Service managers in mechanical engineering know the problem: requests come in via various channels, the assessment of urgency depends on the dispatcher’s wealth of experience, and valuable time has passed before the right person is working on the right device. Artificial intelligence can fundamentally change this process – but only if the database is right and the application fits the company’s service structure.
This article shows what AI can actually achieve in the technical service of mechanical engineering companies today, what the requirements are and what expectations are realistic.
What AI can achieve in technical service
AI in service management is not an end in itself. The useful fields of application can be narrowed down to three core areas:
Automatic classification and prioritization of requests
When a customer submits a service notification – via portal, e-mail or telephone – it must be classified: Which system? What problem? How urgent? AI can take over this step by analyzing the content of the request, comparing it with historical cases and suggesting a category and priority level.
A practical example: If an operator reports fluctuating pressure on a hydraulic pump, a trained system recognizes this as a mechanical problem with medium priority – and automatically upgrades it if the device is stored as production-critical.
Intelligent routing to the right technician
In many companies, the assignment of service orders is still manual or rule-based. AI-supported routing takes several factors into account at the same time: the technician’s qualification profile, geographical availability, current workload and the service history of the machine in question. The result is faster assignment with a higher first-time resolution rate.
Knowledge retrieval for technicians in the field
Much of the service knowledge in mechanical engineering companies is not documented in a structured way – it is in the heads of experienced employees. AI-supported knowledge systems can automatically provide relevant documentation, previous solution steps and circuit diagrams when a specific report is received. When opening the case, the technician can see what has worked for similar problems in the past.
The decisive basis: structured machine data
AI can only make decisions as well as the data on which it is based is structured and complete. In mechanical engineering service, this is primarily:
- The installed base: Which machines are at the customer’s premises, in which configuration, how long have they been in operation?
- The service history: What problems were there, how were they solved, who was responsible?
- Telemetry data: What operating parameters does the machine currently provide?
If this data is missing or distributed across different systems, every AI solution works on an insecure basis. Classifications become less accurate, routing decisions worse, knowledge retrieval incomplete.
In this context, the digital machine file is not an optional feature, but a prerequisite for the meaningful use of AI in service. It bundles machine data, service history and contract information in one place – and makes it usable for AI systems.
If you do not yet know how complete your installed base is, an installed base assessment provides a structured introduction.
From classification to decision support
The use cases described – classification, routing, knowledge retrieval – are individual functions. The next step is to combine these functions into an end-to-end decision support system: Which technician gets which assignment, based on which machine status, which knowledge and which contractual basis?
This approach is summarized under the term Service Decision Intelligence (SDI). It combines structured service knowledge, machine data and AI models to create an intelligence layer that actively supports service teams in making decisions – without taking responsibility away from them.
SDI is not a separate product that is operated in parallel to existing systems, but a layer that is based on the existing service platform. Salesforce Service Cloud forms the technological basis: cases, installation history, IoT data and dispatching rules converge on a shared database.
Knowledge management and remote support as an AI level
Two areas that are often underestimated in AI-supported service:
Structured knowledge as AI input
AI needs access to structured service knowledge in order to make meaningful suggestions for solutions. Empolis Service Express is a specialized knowledge management solution that bundles technical documentation, error codes and proven solution steps in a searchable system. Combined with AI-powered search, technicians in the field can access relevant information in seconds – without having to pore over manuals or call colleagues.
Remote support with AR support
Not every service call requires a technician on site. TeamViewer enables remote diagnostics and, where appropriate, AR-supported support: an experienced specialist sees what the technician on site sees in real time and can provide step-by-step guidance. AI can automatically suggest a remote session if the classification shows that the problem can be diagnosed remotely.
Both solutions are integrated into the Salesforce platform and work on the same database as the ticket system and the machine file.
What machine builders need to clarify before using AI
AI in service is not plug-and-play. Three prerequisites must be met before the technology can be used effectively:
1. check the database
Are machine data, service history and customer information recorded in a complete and structured manner? Fragmented or inconsistent data is the most common reason why AI projects in service fall short of expectations.
2. standardize processes
AI improves processes – it does not replace missing ones. If you don’t have clear rules for ticket prioritization and technician approval today, you need them first. AI can then automate and scale them.
3. involving employees
AI in service is changing the work of dispatchers and technicians. Anyone who introduces the system without involving the teams concerned will generate resistance – and lose the benefits that the technology could bring.
Conclusion
AI in technical service is no longer a project for the future. The use cases are clear and the technology is available. The decisive factor is not the software – it is the maturity of the database and the quality of the underlying service processes.
A step-by-step model is ideal for machine manufacturers who want to proceed step by step: First structure the installed base, then digitize service processes, then build AI-supported decision support. If you follow this path consistently, you will create the basis for a service that does not become linearly more expensive as the customer base grows – but smarter.
FAQs
When does AI start to pay off in technical service?
AI in service is not worthwhile from a certain company size, but from a certain level of maturity of the database. Anyone who has recorded their installed base in a structured manner and has at least two to three years of service history has a sensible basis. Anyone who still keeps machine data and customer data in separate systems should start there first.
Do we have to completely digitize the installed base first?
Completeness is not a prerequisite – but a reliable basis is. Even with a partially recorded installed base, the first AI functions can be used sensibly, for example for the automatic classification of requests. For more advanced applications such as predictive maintenance or AI-supported routing, a structured, complete database is crucial. An installed base assessment shows where you currently stand.
What is the difference between an AI chatbot and Service Decision Intelligence?
A chatbot answers standard queries – spare parts status, appointment booking, simple error codes. Service Decision Intelligence goes further: it combines machine data, service history and structured knowledge to provide decision support for complex service cases. Not “Which spare part?” but “Which technician, with which knowledge, for which machine, at which priority level?”
How long does the introduction take?
This depends directly on the maturity of the data. Initial results – such as automatic ticket classification – can be achieved in a few months once the database is in place and processes are standardized. Building a complete decision intelligence layer takes longer and is best done in stages: Structuring data, digitizing processes, setting up the AI layer.
Will dispatchers and technicians lose control over their work due to AI?
No – and that is an important difference in the concept. AI in the service supports decisions, but does not make them autonomously. The dispatcher sees a routing suggestion with reasons and can override it. The technician receives advice on solutions and decides for themselves. Companies that communicate this clearly to their teams experience significantly less resistance during implementation.
