AI is changing how your participants learn, how you assess them and what you are accountable for. Academies should answer five questions now.
A participant on an accountancy qualification course pastes a practice task into ChatGPT. The answer sounds convincing, but it follows a different structure from the course notes and misses a rule that matters in the exam. Her trainer never finds out.
Situations like this are routine. 58 percent of people in Germany aged 16 and over use AI at least occasionally (Bitkom Research 2026). Among university students, 46.5 percent use AI for their studies every day (CHE, July 2026). Comparable surveys for continuing education are missing. There is little reason to think adults on a professional course learn differently, though: in the evenings, at weekends and with questions that come up between classroom sessions.
At the same time, the sector is under economic pressure. The wbmonitor business climate score from BIBB and DIE fell to minus 4 points in 2025, 17 points below the previous year (bildungsklick). That makes it all the more important to tackle the right AI questions first. This article sets out the five most pressing ones, based on current studies and on the legal position after the EU AI Act was amended in July 2026.
Key takeaways
- Explanations are no longer scarce. A course earns its value through content aligned with the curriculum and the exam, and through the support of its trainers.
- Take-home papers and written practice tasks say less about competence than they used to. Experts favour oral and practical formats.
- Article 4 of the EU AI Act has applied since February 2025 and requires measures that support the AI literacy of your staff and trainers.
- AI that evaluates learning outcomes or monitors exams counts as high-risk. Those obligations apply from 2 December 2027.
- Your content and your participants' data need clear rules before they end up in third-party AI tools.
Learners already ask general-purpose AI. What is left for the course?
In the mmb Learning Delphi 2025/2026, an annual survey of 77 digital learning experts, 95 percent see chatbots and learning assistants playing a central role in vocational education. 75 percent expect them to be the most commercially successful learning format (mmb-Trendmonitor 2025/2026). Video tutorials and learning nuggets, by contrast, are losing support.
The authors raise an uncomfortable point. Most of these chatbots come from companies outside the education sector. Many tasks for which learners once needed material from an education provider are now handled by a general-purpose language model, from explanations to exam flashcards.
For academies, this means an explanation alone is no longer a reason to book a course. What a general tool cannot supply is content aligned with the curriculum, the current legal position and the exam, and trainers who know where their group stands. The question for your academy is therefore: does your content reach learners at the moment they ask their questions?
Take-home papers and practice tasks say less about competence
Respondents to the mmb Learning Delphi consider written formats such as take-home papers and written exams less reliable as proof of competence now that AI is widely available. On a scale from 1 (very suitable) to 6, they rate practical demonstrations (1.8) and oral exams in person (1.9) highest. Take-home essays and papers come last at 3.7 (mmb-Trendmonitor 2025/2026).

Synchronous formats do take more examiner time and more scheduling effort. 41 percent of respondents also see gaps in how AI in assessment is regulated.
In regulated professional qualifications, exam regulations set the format of the final exam. Your room for manoeuvre lies in exam preparation and in internal assessments. This is where it pays to ask which exercises still show what participants can do. One pointer from the same survey: work samples with a documented process (2.4) score clearly better than work samples where only the final result is available (3.2). What counts is a visible learning path.
EU AI Act: what has applied since 2025 and what follows from December 2027
The EU AI Act (Regulation (EU) 2024/1689) affects academies in three places. The Digital Omnibus amending regulation (Regulation (EU) 2026/1744) has been in force since 27 July 2026 and changed two of them.
AI literacy (Article 4). Since 2 February 2025, providers and deployers of AI systems have had to take measures on the AI literacy of their staff and of people using AI on their behalf. For academies, that means staff and trainers. The Omnibus rewrote the article. It now requires measures that support the development of AI literacy, without having to guarantee a specific level for any individual (Regulation (EU) 2026/1744). The duty remains, and your measures should be documented. The article does not require a certificate or a test. Which measures make sense and what to record is covered in our guide to AI literacy under Article 4.
Transparency (Article 50). Since 2 August 2026, people interacting with an AI system must be able to recognise that they are doing so (Article 50). This applies, for example, to a learning assistant in your course.
High-risk uses (Annex III, point 3). AI systems that decide on admission, evaluate learning outcomes, assess the appropriate level of education or monitor exams for prohibited behaviour count as high-risk (Annex III). Under the Omnibus, those obligations apply from 2 December 2027 instead of August 2026. This is relevant for academies planning AI-assisted grading or exam proctoring. In the mmb Learning Delphi, 61 percent of experts see a central role for AI proctoring, up from 47 percent the year before.
In our reading, a learning assistant that explains content and offers practice does not fall under these cases. The specific purpose is what decides. This assessment does not replace legal advice.
Copyright and data protection: your content is your capital
Course notes, case collections and exam questions are what set an academy apart from a general AI tool. This is exactly where experts see the biggest gaps. 57 percent consider AI use insufficiently regulated with regard to copyright, and 39 percent say the same of data protection. 69 percent expect education providers to pay more attention to the data security of their infrastructure (mmb-Trendmonitor 2025/2026).
In practice, the risk often goes unnoticed. A trainer uploads her course notes to a personal chatbot account to generate practice questions. A participant pastes in a case study that includes names from his workplace. You should therefore settle three points:
- Rights to materials. Who holds the usage rights to course notes written by freelance trainers and to licensed third-party content? Does the licence cover processing by AI?
- Data processing agreements. Every tool that processes your participants' personal data on your behalf needs a contract under Article 28 GDPR.
- Usage rules. A short policy sets out which tools staff and trainers may use with which content. It is also an obvious part of your measures under Article 4.
Trainers between relief and extra work
77 percent of respondents expect AI to save teaching staff time. Almost as many (73 percent) expect teaching staff to spend a lot of time checking the quality of AI output. 75 percent fear that learners will lose important skills through AI use. In 2024, the figure was 44 percent (mmb-Trendmonitor 2025/2026).
The human role gains weight as a result. 84 percent expect learners to seek more direct exchange with teaching staff and fellow learners. 83 percent consider it likely that chatbots and teaching staff will work in tandem to handle learners' subject and motivational questions.
For trainers, the work shifts: less repeating the same explanation, more context, feedback and conversation. They need two things for this. Confidence in working with AI, and insight into where their group is actually getting stuck.
What academy leaders should prioritise now
| Question | What to clarify | First step |
|---|---|---|
| Value of the course | Where is your content better than a general AI answer? | Put typical participant questions to a general chatbot and compare the answers with your course notes |
| Assessment and practice | Which exercises still show what participants can do? | Review internal assessments for formats with feedback and conversation |
| AI literacy | Which measures do you document under Article 4? | Plan and record measures such as an introduction or training for staff and trainers |
| High-risk uses | Do you plan AI for grading, placement or exam proctoring? | Check planned uses against Annex III and set a timeline up to December 2027 |
| Content and data | Which tools may work with which content? | Draft a usage policy, check data processing agreements and rights to materials |
| Role of trainers | How does AI support your teaching team? | Choose one course as a pilot and involve trainers from the start |
Many of these questions lead to the same point. Your participants use AI anyway, so that AI should work with your content and remain under your control. This is what acemate works on. The AI tutor answers from the academy's own materials and cites document and page. Educators decide how it responds, and Analytics show the teaching team in aggregated form where understanding breaks down. In the DIHK Bildungs-gGmbH pilot, around 90 percent of participants wanted to keep using the learning assistant for the rest of their course (IHK AILA case study).
None of these questions has to be answered for every course at once. One course with a lot of self-study and a clear exam is enough to gain experience and to base the next steps on your own observations. If you would like to talk this through, the page for academies and continuing education gives an overview and lets you arrange a conversation.



