By Benoît Marchais – Functional Manager, Enovation
AI is no longer a gadget in LMS, but a driver of profound transformation: express content generation, dynamic adaptation, predictive analysis, 24/7 assistants… Open source platforms play a key role as innovation laboratories where productivity, personalisation and digital sovereignty come into play. This transformation is already underway, as evidenced by the growing use of AI-powered plugins and authoring tools. Full disclosure: as I begin writing this article, I would be lying if I said that the writing process wasn’t made easier by a (reasonable) call for help from a generative AI…
Here is an overview of the main use cases where AI is transforming online training.
1. Automatic generation of educational content
Use case: accelerating resource production
One of the major challenges of online training lies in content production: designing quizzes, writing case studies, structuring educational pathways. These time-consuming and repetitive tasks can take several hours or even days of work for an educational engineer.
Generative AI provides a solution by enabling structured content to be created in a matter of minutes from a simple text description or source documents (PDFs, presentations, videos). The trainer describes the educational objectives and provides the basic resources, and the AI generates an initial version that is almost ready to use: quizzes, interactive activities, course materials, assessments.
This automation does not eliminate the role of the educational engineer, but rather reposition it towards tasks with higher added value: validating educational consistency, customising content for the target audience, and creatively enriching content.
Available solutions
These solutions come in two distinct forms: native plugins, which integrate directly into the LMS and allow content to be generated without leaving the platform, and external authoring tools, which require working outside the platform before importing the modules created in SCORM format or via integration protocols such as LTI (Learning Tool Interoperability).
Integrated solutions in Moodle
Additional plugins include Dixeo (developed by Edunao) and Course AI (offered by eLearning Touch’). These solutions allow structured courses to be generated directly in Moodle from a simple text description or source files, automatically creating a variety of native activities (quizzes, forums, pages, H5P resources, etc.).
External authoring tools
Several SaaS platforms are positioned in this segment, offering an AI-assisted instructional design approach with export to LMS:
- Autrice allows you to design complete training modules with instructional architecture and assessment activities, exportable in SCORM format for integration into Moodle or any other LMS.
- Nolej, a French start-up that won an award at BETT 2024, transforms existing content (PDFs, videos, documents) into a variety of interactive activities (flashcards, quizzes, interactive videos, crossword puzzles) that can be exported in SCORM or H5P format.
- Edtake stands out for its focus on data sovereignty: GDPR-compliant with hosting in France, the tool analyses source documents (PDFs, PowerPoints, videos, URLs) to automatically generate training content and educational activities, exportable in multiple formats (H5P, SCORM, Word, HTML, Moodle).
- Caramel (Moodle and ÉLéa Resource and Activity Creation), developed by Emmanuel Gaunard (Grenoble Academy) and Gauthier Remande (Nantes Academy), stands out from other solutions due to its status as an open source tool. Completely free and with no registration required, it allows users to generate H5P activities and courses for ÉLÉA or Magistère from provided documents. The tool relies on sovereign AI hosted by GDPR-compliant universities (Mistral Small, LLaMa or GPT-OSS, depending on requirements).
Limitations to consider
While generative AI does indeed make it possible to create a robust framework without investing time in initial technical configuration – an undeniable advantage for quickly getting a training project off the ground – this apparent ease comes with several major pitfalls.
A worrying technical dependency
The first risk concerns the competence of the educational engineer: accustomed to relying on AI to build their courses, they gradually risk losing technical mastery of the design process. This dependency can become problematic when the tool is no longer available or when fine-tuning is required. Professional autonomy is thus compromised in favour of automation, which, while speeding up the process, can also diminish professional expertise.
Sensitive data in transit abroad
Beyond this question of jurisdiction, there is a crucial issue of data sovereignty. Most current generative AI solutions rely on servers located outside Europe, particularly in the United States or Asia. However, educational content passes through these infrastructures during its generation, potentially exposing sensitive information: learner data, specific organisational know-how, proprietary methodologies or strategic content. This circulation of data raises questions about GDPR compliance, but also about intellectual property protection and the confidentiality of the educational practices developed by the institution.
An unavoidable verification time
Finally, the promise of creation ‘in just a few clicks’ obscures an unavoidable reality: generated content always requires thorough proofreading and educational validation. Factual errors, conceptual approximations, clumsy wording or wording unsuitable for the target audience: imperfections are frequent and sometimes insidious. The time theoretically saved in generation must therefore be reinvested in meticulous verification, correction and adjustment work – a process that can be time-consuming and partially negates the initial expected efficiency gains. Without this critical step, there is a risk of disseminating content of uncertain quality, compromising the credibility of the training.
2. Automatic adjustment of difficulty levels
Use case: individualising assessment pathways
In traditional training, all learners receive the same assessment questions, regardless of their level of proficiency. This uniform approach poses two problems: it can discourage learners who struggle with overly complex questions, and it underutilises the potential of advanced learners with exercises that are too simple.
