, it can be truly transformative for life outcomes” – Marius Frank, Education Director, Microlink PC

LLMs are also the technology driving generative AI chatbots such as ChatGPT and Copilot, used by some teachers to adapt materials for different learner profiles, as discussed in the previous section. The same tools are used by many neurodivergent VET learners to engage with and to improve the accessibility of learning materials. Interviewees described how these tools could be used to simplify or summarise learning materials and to assist with interpretation. Christopher Patnoe described how some learners with ASD use generative AI chatbots to, for instance, help explain metaphors, jokes and cultural references. Learners can use the same tools to engage much more closely with the learning materials, according to Lorenzo Desideri (Professor in Psychology and Artificial Intelligence, Sigmund Freud University Milano). He said that many learners with dyslexia are already using ChatGPT to generate custom quiz questions to test comprehension. By framing queries like “Create three multiple-choice questions about this paragraph”, students can engage with the material actively, reinforcing their understanding of syntax and content. A number of interviewees mentioned that learners could use Google’s Notebook LM to generate podcasts based on inputted materials, providing a new and enjoyable way to learn. David Voss, who works on digital learning in higher education, explained that this tool could be especially powerful for learners with dyslexia who struggle with the speed of reading as it converts materials into a format that is much easier for them to receive. Concerns about the risks associated with generative AI, such as misinformation and overreliance hindering skill development, are discussed in Chapter 4.

Some tools directly address the difficulties that neurodivergent learners may face in VET

Some tools are designed to directly address the difficulties that learners might experience in VET (and in general education) because of their neurodivergences, such as difficulties with reading, writing and mathematics, executive function, with social interaction and communication and with managing stress. These tools help to develop skills in these areas or provide a work-around, whereas the previous section focussed on tools that enhance the accessibility of learning materials and instruction.1 This section highlights the tools mentioned by interviewees while the repository published alongside this report provides a longer list of available tools.

Tools to assist with reading

Advances in text-to-speech and in speech recognition (similar to speech-to-text) could also drive enhancements in tools aimed to develop reading skills among learners with reading difficulties. Read&Write is an example of a tool specifically designed for students with literacy challenges, which combines a text-to-speech feature with other features such as screen tinting, which can be helpful for learners with dyslexia. Speech recognition tools can be used to practise reading out loud, delivering instant pronunciation feedback, and reducing reliance on a human tutor. As students improve their ability to pronounce words correctly, they become better at recognising written words when reading. Thus, simple AI-driven speech tools can significantly accelerate reading proficiency and lessen the burden on specialised instructors, according to Lorenzo Desideri. EyeJustRead is a reading tool that combines eye‑tracking, speech recognition, and metadata from e‑books to provide personalised insights into reading strategies, challenges, and progress, especially for learners with reading difficulties.

Tools to assist with writing

Interviewees mentioned writing assistance tools specially aimed at learners who experience difficulties in this area. By converting spoken language into written text, speech-to-text tools such as Dragon Speech Recognition allow learners to express ideas verbally rather than through typing or handwriting. This can enable neurodivergent learners to express themselves and demonstrate their knowledge more fluidly and to sidestep any issues with spelling. Other tools are specifically designed to address spelling accuracy, grammar and punctuation accuracy, such as Grammarly, which has recently integrated generative AI capabilities (as discussed in Chapter 4).

Many interviewees drew attention to the potential for generative AI chatbots to be used as a tool to assist with writing, given their capacity to refine written text. Learners who struggle with spelling, punctuation or grammar can use the chatbot (where spelling, punctuation or grammar are not the skills being assessed) to detect and correct any errors so that an assignment can be assessed on the basis of content alone. David Voss spoke about the potential to use generative AI chatbots to overcome the “blank page syndrome” which can be an issue for learners with ADHD, by providing a template and some suggested steps to follow. Many interviewees also highlighted the use of generative AI chatbots to improve tone. Deborah Millar described a framework she developed, called PASTA AI, to guide learners with special education needs in Hull College to use generative AI chatbots in a thoughtful way. Learners use the framework to specify the desired Profession, Audience, Style, Tone and Action when writing a prompt. The framework improves the quality of the output and gives learners a different way to engage with the writing process. Deborah Millar felt that, with the right frameworks in place, generative AI chatbots could help empower neurodivergent learners and help them develop literacy skills.

