\nVisualization: A robotic arm on the workbench and above it its interactive 3D model with a highlighted cable connection | Image: \u00a9 Ulrich Buckenlei | XR Stager Online Magazine | VISORIC GmbH
\n<\/sup><\/p>\n
<\/p>\n
Anyone who maintains a machine, repairs a device or assembles a component usually works with instructions that describe the product but do not show it. Exploded views, item numbers and blocks of text deliver the information, yet the actual transfer onto the real part remains the job of the person doing the work. It is at this point of translation that questions, search times and errors arise, especially when connections are hidden, variants differ only in detail or a work step occurs only rarely. For manufacturers of complex products, these small uncertainties add up to noticeable costs, in assembly, in field service and in every training session for new employees.<\/p>\n
An example from robotics shows how this gap can be closed. The tnkr platform brings together assembly instructions, bills of materials and interactive 3D visualization for robotics projects in one place.[1]<\/sup> In a recent demonstration, the cables of a robot appear one after another in the 3D model, individual components can be selected, real photos complement the digital representation, and a connection list states cable number and cross-section. An integrated AI assistant called Leonardo also analyzes first-person videos and generates step-by-step assembly instructions from them.[2]<\/sup><\/p>\n What makes this example interesting is less the robotics than the principle behind it. The instructions are no longer a separate document lying next to the product, but an interactive view of the product itself. They can be rotated, played back and checked step by step. They are fed by data that already exists in many companies: CAD models, bills of materials, assembly sequences and the practical knowledge of employees, which can increasingly be captured on video.<\/p>\n On top of this comes a regulatory impulse. From 20 January 2027, the new Machinery Regulation applies in the EU, which for the first time explicitly permits operating instructions in digital form.[9]<\/sup> For manufacturers, the question of what good digital instructions look like thus shifts from an innovation question to a strategic one. This article puts the development into sober perspective, because research shows clear advantages, but equally that the benefit depends on the task, data quality and implementation.<\/p>\n The following article shows where classic instructions fail, what interactivity and context can achieve, how CAD data and AI make creation scalable, where the practical limits lie and how companies can get started with a measurable pilot.<\/p>\n Printed assembly instructions are a remarkably robust tool. They work without power and without a device, can be placed next to the workbench and are still the standard in many quality management systems today. Their weakness lies in the form of presentation: a spatial process is reduced to a flat page. The reader has to assemble a spatial picture from several views, derive the correct sequence and mentally reconstruct the movement of a part before even reaching for a tool.<\/p>\n The explainer video has taken over part of this work because it shows movement directly. But it brings its own limits. A video follows a fixed narrative, a fixed camera position and a fixed pace. Anyone who is unsure at a particular point rewinds, pauses and searches for the decisive second, instead of looking at the connection from the angle that would be helpful for their own situation. Hidden parts remain hidden, and variants that differ from the filmed unit simply do not appear in the video. Added to this is the maintenance effort: if a part changes, the entire sequence often has to be reshot and re-edited.<\/p>\n How large the difference between printed and digital instructions can be was examined in an experiment by RWTH Aachen University, published in the Journal of Operations Management. There, digitally animated, interactive work instructions led to better performance than paper-based instructions. A second finding is remarkable: a combination of digital and printed instructions brought no additional advantage, which is why the authors recommend consistently replacing paper rather than merely supplementing it with digital content.[3]<\/sup><\/p>\n <\/p>\n A systematic review of augmented reality in maintenance, which evaluated 30 primary studies, confirms the fundamental potential of bringing information closer to the object and to the specific work step. However, it equally identifies technical hurdles that have so far stood in the way of broad adoption.[4]<\/sup> The message is therefore nuanced: paper and video do not lose their justification, but for spatially complex tasks, considerable potential for improvement remains untapped.<\/p>\n The decisive question is therefore not whether instructions will become digital, but which properties digital instructions must have so that they actually take over the translation work instead of merely moving a PDF onto a screen. The next chapter shows what these properties are.<\/p>\n The direction is therefore clear, but the path there is not trivial. A PDF on a tablet is not yet better instructions, merely a different place for the same drawing.<\/p>\n <\/p>\n The core of interactive 3D instructions is a shift in control: it is no longer the author who determines what the user sees at which moment, but the user themselves.