E-learning 2026: AI and professional training

E-learning enters a new phase

Digital professional training is no longer simply about uploading materials to a platform. Interactive scenarios, blended programmes and generative assistants are opening up new possibilities. But producing more content does not guarantee that learners understand more, practise enough or apply what they have learnt at work.

As of 13 October 2026, two developments in France invite us to examine this shift: the DEFFINUM closing review, published on 1 October, and the 18th Rencontres de la eformation, announced for 21 and 22 October. Both highlight the same question: how can technology serve learning, rather than the other way round?

For a training provider, consultant or business, the challenge is to rethink the whole approach: objectives, activities, support and evidence of progress. The examples below are suggested applications, not results attributed to Niza Learning or practices assumed to be widespread.

What developments in e-learning in 2026 tell us

DEFFINUM: documented achievements and a lesson for learning design

The official press release of 1 October 2026, issued jointly by the partners of the programme launched by the Ministry of Labour, reviews its achievements following its closure on 29 September. Managed by Banque des Territoires, DEFFINUM aimed to transform practices, stimulate innovation and encourage cooperation and the pooling of resources.

It distinguishes between 45 projects supported through a €100 million budget and 12 projects redirected to the ‘Skills and Jobs of the Future’ programme. The aggregate review reports more than one million learners reached, 6,700 organisations using the resources and over 1,800 blended programmes created or redesigned. These figures do not describe the entire French market and do not, in themselves, demonstrate a causal improvement in learning.

The outputs include learning management systems (LMS), e-learning modules, assessment tools and immersive solutions. The press release reports AI in 70% of projects, with applications still emerging. Its central lesson is more measured: digital effectiveness depends on sound instructional design, contextualisation and support. The publication of an evaluation summary and a booklet of 16 inspiring practices offers an opportunity to build on experience, rather than an invitation to copy tools without first assessing needs.

Rencontres de la eformation: an event still to come

The ministerial announcement published on 15 September and updated on 16 September schedules the 18th Rencontres for 21 and 22 October 2026 at Université Gustave Eiffel in Champs-sur-Marne. At the time of writing, the event has not yet taken place, so no lessons from its discussions can be presented as established findings.

The announced programme dedicates 21 October to the impact of AI on training professions: instructional design, content production, learner support and skills. On 22 October, it will focus on remote interactions, cooperation and the collective construction of knowledge. These issues remind us that an assistant available around the clock does not automatically replace dialogue with a trainer.

Another development in 2026: distinguishing performance from learning

Published on 19 January, the OECD Digital Education Outlook 2026 analyses educational uses of generative AI. It stresses that a better answer produced with a tool does not necessarily mean a skill has been acquired. Its research notably covers school education: applying it to adults requires caution, but it usefully encourages us to measure what learners can do without assistance.

Five major transformations in e-learning

These transformations combine documented developments with possibilities to explore through experimentation. They do not imply uniform adoption of tools or the automatic personalisation of all training.

1. Generative AI supports instructional design

The OECD describes AI as an assistant, partner or tutor when its use serves an explicit learning purpose. A provider might, for example, request three versions of a management case study based on a fictional situation. The designer then selects the one that genuinely develops the intended skill and checks the model answers: generation remains a preparatory step.

2. Personalisation becomes a design choice

DEFFINUM documents more flexible, individualised programmes without establishing that every adaptation relies on AI. Within a business, an initial assessment can direct learners towards a vocabulary refresher or an advanced simulation. Adding an assistant that offers alternative explanations is an option to test; learners must retain the ability to request human support.

3. Blended learning combines independence with human interaction

The programme’s blended learning pathways combine face-to-face and remote learning with synchronous or asynchronous sessions. For team leaders, a self-paced module can prepare them for a virtual class analysing situations, followed by a review of workplace application. Group time is then used to discuss reasoning and receive feedback, not to repeat a video already watched.

4. Assessment seeks evidence of skills

The tracking tools identified by DEFFINUM make it easier to observe the learning journey; they do not make completion rates equivalent to mastery. Data protection training can combine quizzes, spotting errors in a fictional document and an oral explanation of the learner’s reasoning. AI can offer preliminary feedback; explicit assessment criteria and human judgement make it possible to evaluate the actual reasoning.