“Being able to generate ideas with something to start, say, an essay assignment or written assignment is something that will be helpful, I think, particularly with conditions like ADHD. It's “where do I start?”, “how do I get started?” and then once you do get started, you can fly” – David Voss, who works on digital learning in higher education

Tools to assist with mathematics

Interviewees highlighted a small number of tools aimed to address difficulties with mathematics or in writing mathematic notation. The Luovi Vocational College features drawing screens that can make mathematics learning more visual and active floors that can make it more tactile and engaging, according to Piia Jokelainen. Clayton Lewis spoke of his experience developing software to help with algebraic notation by taking pen-and-tablet handwritten input, fixing errors and producing output that could be assessed by teachers. Not only did the software help teachers assess learners’ work, but it also helped learners reflect on their own learning processes and identify strategies that would be useful for them. While the software never developed beyond the testing phase, he saw this as an idea that could be revisited given recent advances in AI. In his view, decoding partial information is a particular strength of AI. This is what enables text-to-speech tools to operate smoothly without stopping to correct spelling mistakes, for example. In a similar way, AI-enabled software could make algebraic notation far more natural.

Tools to assist with executive function

Planning, time management, working memory and attention (all elements of executive function) can be challenges for some neurodivergent learners in VET, which make it harder to organise, prioritise, and complete tasks in learning and workplace settings. Interviewees spoke of the potential for AI and other advanced technologies to help learners address these challenges, which can particularly affect learners with ADHD and ASD.

Tools such as Goblin Tools and Tiimo can help support planning and time management. Goblin Tools enables users to create to-do lists and to estimate how much time each task will take. VET learners can use these tools to break down assignments into manageable chunks and to signal “what’s coming next”, according to Nwanneka Udeka. Tools such as this can enable learners to complete tasks more independently and avoid frustration stemming from missed deadlines or unclear expectations. Geena Vabulas identified accessibility as one of the key strengths of Goblin Tools; it requires no login, is free, has a simple interface and accepts speech-to-text input. She noted that some upper secondary schools were already using Goblin tools, helping to prepare learners for the transition to work and introducing them to a tool that can continue to support them once they are in work.

Many learners who experience difficulties with working memory use alerts and notifications in their daily life. Francesc Sistach made the point that even very simple technologies – like alarms on a smartwatch or calendar apps – can dramatically help neurodivergent learners organise their time. In his view, small, readily available technological aids often yield outsized benefits compared to more complex systems. Other interviewees highlighted ways that AI could enhance the same tools. For example, David Banes (David Banes Access and Inclusion Services) spoke about how the same systems could use AI adapt notifications and alerts by recognising patterns in the users’ behaviours or regular schedule and by voice assistants (e.g. Amazon Alexa, Siri) that can report back verbally. Alisdair Gurling suggested that wearables such as the Ray-Ban Meta glasses, used by blind and low-vision users, could be useful for neurodivergent learners, providing embedded reminders and an ongoing connection between digital and physical life.

Some tools provide video and written reminders, which learners can access via QR codes or NFC tags placed in strategic locations, removing the need to rely on memory alone. Step-by-step instructions enable learners with memory difficulties to carry out sequential tasks, which is often required in vocational and protected work settings, according to Lorenzo Desideri. By following the instructions, learners can operate with greater autonomy and lower cognitive load. He described strategies such as this as particularly useful for learners with ASD and with learning disabilities. One example of such a tool is AssistiV, which Geena Vabulas described as highly supportive, although she noted that it was best suited for users receiving government grants since it requires significant initial setup. Robert McLaren saw the potential to enhance these tools further by embedding the latest advances in AI and augmented reality so that support could be personalised to the user and to the environment (addressing a barrier discussed further in Chapter 3).

Mind-mapping software can help learners to transform existing learning content into a visual graphic, helping them to engage with the content and to organise their thoughts. David Banes described these tools as particularly useful for learners with dyslexia. While these tools are not new, some mind-mapping tools have recently integrated generative AI features. Helen Nicholson-Benn gave two such examples: Ayoa allows the user to generate images and mind maps based on a free text prompt, while MindView uses generative AI to simplify and transform inputs. Marius Frank saw mind-mapping software as useful for all learners regardless of neurotype. He made the point that making tools such as this available to all learners improves the inclusiveness of the educational environment overall, as all learners benefit while those with neurodivergences can get the support that they need.

