CADEXA
The ZIM cooperation project CADEXA - CAD Data Extraction and Automation Agent is the start of a joint research and development project between agitum GmbH and the CIMTT Center of Production Engineering and Management at TU Dresden. The aim of the project is to develop an AI-based automation agent that intelligently evaluates technical drawings, provides relevant information in a structured manner, transfers it to existing ERP and PDM systems and makes it accessible to users via a dialog-capable interface.
The project kick-off took place on 02.04.2026. This marked the official start of the joint work process between the project partners. CADEXA is designed as a ZIM cooperation project and combines technological development expertise with ergonomic research into the human-centered design of digital assistance systems. Further information, updates and project contents will be accessible via our official project website: CADEXA
Initial situation
Technical drawings are a central source of information in industrial companies, especially in mechanical and plant engineering. They contain dimensions, tolerances, material specifications, parts list information, symbols and other technical features that are relevant for downstream processes in design, production planning, purchasing, documentation or logistics.
In operational practice, however, this information is often still checked manually, transferred and updated in ERP or PDM systems. This results in media discontinuities between CAD documents, production-related databases and business management systems. These manual activities tie up skilled workers, increase the susceptibility to errors and make end-to-end digital information processing more difficult.
At the same time, the complexity of technical documentation and information flows is increasing. CAD drawings vary considerably depending on the company, design method, software environment and individual conventions. Standardized solutions quickly reach their limits here. It is therefore not enough to simply extract technical information automatically. It must also be provided in a form that suits the respective usage situation.
Objective of the project
CADEXA aims to develop an adaptive automation agent that analyzes technical drawings multimodally and converts relevant content into structured data formats. This information is then to be automatically transferred to existing ERP and PDM systems or made available for further queries.
The planned Automation Agent combines three central development modules:
Firstly, an AI-supported extraction module is being developed that recognizes the textual and visual content of technical drawings. This includes dimensions, tables, symbols, material information and other drawing-related data.
Secondly, the extracted information is integrated across systems. As small and medium-sized companies in particular often work with heterogeneous IT landscapes and do not always have standardized interfaces, RPA-based automation solutions are considered for transfer to ERP and PDM systems.
Thirdly, a dialog-capable user interface is being developed that allows users to query information in natural language. For example, it should be possible to check whether a component has already been created in the target system or which technical features have been extracted from a drawing.
CIMTT's contribution
In CADEXA, CIMTT is contributing its expertise in ergonomics to the development of context-sensitive information provision. The focus is on the question of how technical information can be prepared and output in such a way that it is relevant, comprehensible and action-supporting in the specific work process.
Context is not understood as a single influencing factor. Instead, a context model is developed that describes how the relevance of information results from the interplay of person, task and situation. For example, the decisive factors for the provision of information are whether a person from the design, production planning or purchasing departments is working with the system, which task is currently being processed and the situational conditions under which it is being used.
A technical drawing contains different relevant information for different people and work steps. While detailed dimensions, tolerances and change information may be the main focus in design, material details, cost indicators or delivery relevance are more important for purchasing or quotation processes. Under time pressure, a reduced, prioritized presentation can also be useful, while a more in-depth presentation of information is required in a detailed technical review.
CIMTT is developing concepts for context modeling, situation classification and adaptive information output based on ergonomics. The aim is an output logic that does not present technical information statically, but prioritizes, reduces, highlights or makes it available dialogically depending on the context of use. In addition, a user-centered evaluation is planned in order to test the practicability, comprehensibility and support effect of the developed solution.
Cooperation between the project partners
The project partners bring complementary expertise to the project. agitum GmbH is responsible for the technological development of AI-supported information extraction, RPA-based integration and the prototype system architecture. CIMTT complements this technical development with the ergonomic perspective on use, context, information relevance and human-centered system design.
It is precisely this connection that is central to CADEXA. A technically functioning system is not yet an effective assistance system. Only when information can be meaningfully provided, understood and used in the work process does real added value arise for companies and employees. CADEXA therefore addresses not only the automation of technical data flows, but also the question of how digital assistance systems can be designed in a human-centered way in real industrial work contexts.
Importance for medium-sized industrial companies
CADEXA is aimed in particular at small and medium-sized companies that want to make technical documentation and information processes more efficient. The automated extraction and structured transfer of technical drawing information should reduce manual transfer efforts, improve data quality and accelerate subsequent processes.
In addition, the context-sensitive provision of technical information can help to relieve employees in a targeted manner. Relevant information should be easier to find, more comprehensible and better adapted to specific tasks. CADEXA therefore not only supports the digitalization of technical processes, but also the human-centric design of AI-supported work systems.
For CIMTT, the project ties in with existing core research areas in the fields of knowledge management, work design, digitalization, AI applications and transfer to small and medium-sized enterprises. CADEXA also offers a concrete use case for the further development of labor science research on context-sensitive information provision in industrial production and development processes.
Funding
CADEXA is being implemented as a cooperation project as part of the Central Innovation Program for SMEs (ZIM). The project combines the development of innovative AI and automation technologies with ergonomic research into the context-sensitive and human-centered design of digital assistance systems.