Munich, Germany is at the center of ASMPT’s latest push to bring AI deeper into electronics manufacturing support. The company is expanding its Virtual Assist software with AI-powered spare-part identification, targeting a very practical problem on the factory floor: finding the right information quickly when a machine needs maintenance.
Factory technicians often work with large equipment catalogs, service documents, troubleshooting instructions and part numbers. The information may exist, but finding the correct answer under production pressure can take longer than the repair itself. ASMPT’s approach is to let technicians interact with a knowledge system more like they would ask an experienced colleague.
AI aimed at the maintenance workflow
Virtual Assist is built around voice-based and conversational access to manufacturing knowledge. ASMPT describes the system as a toolkit for technicians and engineers that can search documents, tutorials, troubleshooting guides, service reports, web pages and videos.
The new spare-part identification capability adds another practical layer. Instead of forcing a technician to know the exact terminology or internal part number, the system can help connect the physical maintenance problem with the relevant component information.
Why identification matters
Spare parts are a deceptively difficult data problem. A single production line can contain thousands of components, and the same machine family may have multiple revisions. A technician may know what a component looks like or where it sits in a machine without knowing the exact catalog identifier.
AI can help bridge that gap by combining natural-language descriptions with structured product information. The goal is not simply to produce a plausible answer. The system has to retrieve the right technical record, because an incorrect replacement can create another failure or extend downtime.
- Natural-language questions reduce the need to memorize internal terminology.
- Search can combine manuals, service reports and other technical documents.
- AI assistance can shorten the path from fault identification to the correct part record.
- Technicians can use the same knowledge system for troubleshooting and learning.
This type of application is different from a general chatbot. The value comes from the quality of the underlying knowledge base. The system has to understand machine documentation, component relationships and maintenance context. If those records are incomplete or outdated, a more fluent AI response does not solve the operational problem.
A factory knowledge layer
ASMPT’s wider Virtual Assist system is designed as a centralized knowledge platform. That is significant because industrial companies often accumulate technical information across many systems. Manuals may live in document repositories, service reports may sit in databases and experienced technicians may hold knowledge that was never formally documented.
A conversational interface can become a common access layer across those sources. Instead of asking a technician to know which database contains the answer, the assistant can search across the available information and return a focused result.
That also creates an opportunity for continuous improvement. ASMPT says its AI learns from interactions and that manufacturer-independent information can be integrated. In a mature deployment, recurring questions can reveal documentation gaps, confusing procedures or parts that are frequently misidentified.
The economics of downtime
Manufacturing equipment has a different tolerance for delay than consumer software. If a production machine stops, every minute can have a measurable cost. A support tool So does not need to perform spectacular reasoning to be valuable. It needs to reduce the time between a fault appearing and a technician taking the correct action.
That makes spare-part identification a good target for AI. The workflow is repetitive, information-heavy and dependent on connecting multiple pieces of technical context. Those characteristics are well suited to search and retrieval systems augmented with language models.
Where the technology still needs discipline
Industrial AI cannot rely on confident language alone. A maintenance recommendation needs traceable technical information and appropriate validation. A technician should be able to see which documentation supports an answer and confirm that the proposed component matches the machine configuration.
The same principle applies to voice interfaces. Hands-free access is useful when a technician is working on equipment, but spoken instructions can be misheard. Critical procedures should remain easy to verify, especially when they involve electrical systems, moving machinery or safety interlocks.
ASMPT’s direction is So less about replacing technicians and more about reducing the information burden around them. The experienced engineer remains important, but the software can make years of accumulated documentation easier to access.
The broader manufacturing trend is clear. AI is moving from office productivity into operational environments where the value of a system is measured by response time, accuracy and reduced downtime. Spare-part identification may sound narrow compared with a general-purpose AI assistant, but narrow workflows are often where enterprise AI can provide its most measurable results.
From manuals to an operational assistant
The more interesting part of the Virtual Assist approach is the attempt to turn scattered technical documentation into something technicians can use during an actual repair. Industrial organizations often have years of accumulated knowledge, but that knowledge is not always organized around the question a technician is asking at the moment a machine stops.
A conversational search layer can change that relationship. Instead of searching several portals for a part number, a technician can describe the problem and work backward toward the relevant documentation and component. The system can also help newer employees access knowledge that previously depended heavily on experienced staff being available.
That does not remove the need for engineering judgment. A part identification result still needs to be checked against the machine configuration, revision and maintenance procedure. The value is in reducing the time spent finding information, not in replacing the technician’s responsibility for the final action.
For manufacturers, that distinction matters. The most useful industrial AI products are often narrow systems tied to measurable operational outcomes. If an assistant can consistently shorten troubleshooting and parts-identification time, the benefit can be observed directly in maintenance workflows. That is a more concrete test of enterprise AI than a generic chatbot demonstration.