“If we open up minds to the possibility of using mind mapping tools and organisational tools to every child, then what we are doing is equipping young people to reach for the tools that they need at the time that they need it most” – Marius Frank, Education Director, Microlink PC

Alisdair Gurling suggested that learners who struggle with focus could benefit from apps that generate soundscape background music. One example is the Endel app, which uses AI to generate music taking into account the weather, the time of day and the heartrate of the user (using a smartwatch). The music adjusts to the heartrate and then reduces the beats per minutes in an effort to promote calm and focus.

Tools to assist with communication and social interaction

Interviewees identified some tools aimed at addressing difficulties with social interaction and communication, ranging from alternative and augmentative communication (AAC) apps to social robots.

Nwanneka Udeka highlighted the important role of AAC apps, such as Proloquo2Go and TouchChat, in giving learners who struggle with verbal communication a way to express themselves. Users preload relevant vocabulary onto the app on their phone or tablet, such as words for requesting a break, answering a question about a lesson topic, or greeting peers. These apps thus enable learners to participate alongside neurotypical classmates, despite verbal limitations, ensuring they do not miss critical instruction.

Interviewees described the potential for social robots to support the development of social, emotional and language skills, particularly for learners with ASD. While social robots have been traditionally aimed at children, interviewees highlighted their potential to be used in VET and among users of all ages. These robots interact with learners through speech, facial expressions and gestures, which many learners find engaging and comfortable. Another advantage is that they allow the user to practise interactions repeatedly, an advantage also associated with the XR tools described in the previous section, designed to simulate interview and other stressful scenarios. By encouraging learners with ASD to engage within the educational setting, social robots can also facilitate inclusion and encourage interaction with peers, according to Lorenzo Desideri. As a side effect, he noted also that social robots often motivate teachers to innovate and to promote more inclusive learning.

• LuxAI produces social robots, such as their QTrobot, to help individuals with ASD, among others, develop social, emotional and language skills in schools, therapy centres and homes. Chief Operating Officer Aida Nazari described how the robot uses a camera to monitor gestures, facial expressions and behaviour and uses AI to personalise interactions. The degree of freedom of the AI can vary by use case, from facilitating open-ended conversations where appropriate, to delivering pre‑programmed activities where a specific evidence‑based approach is required. In vocational education, she described how QTrobot could be used to practise an interaction with a receptionist, a doctor or a call centre worker, for example.

• Another example is Felix, a social robot designed to support people, especially those with psychosocial or communication difficulties, in expressing and tracking their emotions. Felix is connected to an online platform that logs emotional input and helps identify emotional patterns over time, facilitating discussions with coaches or teachers. Rob van de Ven, whose company Happybots produces Felix, mentioned that they were exploring using AI to enhance the analysis of emotional data. For example, AI could help detect trends such as a 10% decrease in emotional outbursts over the past year, patterns that might otherwise go unnoticed. In educational contexts, Rob van de Ven described how Felix could be used to help learners to express how they feel in a simple, non-verbal way, which helps prevent emotional build-up and behavioural incidents. Rob van de Ven also saw potential to use Felix in internships or work-based learning settings.

Tools to assist with managing stress

Stress can be a response to having to navigate education systems and workplaces that are not designed for neurodivergence. Interviewees described how digital support systems could help neurodivergent learners in VET to navigate potentially stressful situations that can arise in VET and in the transition to work, and the potential to develop these tools further. Geena Vabulas described a situation that could be overwhelming for a neurodivergent learner: the bus they usually take to go to a work placement does not arrive. Stress and overwhelm could prevent the individual from finding a solution in that moment. Digital support systems such as Brain In Hand are designed to help the individual access pre‑defined strategies (e.g. text work, call home, wait for the next bus) in such moments. Geena Vabulas spoke of differing perspectives she had encountered in her research: some neurodivergent learners wanted digital support systems to be developed further so that they could operate much more autonomously (e.g. calling a taxi on your behalf) whereas others preferred to retain control. David Banes suggested that in future, digital support systems could be linked to wearables that detect behavioural or physiological indicators of stress, to enable earlier intervention.

AI and advanced technologies can improve VET systems so that they can better accommodate diverse needs

An OECD report (2023[31]) on future‑ready VET describes the potential for digital technologies to better manage and communicate in VET by automating administrative tasks, improving the exchange of information, and by tracking learning activities and progress across different learning environments. Some interviewees made the point that these efficiencies and improvements to VET systems could also benefit neurodivergent learners. They gave examples whereby AI and advanced technologies could: alleviate time and staffing pressures and thereby enable teachers to spend more time with learners who need more support; improve training providers’ management systems so that additional learning needs are documented and accommodated; help neurodivergent learners to navigate the VET system; and mitigate early school-leaving.