<\/p>\n In concrete terms, this means: the model can be freely rotated and enlarged, housing parts can be hidden to reveal internal connections, and motion sequences can be played back as often as needed and at any speed. Anyone who wants to check a connection selects it directly in the model and receives the associated data, such as part number, tightening torque or cable cross-section. The instructions are thus transformed from a predefined narrative into a tool that adapts to the understanding needs of each individual. It is important that this freedom does not lead to overload. Good interactive instructions therefore specify a recommended path, but allow deviations at any time, for example to view a step from a different perspective or to examine a detail more closely.<\/p>\n That this additional control can have measurable effects, but does not automatically have them, is shown by an experiment published in the journal Automation in Construction. Using the example of pipe maintenance, the authors compared 2D documents, interactive 3D representations and virtual reality with a total of 120 participants and examined both learning the task and its later execution.[5]<\/sup> The value of such studies lies in the fact that they do not assess information formats by their attractiveness, but by their effect on a real activity.<\/p>\n A scoping review in the journal Human Factors, which evaluates 24 studies on AR-, VR- and MR-supported training for manual assembly tasks, reaches a balanced verdict: the results are promising, but by no means consistent.[6]<\/sup> Depending on the task, design and device, the effects vary. Interactivity is therefore a lever, not an automatism.<\/p>\n <\/p>\n For companies, this leads to an important distinction between presentation and instruction. A 3D model that impresses at a trade fair does not necessarily help in the workshop. Effective instructions are designed to answer precisely the questions that actually arise during a specific task: which part comes next, how is it aligned, and how do I know that the step has been completed correctly?<\/p>\n Such questions rarely arise at the beginning of a task, but in the middle of it, often at exactly the part that is currently in one’s hand. The next chapter shows how instructions can provide targeted support at this point.<\/p>\n Interactivity therefore only unfolds its value when it is tailored to the specific questions of a specific task. This is less a question of technology than a question of instructional design.<\/p>\n <\/p>\n Anyone in the middle of a repair rarely needs the complete instructions. What is needed is a single view, a single connection or a single safety note, and exactly now.<\/p>\n Linear instructions force the user to search for this point. Interactive instructions, by contrast, can offer it directly at the relevant part. The person pauses, jumps back, isolates a single part and triggers the next step only when the previous one has really been completed. This creates support that adapts to the pace and level of knowledge of each individual: experienced technicians skip what they already know, while new employees go into depth exactly where they are uncertain.<\/p>\n In research, this principle is discussed under the term context-aware documentation. A paper published in the journal Applied Sciences describes next-generation technical documentation in which an information manager selects which content is displayed in which situation, instead of always providing the complete documentation.[7]<\/sup> The instructions thus respond to the work context instead of ignoring it.<\/p>\n This approach becomes particularly relevant as workforces become more heterogeneous. A study on projection-based work instructions published in 2026 refers to earlier findings according to which animated digital instructions improved productivity by 12 percent and product quality by 52 percent. The study itself shows that digital instructions can reduce education-related performance differences in complex assembly.[8]<\/sup><\/p>\n <\/p>\n This fundamentally changes the role of instructions. They are no longer a reference work that is read before the job and put aside afterwards, but a companion that is present during the task, holds back during routine work and offers targeted help when uncertainty arises. For companies facing skills shortages and changing teams, this is more than a gain in convenience. There is also an often underestimated effect: when information appears in the right place, the number of questions to experienced colleagues decreases, and their time is already scarce in many companies.<\/p>\n What remains open is whether this principle carries beyond special cases such as robot building and why it is becoming relevant for many manufacturers right now. The next chapter addresses both questions.<\/p>\n Context is therefore the real added value compared with any linear presentation. However, it requires the instructions to know which part and which step a piece of information belongs to.<\/p>\n <\/p>\n The principle is not limited to robotics. Wherever people carry out spatially demanding, error-prone or rarely recurring work steps, the same question arises. Consumer electronics and household appliances, machine and plant service, cable harness production, plant assembly and technical training all follow the same pattern. In service, the affected assembly can be isolated, in cable harness production the cable routing becomes traceable step by step, and in training, assembly sequences can be practiced before the real product is even available.