AI solves this educational equation by dynamically adapting the difficulty of the questions. The system analyses responses in real time: if a learner easily answers several questions correctly, the algorithm gradually increases the complexity to maintain an optimal level of challenge. Conversely, if a learner repeatedly struggles, the system offers more accessible questions and automatically generates reinforcement exercises targeting the concepts they find difficult.
This automatic adaptation transforms assessment into a genuine learning tool: each learner progresses at their own pace, within their zone of proximal development, thereby maximising educational effectiveness.
Example of a player available
Stellia.ai, a French start-up founded in 2019, is developing an adaptive learning solution that personalises learning paths according to each learner’s level of knowledge. The system assesses skill mastery through a database of exercises and dashboards, identifying learners who are struggling and offering content tailored to their level. According to the publisher’s data, quiz results are 20% higher and learning time is reduced by 30%.
3. Personalising learning paths: adaptive learning in action
Use case: recommending the right training courses to the right people
With such a wide range of training courses available, learners and training managers are often faced with information overload: which training course should they choose first? Which learning path is most relevant given the learner’s profile, position and professional goals?
AI transforms the LMS from a simple static catalogue into an intelligent recommendation system. By analysing learning histories, educational preferences, assessment results, job positions and the organisation’s strategic skills, algorithms automatically suggest the most relevant modules for each employee.
This personalisation goes beyond simple recommendations: it can prioritise training courses according to urgency (certifications to be renewed, new regulations), identify missing prerequisites and suggest refresher courses, or even recommend additional training courses to develop cross-functional expertise.
Use case: adapting the pace of learning in real time
Even more innovative, AI can dynamically adjust the pace of progress for each learner. The system detects signs of disengagement (decreased connection time, activity dropout rates, declining quiz results, long periods of inactivity) and reacts proactively.
Rather than waiting for a learner to drop out permanently, AI can:
- Automatically suggest educational breaks to avoid cognitive overload.
- Suggest targeted revisions of poorly understood concepts.
- Offer alternative formats (short videos instead of long texts, infographics instead of tables).
- Temporarily lighten the schedule for learners who are struggling
- Recommend group or individual tutoring sessions
This real-time adaptation maintains engagement and significantly reduces dropout rates.
4. Predictive analytics: anticipating to improve training
Use case: identifying learners at risk of dropping out
Dropping out of distance learning is an insidious phenomenon: by the time a tutor notices that a learner is no longer active, it is often too late to reverse the trend. Predictive analytics makes it possible to anticipate this risk several weeks before the learner drops out for good.
By analysing dozens of indicators (connection frequency, forum participation, interim results, time spent on resources, regularity of engagement, activity completion rates), machine learning algorithms calculate a risk score for each learner. Tutors receive automatic alerts as soon as a profile shows weak signs of dropout, allowing them to intervene in a targeted and preventive manner: personalised message, catch-up session, enhanced support.
This approach transforms the pedagogical approach: from reactive intervention (after failure) to proactive intervention (before dropout).
Use case: anticipating the organisation’s skills needs
Beyond the individual, AI can be used to analyse collective trends and anticipate future skills needs. Analytical dashboards aggregate training data (skills acquired, training courses taken, average results by area) and cross-reference it with business data (upcoming projects, technological developments in the sector, organisational transformations).
The system can thus identify critical skills gaps for the coming months and automatically suggest:
- The creation of new training modules.
- The adaptation of existing content that has become obsolete.
- The recruitment of trainers specialising in certain emerging topics.
- The implementation of retraining programmes for professions in decline.
This strategic approach to training enables organisations to stay ahead of skills needs rather than being overwhelmed by them.
5. Educational chatbots: enhanced tutoring
Use case: offering 24/7 support
The availability of trainers and tutors is inherently limited: office hours, holidays, volume of requests. This constraint creates frustration among learners, particularly in asynchronous distance learning where questions arise at any time.
Educational chatbots integrated into the LMS offer a first level of instant and permanent support. These virtual assistants can:
- Answer frequently asked questions (how to access a module, where to find additional resources, when the next virtual class is scheduled)
- Provide guidance on navigating the LMS (where to find my results, how to register for a training course)
- Suggest additional resources based on the difficulties expressed
- Collect recurring questions to improve training content
- Refer learners to a human tutor when the question exceeds their capabilities
This constant availability reassures learners and relieves trainers of first-level questions, allowing them to focus on complex support and in-depth educational relationships.
Use case: intelligent and contextual tutor
The most advanced chatbots go beyond simple automated FAQs. Connected to LMS data, they become true intelligent tutors capable of:
- Understanding the context of the learner’s progress (where are they in the course?)
- Adapt their responses to the level of mastery demonstrated
- Offer additional personalised explanations
- Suggest targeted revision exercises
- Encourage and re-motivate when faced with difficulties
The chatbot thus becomes a personalised learning companion, available on demand, patient and caring.
