de
Year-end trend articles have a problem: they list ten developments, and a year later nobody remembers which came true. This piece tries something else. It describes four shifts that are already visible, and for each the question L&D teams should be asking for 2027.
Production gets cheap, steering gets expensive
The biggest change is also the least discussed. With AI-assisted production, effort per training hour typically drops by up to 80%. When a course takes hours rather than weeks, creation is no longer the bottleneck.
The bottleneck moves. Who decides which training should exist? Who checks it is correct? Who keeps it current? And who makes sure a hundred new courses do not come with a hundred different quality standards? These tasks used to be a side effect of production. In 2027 they become the main job.
The question for 2027: which approval and quality rules apply when business units create training themselves?
Experiential knowledge becomes the scarcest resource
By 2036 Germany will be short of roughly 4.3 million workers, as around 1.3 million people retire each year while only about 800,000 enter (IW Köln, 2026). For L&D that means a shift in sources. Training increasingly comes not from textbooks and manuals but from what experienced colleagues know and will soon no longer be around to tell.
Knowledge capture and training creation therefore move closer together. What that looks like in practice is shown in the piece on phased retirement.
Regulation demands speed and proof
NIS-2 has been in force since 6 December 2025, and the AI literacy duty under Article 4 of the EU AI Act has applied since February 2025. Both increase the pressure to adapt training quickly and document every completion cleanly. What that means for L&D is covered in the piece on NIS-2 and the EU AI Act.
At the same time infrastructure is thin in many places. 52% of companies in the DACH region have no digital learning platform, although 70% consider digital topics relevant (Masterplan / Amadeus Fire, 2025).
Impact is asked for, not attendance
When training becomes cheap, expectations of its impact rise. Participant numbers and completion rates count for less and less as proof of success, especially to leadership that is simultaneously seeing 95% of AI pilots deliver no measurable ROI (MIT NANDA, 2025). L&D will have to show what changes in operations. How to do that without a large evaluation project is described in Measuring learning success.
What we do not know
We deliberately leave one shift open: how learners themselves will use AI. Whether employees in 2027 prefer a short video, asking an AI chat, or both, cannot be predicted seriously today. Whoever claims to know usually sells the matching product.
What can be said: the teams that are well positioned in 2027 have shifted their role from producer to steerer. Why a patchwork of single tools is a poor basis for that is covered in the piece on best of breed in the L&D stack.