<\/p>\n What is new is a regulatory push. From 20 January 2027, the Machinery Regulation (EU) 2023\/1230 replaces the previous Machinery Directive and applies directly in all member states. For the first time, it explicitly permits the operating instructions to be provided in digital form, and digital assembly instructions also become permissible for partly completed machinery.[9]<\/sup><\/p>\n However, this relief is tied to conditions. The manufacturer must state how the digital instructions can be accessed, they must remain available for a defined period, and a paper version must still be provided free of charge at the customer’s request.[10]<\/sup> The regulation therefore does not prescribe interactive 3D instructions, but it opens up the framework within which digital formats can become the norm.<\/p>\n <\/p>\n For manufacturers, this shifts the perspective. Anyone who has to build digital instructions anyway inevitably faces the question of whether a downloadable PDF is sufficient or whether the transition should be used to make instructions genuinely easier to understand. This is where the strategic opportunity lies: the digitalization that is coming anyway can become a quality improvement instead of remaining a mere change of medium. At the same time, user expectations are changing. Anyone who configures products via an app and calls up instructions on a smartphone in everyday life is increasingly reluctant to accept an exploded view in a binder in a professional environment.<\/p>\n Decisive for economic implementation, however, is how much effort the creation of such instructions requires, especially with many products and variants. The next chapter shows how this effort can be limited.<\/p>\n The Machinery Regulation is therefore not a driver for 3D technology as such, but it is an occasion to fundamentally reassess one’s own technical documentation. Those who use this occasion combine compliance with measurable added value.<\/p>\n <\/p>\n The real economic bottleneck of interactive instructions lies not in presentation, but in creation. If every set of 3D instructions is built, animated and labeled by hand, the effort quickly exceeds the benefit with a broad product portfolio. In addition, every product change can potentially make the instructions outdated, often without anyone noticing.<\/p>\n The key lies in data that already exists in many companies. CAD models contain geometry and assembly structure, bills of materials contain parts and quantities, and routing sheets contain the sequence. Researchers from Flanders Make and Hasselt University have developed a tool that semi-automatically derives an assembly sequence from a CAD model. The method works backwards: it determines the order in which parts can be removed from the assembly and then reverses this sequence. On this basis, digital work instructions including visualizations and animations are created, which the process engineer can adapt and supplement.[11]<\/sup><\/p>\n Fraunhofer IGCV pursued a complementary approach in the SynDiQuAss project. Starting from a template for an entire product family, the assembly process is broken down into elementary work steps, after which existing information such as CAD representations, tools and parts is linked automatically. According to the authors, this tool-supported generation can increase productivity and product quality in manual assembly, especially for small batch sizes.[12]<\/sup><\/p>\n <\/p>\n Both studies show what matters: not individual, perfectly designed instructions, but a reusable data flow. CAD models, bills of materials, assembly sequences, media and standardized step modules form a pipeline that covers new products and variants with manageable additional effort. Feedback is equally important: if a part changes, this must result in a traceable version of the instructions, otherwise the instructions describe a product that no longer exists in that form. In practice, this means close integration with existing product data management systems, so that changes do not have to be updated manually, but automatically end up as a review task in the documentation.<\/p>\n However, CAD data only provides the geometry and a possible sequence. The practical knowledge of how a manual operation is actually best performed is not contained in the model, but in the heads and hands of experienced employees. The next chapter shows how artificial intelligence can unlock this knowledge.<\/p>\n Scalability is therefore above all a question of data architecture. Those who maintain their product data in a structured way lay the foundation on which every further automation builds.<\/p>\n <\/p>\n Artificial intelligence comes in where CAD data reaches its limits. It can search existing manuals, recognize work steps in videos, identify components and formulate initial texts for individual steps. For companies, this is particularly interesting because it makes it possible to capture implicit expert knowledge that previously had to be painstakingly transferred through interviews and training materials, and that threatens to be lost when experienced specialists leave.<\/p>\n In 2024, a research team led by Peter Burggr\u00e4f and Tobias Adlon presented a concept in Procedia CIRP that aims to automatically derive assembly instructions from recordings of real work processes.[13]<\/sup> The idea behind it: instead of drafting instructions at a desk, an experienced worker is recorded at work, and the system derives the individual work steps from the recordings. The assistant described in the opening example uses the same principle, translating first-person videos into assembly steps.