Examples of available actors
- Raison (formerly Corolair) offers an agentic platform that allows users to create and share AI teaching assistants tailored to the objectives, teaching methods and needs of learners. Raison’s Moodle plugin transforms course content into interactive AI assistants that are natively integrated into Moodle. Learners can ask course-related questions and receive instant answers, as well as self-assess via a bank of questions generated by AI and validated by the trainer. The assistants are deeply integrated into everyday tools, including LMS (Moodle, Blackboard, Canvas) and messaging applications (Teams, Slack, WhatsApp). The solution is GDPR-compliant, with servers hosted in France and the option of self-hosting.
- Stellia.ai also offers a 24/7 knowledge assistant that answers learners’ questions, allows them to practise and access tailored multimodal content (videos, podcasts). The assistant integrates with existing LMSs and allows trainers to track interactions, question trends and learner performance via analytical dashboards.
6. Integration into work environments: the example of Microsoft 365 Copilot
Use case: transferring Moodle to the everyday ecosystem
An alternative approach to adding more AI features to the LMS is to integrate AI directly into the everyday working environments of teachers and learners. Rather than forcing users to log in to Moodle every time, AI allows certain features to be transferred to the tools they use on a daily basis.
Enovation Solutions, in collaboration with Microsoft, has developed the local_copilot plugin, which allows users to interact with Moodle directly from Microsoft 365 applications (Word, PowerPoint, BizChat). Two separate AI agents – one for teachers and one for students – offer course management, educational monitoring and content access features, respectively, without leaving the Microsoft ecosystem.
From Word, teachers can:
- View the structure of their courses
- Create assignments and discussion forums
- Track student progress
- Post announcements
From any Microsoft 365 application, students can:
- View the courses they are enrolled in
- Self-enrol in new courses
- View overdue assignments
- Check their progress and grades
This approach differs from automatic content generation: it aims to streamline access to Moodle features by integrating them into an environment familiar to users. The LMS thus becomes invisible but omnipresent, accessible from any entry point in the digital work ecosystem.
Moodle and AI: prospects for 2026
Growing technological maturity
Recent versions of Moodle natively integrate APIs that facilitate connection with external AI services (OpenAI, Azure, Ollama). This openness accelerates innovation and allows companies to build tailor-made solutions without redeveloping the basic infrastructure.
The plugin ecosystem is rapidly expanding, driven by an active community of developers and specialist publishers. This diversity allows each organisation to build its own suite of AI tools based on its specific needs, technological maturity and regulatory constraints.
Ethical and regulatory issues
The adoption of AI in training raises key questions that cannot be ignored:
- Personal data protection: learning data is sensitive and must be processed in accordance with the GDPR. The issue of hosting (in Europe or outside Europe) and the purpose of processing becomes central.
- Transparency of algorithms: learners and trainers must understand how recommendations are generated, on what criteria courses are personalised, and how the risks of dropout are calculated. Algorithmic opacity can generate mistrust and compromise adoption.
- Fairness in recommendations: algorithmic biases can reproduce or amplify existing inequalities. An algorithm trained on historical data may favour certain profiles at the expense of others. Vigilance and regular auditing of systems are essential.
- Digital sovereignty: dependence on foreign technology players raises strategic questions for organisations, particularly in the public sector and sensitive industries. Sovereign solutions (European hosting, locally trained AI) are becoming a major criterion for choice.
Moodle, thanks to its open-source governance and GDPR compliance, offers a reassuring framework for experimenting with these technologies while controlling risks. The transparency of the code, the ability to audit algorithms and the freedom to choose hosting providers are decisive advantages in an increasingly strict regulatory environment.
In conclusion: AI, a catalyst for more human training
Paradoxically, artificial intelligence in Moodle does not dehumanise training: it frees up time for trainers, who can focus on personalised support, mentoring and group dynamics. Automatically generated content, personalised learning paths and predictive analytics do not replace pedagogical expertise, they enhance it.
AI takes care of repetitive, time-consuming tasks with low human added value: production of standard content, automatic quiz correction, detection of weak signals of dropout, recommendations based on algorithmic rules. This allows trainers to refocus on what makes their job unique: the educational relationship, fine-tuning to individual needs, motivation, inspiration, and conveying a passion for a subject.
For organisations preparing for 2026, the message is clear: AI in training is no longer an option but a strategic necessity. Moodle, with its flexibility and rich ecosystem, offers an ideal environment for experimenting with, deploying and measuring the impact of these technologies. Organisations that know how to combine artificial intelligence with pedagogical intelligence will gain a head start in the race for tomorrow’s skills.
The challenge is not to choose between humans and machines, but to find the right balance: AI as an accelerator of efficiency, humans as guarantors of meaning and quality. It is in this complementarity that the training of tomorrow will be built.
For more information, contact us at info@enovation,ie