If AI and advanced technologies can remove time and staffing pressures in VET, teachers may be able to spend more time with learners who need more support and explore technologies that could make the learning environment more inclusive. A handful of interviewees spoke of the time savings for teachers using AI for lesson planning, creating lesson resources and for administrative tasks. Kellie Mote described how many teachers in higher and further education are currently using TeacherMatic, an interface that helps them build prompts for generative AI. A Jisc pilot (2024[44]) estimated time savings of two hours ten minutes per week for the average teacher. In the context of staff shortages in rural German schools, which are so severe that some education authorities have already reduced the school week to four days a week, Thomas Köhler said that some had suggested that AI assistants could fill critical gaps. He was careful to specify that this was proposed only as an emergency measure, rather than as the future of teaching.

AI and advanced technologies could improve training providers’ management systems so that additional learning needs are documented and accommodated. Bud Systems is a training management platform built originally for apprenticeships but now serving broader vocational education and skills delivery. The platform is sold to training providers as an integrated solution for programme design, enrolment, back-office administration as well as training delivery. While the system works well without AI, according to Chief Operating Officer Brad Tombling, in recent months, they have found a few use cases where AI can offer improvements: a chatbot work assistant for training providers (which they will soon extend to learners) and AI-generated qualitative feedback to learners’ written submissions (discussed earlier in this chapter). A partnership with Cognassist (a neurodiversity assessment tool) enables training providers to document learners’ additional support requirements and needs and to adapt training delivery accordingly. This means that when a teacher sets an exercise for a learner through the platform, they will see that the learner should be given additional time to complete the exercise, for example, or that the learner requires the exercise to be sent in a particular format.

“We're not looking to leverage AI and then find something that it solves. We're trying to identify a specific problem and then evaluate whether AI plays a role in the solution to it” – Brad Tombling, Chief Operating Officer, Bud Systems

Two interviewees gave examples of AI that could help retain neurodivergent VET learners, one by helping them navigate VET environments and the other to address an important risk factor for early school-leaving. One project explored by Thomas Köhler (TU Dresden Institute for Vocational Education and Vocational Didactics) together with Austrian and Norwegian colleagues, was to develop an organisational assistant for VET learners, i.e. a chatbot that they ask questions such as “where I can find my learning resources?”, “do we have an exam next Monday?”, and “who can help me with this subject?”. He described the project as working well in virtual learning environments but needing more work to adapt it for physical learning environments. Motti Sigel (Managing Director, MassChallenge Israel) spoke of the existence of AI-enabled anti-bullying tools. One startup supported by MassChallenge aimed to identify isolated students in schools in order to target interventions to prevent bullying. Such tools could help educational institutions, including VET institutions, safeguard and retain neurodivergent learners.

AI and other advanced technologies can support the transition from VET to employment

The use of AI and other advanced technologies can help neurodivergent learners build employability skills within VET and can also directly support the transition to work by supporting them in the job application process and in the job itself. Interviewees highlighted two important features of generative AI: that its widespread use helps remove some of the stigma associated with use of assistive technologies and that individuals may appreciate being able to continue using the same tools as they transition from VET to employment.

A few interviewees made the point that because neurodivergent learners can be very talented at using new technologies, greater use of technology in VET would enable them to play to their strengths and to build independence, confidence and other employability skills that would help in the transition to employment. Ann Kennedy, a tutor, hoped her further education students would ask generative AI chatbots questions they would not feel comfortable asking in class, enabling more independent learning. In the (general rather than vocational) lower secondary school where Eleni Damianidou is head teacher, teachers set aside time during special classes (with learners with ASD) for learners to investigate the topic they are being taught using a generative AI chatbot. Classes are small, allowing teachers to supervise the process. In Eleni Damianidou’s view, this exercise helps to build the self-confidence necessary to later seek good job opportunities where their talents will be valued, something that could also be beneficial for VET learners. David Banes suggested that learners who struggle with social interactions could use generative AI chatbots to script small talk to help them navigate professional environment. Thomas Köhler spoke of a woman he had interviewed years before who was too shy to communicate with colleagues in person. She started to communicate with people over the early internet and, with time, used this practice to adapt her behaviour in face‑to-face interactions. He saw the potential of AI to be used in similar ways.