<\/p>\n How far this development has come is shown by a paper on AI-supported worker guidance published in August 2026. It combines recordings from first-person and external perspectives with multimodal video understanding and structured task knowledge. In a case study on the disassembly of a workstation, the system reliably recognized actions and suggested valid next steps. It is remarkable that a single expert demonstration was already sufficient to derive meaningful action guidance, while multiple recordings further increased robustness.[14]<\/sup><\/p>\n <\/p>\n As impressive as this progress is, the reliable statement remains: AI delivers candidates and drafts, not approved instructions. Geometry, sequence, variants, safety notes and real-world feasibility must be checked by experts. An incorrectly described manual operation in assembly instructions is not a cosmetic flaw, but a quality and safety risk and, in case of doubt, a matter of manufacturer liability. A release process is therefore advisable that clearly defines who reviews an AI draft, which criteria apply and how corrections are documented. These corrections in particular are valuable because they show where the tools work reliably and where they do not.<\/p>\n In practice, this results in a clear division of labor: AI shortens the path to the first draft, while approval remains with people who bear professional responsibility. Yet even well-created and carefully reviewed instructions sometimes fail in everyday use for entirely different reasons. The next chapter shows what these are.<\/p>\n Artificial intelligence therefore mainly changes the speed at which instructions are created, not the responsibility for their content. Those who draw this line clearly can use the tools to great benefit.<\/p>\n <\/p>\n A convincing demonstration is not yet a working series deployment. The most common stumbling block is data quality: CAD models are incomplete, represent a design state instead of the product as actually built, or are too detailed for smooth display on tablets and in browsers. Instructions that deviate in detail from the real product quickly destroy the trust on which their benefit is based. Performance is another factor: a model with millions of polygons may run smoothly at the design workstation, but not on a tablet on the shop floor or in the customer’s browser. Data preparation is therefore a separate work step that should be planned from the outset.<\/p>\n The second stumbling block is maintenance effort. A paper on augmented reality in industrial maintenance describes how scaling remains costly because content creation requires skills in 3D modeling and programming. In addition, processes become outdated so quickly that maintenance technicians sometimes abandon written instructions altogether.[15]<\/sup><\/p>\n A case study on personalized work instructions in assembly comes to a similar conclusion. Companies struggle with personalized instructions because maintaining them is time-consuming and a large number of instructions must be kept up to date. In practice, personalization mainly takes place through personal mentors, while technology played hardly any role in creation and delivery.[16]<\/sup> The authors conclude that the creation, introduction and maintenance of work instructions must become significantly more accessible.<\/p>\n <\/p>\n Variants and design add to this. Every product variant potentially multiplies the number of instructions to be maintained, and more visual effects do not automatically mean more understanding. Well-made interactive instructions are therefore reduced rather than spectacular: they show exactly what is needed for the respective step and leave everything else out. Tests with real users are therefore just as much part of the project as the 3D model.<\/p>\n None of these stumbling blocks speaks against interactive instructions, but they do speak against an approach that starts with the technology. The next chapter shows what an approach that makes these risks visible early on looks like.<\/p>\n Those who know these stumbling blocks plan more realistically and avoid the most common mistake of inferring smooth series deployment from a successful demo. Data situation, responsibilities and maintenance processes therefore belong at the start of every project.<\/p>\n <\/p>\n The most sensible entry point is not the digitalization of the entire product portfolio, but a single process on which the benefit can be measured. Suitable candidates are work steps that occur frequently, regularly lead to errors or questions, or cause particular difficulties for new employees. That is where the leverage is greatest and the effect becomes visible fastest. The selection of participants is just as important: those who will later use the instructions should already be involved in the pilot, because only then will it become clear whether the solution answers the actual questions at the workplace.<\/p>\n Before implementation comes the choice of the right format. A paper on the design of digitalized work instructions points out that a one-to-one transfer of paper-based instructions into a digital format often fails to exploit the potential of digitalization and does not suit the hardware used. Instead, it provides a decision aid that matches digitalization options with the environmental, work and process conditions of the respective company.