“We have learned to structure our documentation to be equally clear for both people and machines – strengthening our organizational knowledge in the process.” (Photo: © K-Solutions)
May 2026
K-solutions – the IT company of the controller manufacturer Kollmorgen – has developed its own chatbot, based on a large LLM (“Large Language Model”). The aim is to cut reaction times for customers.
At the same time, the chatbot is to provide technically well-founded and comprehensive answers, unlike standard rule-based approaches. Read about the company's experience with this and its recommen-dations.
By Philipp Brüssler
The digital transformation has a firm hold over lift building. But the sector is not only interested in increased networking of lifts. Today, we are living through an unparalleled acceleration in development cycles – not least because their rate of turnover has soared due to the latest AI tools.
As a result, technical documentation is increasing by leaps and bounds and changing at ever shorter intervals. In When combined with growing system complexity and the prevailing shortage of skilled labour, a flood of information has arisen that conventional methods can scarcely handle. This is where a new generation of assistance systems comes in: LLM-based service chatbots that serve as intelligent guides through the digital knowledge thicket.
The friendly chat interface rests on an elaborate architecture. The system operates like a digital colleague who never sleeps. To do so, the artificial intelligence (AI) accesses two decisive knowledge levels: while the general documentation forms the foundation, the real-time connection to specific lift data provides the decisive edge.
As soon as the technician has identified the lift, the AI “reads” the precise circuit diagrams and wiring lists of this particular lift.
What is particularly fascinating is that if the controller is online, the bot analyses the current event memory directly. It not only perceives how the lift is supposed to work in theory but also knows what it needs at that moment.
Although originally developed for our customers and their technicians, the bot has now become an indispensable internal tool. It acts as the first line of defence for our own technical support. The advantages are obvious:
• Immediate response time: the customer or technician gets the solution without delay the moment he is in front of the lift.
• Pressure off the back-office: routine work is automated which enables our experienced support engineers to concentrate on the really tricky special cases.
• Quality boost: troubleshooting is more precise since the AI discovers connections between thousands of pages of documentation faster than any person.
A fascinating side effect emerged during the intensive test phases: the AI performed as an unsparing test for the quality of our own technical documentation.
AI is only as precise as the data with which it is supplied: whenever the bot faltered or misunderstood instructions, we ran into unclear formulations or logical gaps in our manuals.
We have learnt to prepare documentation to ensure that it is equally clear to people and machines. The bot helped us to sharpen our company knowledge.
Innovation requires a clear framework. We have consistently aligned the system with the requirements of the EU AI Act. The bot belongs in the area of limited risks in terms of risk classification.
It does not make any autonomous security decisions but rather acts as a pure assistance system. Thanks to permanent transparency information, users know at all times that they are communicating with an AI and confirm this understanding before first use.
The AI service bot is not a replacement for experienced technicians but rather a powerful tool, whichg bridges the gap between complex technology and fast on-site service. The ability to “understand” complex logical diagrams and interpret event data in real time means we are headed for an era where downtimes will be minimised by immediately available knowledge.
The author is Chief Technical Officer at K-solutions.
If you want to set up your own service bot
• Data hygiene is essential: AI is not magic but instead reflects your technical documentation. Use the introductory phase for a “docu check”. What AI fails to understand due to contradictions also leads to errors by technicians on site. Clean sources are the foundation for precise answers.
• Context creates added value: Pure “handbook chatter” is insufficient. The real payoff only occurs once the AI is familiar with the specific context of the lift. The link to live data, such as the event memory, transforms a generic tool into a genuine expert assistant.
• Transparency instead of flying blind: AI models can assist but cannot (yet) assume the final responsibility for safety-relevant decisions in lift building. Integrate clear guide rails and notice systems from the outset. This boosts users’ trust and ensures that the expert on site always has the last word.
More informations: k-solutions.biz
Digitalpaper
Write a comment