“If I could reassure the students that ‘this is your safe space, this is where you can just relax and you get to learn stuff – you don't have to learn stuff, but you get to learn it if you want it’, that's really the core” – Ann Kennedy, further education tutor

Thorkil Sonne, Founder of the Square Foundation spoke of a programme using Lego Mindstorm robot kits to bridge the transition to work by creating an environment where people with ASD can show their strengths. Participants in the programme are brought to a physical office of a company that might later offer them placements. Participants start building a programmable robot individually (first week), then as a group (second week), learning how to use the Scrum project management framework to organise resources, brainstorm and choose which ideas to pursue. Participants are comfortable in the office environment by the time they meet and present their project to prospective managers and coworkers (third week) and before they are invited into a real work setting where they can understand the tasks they would carry out if offered a job (fourth week).

AI and other advanced technologies could help neurodivergent learners in the process of applying to jobs, interviewing and onboarding, according to interviewees. The potential to use XR-enabled tools for interview practice is discussed earlier in this chapter. Interviewees also highlighted the potential to use generative AI chatbots and writing tools such as Goblin Tools to refine CVs and cover letters. Freya Bevan (Digital Learning Coach, Gloucestershire College) explained that Goblin Tools has a feature called Formalizer, which changes the tone of language, e.g. enabling an apprentice to make a cover letter formal, friendly or sophisticated. Although previous OECD research speaks of the potential for AI to improve labour market matching, including increasing the diversity of applicants (Broecke, 2023[45]) – and to improve matching between learners, institutions and employers in VET itself (2023[31]) – interviewees for this project were more concerned about AI leading to bias and discrimination when used in recruitment tools (as discussed in Chapter 4). Only one interviewee gave an example of the potential of AI to match individuals to opportunities: Auticon, an IT consultancy with a majority-autistic workforce, is currently developing an AI-enabled solution to help match their consultants to prospective projects based on skills and experience, according to Ursula Schemm. They also use AI to improve employee onboarding.

Once in the workplace, widespread use of AI at work could not only help neurodivergent employees but could also remove some of the stigma associated with use of assistive technologies. Interviewees spoke of the wide variety of generative AI tools available to edit emails and other documents, to transcribe and summarise meetings, and to perform many other daily tasks in a desk-based job. Nathaniel Cook (Chief of Information and Technology, Special Olympics) described how these tools level the playing field, effectively “democratising ability” by removing some of the structural barriers to participation and self-expression. When his organisation, the Special Olympics, rolled out Copilot, they found that employees with disability gained twice as much in productivity (12 hours per week saved) compared to their colleagues without disability. Many interviewees highlighted that generative AI tools are accessible to all employees, whereas previously available writing aids and dictation tools might have been limited to those who could prove their need and/or disability status. Not only does this remove a practical barrier for neurodivergent employees, but it removes some of the stigma that surrounded assistive technologies. With the recent enthusiasm around generative AI, people are even excited to share how they use these tools in innovative and playful ways, according to Alisdair Gurling.

“They've reframed it from being something that helps them mitigate a deficit to some strength-based thing that's really exciting” – Alisdair Gurling, researcher at Wonderlab, Monash University

Key to ensuring that the potential of these tools is met, is ensuring that learners can continue using the same tools as they transition from VET to employment (which involves overcoming the barrier of insufficient interoperability, described in Chapter 3). In an ideal scenario, according to Thomas Köhler, AI could assist neurodivergent VET learners in the form of an individualised companion integrated on a mobile device suitable for classroom, workshop and workplace settings, thus providing continuous support as the individual transitions from learning to employment and even across their entire life. Many interviewees saw continuity of use as a particular strength of generative AI chatbots, as well as other tools that support the accessibility of materials, such as Goblin Tools, text-to-speech, speech-to-text and Microsoft Immersive Reader.

“It may accompany the learners to their workshop or even future workplaces. [...] Learning would not need to stop at the end of the school” – Thomas Köhler, Chair of Educational Technology, TU Dresden Institute for Vocational Education and Vocational Didactics; Director, TU Dresden for Open Digital Innovation and Participation (CODIP)

Note

← 1.  While there can be some overlap between the two, the distinction between “disability-centred” solutions and “environment-adaptation” solutions was an important component of previous OECD research (Touzet, 2023[46]) on the use of AI to support people with disability in the labour market.