[17]<\/sup><\/p>\n Measurement is decisive afterwards. Comprehension, error rate, processing time and maintenance effort should be recorded before and after introduction, ideally in direct comparison with the previous instructions. Methodologically, this can build on a frequently cited study that as early as 2003 compared assembly tasks with printed instructions, screen-based instructions and spatially anchored augmented reality in terms of time and errors.[18]<\/sup><\/p>\n <\/p>\n A well-documented pilot delivers more than a yes-or-no decision. It shows which data is missing, which step modules can be reused, how high the maintenance effort realistically is and which processes would benefit next. In this way, an experiment becomes a sound basis for investment decisions, especially with a view to the digitalization of operating instructions that is coming in 2027 anyway.<\/p>\n The technical building blocks, from the CAD data basis and AI-supported drafts to interactive display in the browser, are available today. Whether they pay off is not decided in theory, but on a specific process with specific metrics.<\/p>\n At the end of this path are instructions that no longer lie next to the product, but show the product itself. The following video shows what this can look like in practice.<\/p>\n <\/p>\n <\/p>\n How tangible this principle already is becomes clear from the tnkr platform example described at the beginning. The platform is aimed at developers and makers who build robots and brings together hardware data, code and models in a shared project, including integration with tools such as Onshape and GitHub.[2]<\/sup><\/p>\n In the embedded video, the assembly instructions of a robot are built up step by step in the 3D model. Cables appear one after another, individual components can be selected, real photos complement the digital representation, and a connection list states the cable number and cross-section for each cable.[1]<\/sup><\/p>\n The example thus combines almost all of the building blocks described in this article: an interactive model that the user controls, information anchored to the respective part, and an AI that generates initial instruction drafts from videos.<\/p>\n\n
\n<\/p>\nWhy Paper and Video Reach Their Limits<\/h2>\n
![\"Three-part](\"https:\/\/www.xrstager.com\/wp-content\/uploads\/2026\/06\/visoric-cad-ai-instructionspg-001.jpg\")
<\/p>\nThe same assembly, three very different paths to understanding.<\/h6>\n
\nInfographic: Exploded view, linear video and interactive 3D model in comparison | Graphic: \u00a9 Ulrich Buckenlei | XR Stager Online Magazine | VISORIC GmbH
\n<\/sup><\/p>\n\n
What Makes Interactive 3D Instructions Different<\/h2>\n
![\"Technical](\"https:\/\/www.xrstager.com\/wp-content\/uploads\/2026\/06\/visoric-cad-ai-instructionspg-003.jpg\")
<\/p>\nThe user controls viewing angle, pace and level of detail.<\/h6>\n
\nInfographic: The key control options of interactive 3D instructions at a glance | Graphic: \u00a9 Ulrich Buckenlei | XR Stager Online Magazine | VISORIC GmbH
\n<\/sup><\/p>\n\n
The Right Information at the Right Moment<\/h2>\n
![\"Side-by-side](\"https:\/\/www.xrstager.com\/wp-content\/uploads\/2026\/06\/visoric-cad-ai-instructionspg-003-1.jpg\")
<\/p>\nHelp exactly at the part that matters right now.<\/h6>\n
\nInfographic: From a linear sequence of steps to context-based support at the relevant part | Graphic: \u00a9 Ulrich Buckenlei | XR Stager Online Magazine | VISORIC GmbH
\n<\/sup><\/p>\n\n
Why the Topic Is Moving onto the Agenda Now<\/h2>\n
![\"Infographic](\"https:\/\/www.xrstager.com\/wp-content\/uploads\/2026\/06\/visoric-cad-ai-instructionspg-004.jpg\")
<\/p>\nMany fields of application and a new legal framework from 2027.<\/h6>\n
\nInfographic: Fields of application for interactive 3D instructions and the regulatory framework of the EU Machinery Regulation | Graphic: \u00a9 Ulrich Buckenlei | XR Stager Online Magazine | VISORIC GmbH
\n<\/sup><\/p>\n\n
CAD Data as Raw Material for Instructions<\/h2>\n
![\"Diagram](\"https:\/\/www.xrstager.com\/wp-content\/uploads\/2026\/06\/visoric-cad-ai-instructionspg-005.jpg\")
<\/p>\nExisting product data becomes a reusable pipeline.<\/h6>\n
\nInfographic: How CAD models, bills of materials and assembly sequences flow into a reusable data pipeline for instructions | Graphic: \u00a9 Ulrich Buckenlei | XR Stager Online Magazine | VISORIC GmbH
\n<\/sup><\/p>\n\n
AI as an Assistant to Technical Writing<\/h2>\n
![\"Infographic](\"https:\/\/www.xrstager.com\/wp-content\/uploads\/2026\/09\/visoric-cad-ai-instructionspg-006b.jpg\")
<\/p>\nThe AI creates the draft, the specialist approves it.<\/h6>\n
\nInfographic: From the recorded manual operation via the AI draft to professionally approved instructions | Graphic: \u00a9 Ulrich Buckenlei | XR Stager Online Magazine | VISORIC GmbH
\n<\/sup><\/p>\n\n
Where Interactive Instructions Fail in Practice<\/h2>\n
![\"Infographic](\"https:\/\/www.xrstager.com\/wp-content\/uploads\/2026\/06\/visoric-cad-ai-instructionspg-007.jpg\")
<\/p>\nFour factors determine practical deployment.<\/h6>\n
\nInfographic: Data quality, maintenance effort, variants and design as the supporting prerequisites of interactive instructions | Graphic: \u00a9 Ulrich Buckenlei | XR Stager Online Magazine | VISORIC GmbH
\n<\/sup><\/p>\n\n
A Pilot as the Path to a Sound Decision<\/h2>\n
![\"Process](\"https:\/\/www.xrstager.com\/wp-content\/uploads\/2026\/06\/visoric-cad-ai-instructionspg-008.jpg\")
<\/p>\nOne process, clear metrics, a sound decision.<\/h6>\n
\nInfographic: From process selection via data review and pilot to measuring the actual benefit | Graphic: \u00a9 Ulrich Buckenlei | XR Stager Online Magazine | VISORIC GmbH
\n<\/sup><\/p>\n\n
Robot Instructions That Can Be Rotated, Disassembled and Checked<\/h2>\n