5. Instructional design, accessibility and quality become more important

When content becomes easy to produce, selection and coherence become decisive. The UNESCO AI competency framework, published in 2024 and updated in January 2026, notably emphasises critical judgement and a human-centred approach. Intended for school students, it offers guidance, not a professional training standard. One example is turning a long presentation into short activities with transcripts, clear instructions and contextualised exercises.

How can AI be used without compromising learning quality?

Effective use begins with a brief: audience, prerequisites, observable objective, duration, workplace situation and success criteria. AI can suggest a scenario, distractors for a quiz, a case study or a revised version of learning materials. It does not decide on your behalf what is worth learning.

Consider a module on handling a complaint. The trainer provides an approved procedure, requests a fictional case and several possible responses, then checks their realism. They verify that the quiz explains mistakes and that the case study requires a reasoned choice. A second person can review sensitive material before publication.

The limitations remain tangible: a convincing factual error, an invented source, superficial content, a stereotype or an ambiguous model answer. Confidentiality, copyright and dependence on a supplier add further considerations. A resource available online is not automatically free to reuse; check licences and rights for documents shared or reproduced.

The trainer validates subject-matter accuracy; the instructional designer organises progression and activities. Plan for an exportable version of the content, a review date and exercises without AI. This discipline prevents a change of tool or an automated answer from becoming the weak point in the programme.

Precautions to take: GDPR, security and responsibility

Legal obligations: when an LMS or assistant processes personal data, the GDPR applies. Articles 5 and 6, reproduced by the CNIL, require, among other things, a defined purpose, a lawful basis, data minimisation and limited retention. Consent is not the only possible basis: the choice depends on the processing and must be justified.

The six key principles presented by the CNIL also highlight information, the exercise of rights and appropriate security. Before connecting a tool, examine the data sent, access permissions, retention and the provider’s safeguards. Avoid shared accounts and identifiable personal records in an unapproved assistant; use fictional cases where they suffice.

Recommendations: the AI guidance finalised by the CNIL on 22 July 2025 mainly concerns system development and does not certify tools used in training. Another development: on 9 July 2026, the CNIL shared the EDPB guidelines on anonymisation and web scraping. Open for consultation until 30 October, they must not be presented as a final text. Removing names does not, in itself, guarantee anonymity.

Good learning practices: disclose the involvement of an assistant, explain its limitations, and provide for human verification and an identifiable contact person. Test keyboard navigation, subtitles, transcripts and readability with the audiences concerned. These choices support accessibility; the precise legal obligations depend on the organisation’s status and the service. A checklist is no substitute for analysing them.

What should training providers do now?

Audit content and needs: identify outdated resources, workplace challenges, prerequisites and learners’ access constraints. Also establish how much time trainers actually have available.
Define learning objectives: specify what participants must be able to do, under what conditions and against which criteria. Distinguish awareness-raising, practice and skills validation.
Identify modules suitable for digital delivery: select input and exercises suited to independent learning; reserve complex situations for discussion, guided practice or face-to-face learning.
Choose suitable tools: compare LMS platforms on accessibility, tutoring, assessments, exports, security and total costs. An AI feature must address an explicit need.
Test a pilot programme: involve representative learners, observe obstacles and check content on their devices. Document feedback rather than relying solely on a sales demonstration.
Measure and improve: combine before-and-after progress, the quality of work produced, dropouts and workplace transfer. Review instructions and support when results do not meet objectives.

Operational checklist: eight indicators of maturity

For each point, look for concrete evidence. A ‘no’ indicates an issue to address before deployment.

Audiences, needs and access constraints are documented.
Each module has an objective and a success criterion.
Content is sourced, up to date and validated by an expert.
Activities develop skills through practice, beyond simply watching content.
Human support and response times are planned.
Accessibility is tested on learners’ devices.
Data, providers, rights and AI use are governed by clear rules.
The pilot, indicators and programme review are planned.

Digital technology does not replace instructional design

E-learning in 2026 opens up opportunities for design and support, but the developments examined do not demonstrate that automation alone is enough. Technology becomes useful when combined with subject-matter expertise, human interaction and rigorous assessment. Responsible AI begins with limited, verified uses that are improved through learners’ experience.

To explore this further, discover Niza Learning’s support for training design and digitalisation, along with approaches for training providers and businesses. Your learning needs remain the starting point, not the choice of a tool.

E-learning in 2026: how is artificial intelligence transforming professional training?