AI Opportunity Assessment in Technical Design Work
A technical design consultancy wanted to understand where artificial intelligence could genuinely support everyday work, reduce manual checking and improve the findability of information. The goal was not to adopt AI for the sake of technology, but to identify concrete use cases where AI could create measurable value for expert work, quality assurance and smoother project delivery.
The work focused especially on three areas: checking technical drawings, supporting the early stages of automation and PLC programming, and making better use of internal documentation through AI-assisted information retrieval.
At the starting point, the work involved many repetitive checking phases, scattered documentation and expert knowledge that was difficult to access quickly in everyday situations. At the same time, it was important to assess realistically where AI could work reliably and where better data, documentation and process management would first be needed.
The work helped turn an AI-related idea into a concrete and assessable development path. The customer received a clear view of where AI can create value, under what conditions it should be piloted, and how solutions can be built with cybersecurity, access rights and practical workflows in mind.
Key areas
- Identifying AI use cases
- Mapping current work phases
- Assessing the quality of data and documentation
- Exploring opportunities for checking technical drawings
- AI-assisted information retrieval
- A phased AI roadmap
What was done
Tyylidata carried out a preliminary AI assessment and roadmap project, mapping current work phases, data quality, the system environment and potential AI use cases.
The work assessed, among other things:
- where AI could speed up checking and reduce manual work
- what kind of data and documentation an AI solution would need in order to work reliably
- how the technical solution could be built securely in a Microsoft and Azure environment
- what kind of Proof of Concept experiments would be sensible before full implementation
- how cybersecurity, access rights and documentation boundaries should be considered already during the planning phase
- where processes should first be clarified before AI is used more broadly
During the work, it became clear that using AI in technical design is not only a technical question. It is equally connected to the structure of information, the quality of documentation, the definition of responsibilities, and how well the organisation’s experts are involved in evaluating AI-generated suggestions.
Technical approach
The preliminary assessment examined an Azure-based implementation model where different AI and data services would support different use cases. Possible components included document and table interpretation, computer vision-based symbol recognition, natural language models, semantic search and cloud-based data storage.
An important part of the work was assessing how raw data should be transformed into a usable format. Technical design material is not always easy for AI to interpret directly: file formats, tables, symbols, item names and project-specific practices can vary significantly. For this reason, the work emphasised data preparation, standardisation and the selection of suitable material for test use.
At the same time, the work assessed how an AI solution could operate securely so that access rights, source documents, internal guidelines and technical information remain under control. An AI solution should not become an isolated tool. It needs to connect with the existing system environment, information management and cybersecurity practices.
- AI use cases were identified from everyday work phases
- Data preparation and standardisation were assessed for AI use
- Technical drawing checks were evaluated from an AI perspective
- Cybersecurity and access rights were considered from the beginning of the planning process
Outcome
The project resulted in a clear AI roadmap and recommendations for further development. The work identified the most promising use cases, possible technical implementation options, data requirements, risks and a phased implementation model.
The assessment gave the customer a better overall view of where AI is suitable, where it is not yet suitable, and what should be done before actual implementation. It also created a foundation for a follow-up pilot, where AI can be tested in a controlled way using real use cases, limited data and an architecture adapted to the existing Microsoft environment.
The work also helped identify that successful AI adoption requires not only a technical solution, but also clear responsibilities, shared operating models and consistent information management.