OEE Booster in Practice: How AI Supports Maintenance and Production

Why Is Access to Knowledge Still a Challenge in Manufacturing?

Machine failures cannot be eliminated completely. However, a significant share of the time lost after a production line stops is caused not by the failure itself, but by the search for the information needed to diagnose it. The operating manual may be stored in one place, the history of previous failures in another, process data in a separate system, while some practical knowledge exists only in the experience of individual employees. The problem becomes particularly apparent when the most experienced person is unavailable. The operator needs to determine what they can do on their own, when maintenance should be called, and where to find the correct procedure.

This is exactly the problem OEE Booster is designed to address.

OEE Booster Does Not Replace Existing Systems

One of the key principles presented during the webinar is to use the existing IT and OT environment rather than build another isolated system. ERP, CMMS, measurement systems, technical documentation, failure histories, and other knowledge sources continue to serve their existing purposes. OEE Booster adds a layer on top of them that makes it possible to search, analyze, and present information in a form tailored to a specific user. This matters because a machine operator needs a different answer than a production manager.

An operator should receive a short, step-by-step instruction along with clear information about which actions they are authorized to perform. A maintenance technician needs more technical detail and information about safety requirements. A production manager, on the other hand, is more interested in trends, the impact of a disruption on throughput, or information related to OEE.

A single knowledge base can therefore support different roles, while presenting information in a way that matches their responsibilities.

How Does OEE Booster Use RAG?

During the webinar, we also demonstrated the difference between a general-purpose GenAI tool and a solution that works with an organization’s own knowledge.
OEE Booster uses a RAG approach, or Retrieval-Augmented Generation. After a user submits a question, the system retrieves relevant sections from available sources and only then provides that context to the language model. From the user’s perspective, this primarily means being able to work with specific company documents and data. The answers can also reference the sources on which they were based.

Troubleshooting Machine Failures on the Plant Floor

The first demonstration scenario involved a robot operator. A robot operating on the production line stopped, and a yellow LED ring at its base began flashing. The operator wanted to determine what the signal meant and which actions could be taken before escalating the issue. OEE Booster searched the available documentation and presented an answer based on the predefined rules for the operator role: what the status meant, which actions the operator could perform independently, and when maintenance needed to be involved.

The same mechanism can be extended to additional sources, such as the history of similar failures, internal procedures, or health and safety documentation. The better and more up-to-date the context provided to the assistant, the more useful its response can be.

Support for Maintenance Technicians and Managers

For a maintenance technician, OEE Booster can organize the troubleshooting procedure while taking into account safety requirements and technical data available in the documentation. A maintenance manager can use the solution to work with a broader range of information. During the demonstration, the system was asked, among other things, to prepare an overview of spare parts and categorize them by criticality. In an environment integrated with ERP and CMMS systems, this type of analysis can be enhanced with actual inventory levels, event histories, or replacement frequency. Another example involved creating a monthly schedule of short maintenance tasks for the maintenance team and preparing a checklist for use on the plant floor.
These tasks do not require replacing a specialist’s expertise. AI can, however, reduce the time spent searching for information, organizing data, and preparing documentation.

Faster Employee Onboarding

Distributed knowledge affects more than just troubleshooting. It can also make onboarding new employees more difficult. During the webinar, OEE Booster was asked to prepare a one-page guide for the safe operation of a robot, along with a knowledge test for a new operator. The material was to be created in Polish even though the source documentation was available in another language. This approach can be useful in manufacturing facilities with international teams. Instead of preparing every version of a training document from scratch, existing documentation can serve as the source while the way the knowledge is presented can be adapted to the intended audience.

From PLC Data to Production Management Analysis

Production managers have different needs. Simply knowing what an error code means is usually not enough. What matters are trends, the frequency of disruptions, availability, performance, and quality losses, and their impact on the overall process. In one example presented during the webinar, OEE Booster analyzed data related to previous events and changes in robot parameters. The goal was to identify the direct cause of the stoppage, show the trend leading up to the failure, and present the service intervention. This is where proper data preparation becomes particularly important. Raw values from PLCs should not simply be passed directly to a language model. They first need to be collected, structured, and connected with the relevant business context. Only then can GenAI support analysis in a way that is useful to the person responsible for production.

Integrating Knowledge Across IT and OT

OEE Booster can work with documents and data from a variety of sources. During the webinar, we also demonstrated how it can connect with systems already used within an organization. For operational-layer data, VM.PL also uses solutions that enable communication with PLCs, including through OPC UA. Ultimately, the value does not come from the AI model alone. It comes from combining technical knowledge, documentation, machine data, and business information into a consistent context.

One Technology, Multiple Manufacturing Use Cases

The examples presented during the webinar highlight several areas in which an AI assistant can support a manufacturing facility:

• faster access to information during equipment failures,
• preserving the knowledge of experienced employees,
• supporting operators and technicians according to their roles and permissions,
• preparing instructions, checklists, and onboarding materials,
• analyzing event histories and OEE-related data,
• planning maintenance activities,
• reducing some repetitive administrative tasks.

OEE Booster is not intended to replace specialists or make decisions for them. Its role is to shorten the path between a question and the information needed to take the right action.

See OEE Booster in Action

In the full webinar recording, we demonstrate how assistants can be configured for different roles and present several specific production and maintenance scenarios.

If you would like to see how OEE Booster works with technical documentation, production data, and different user roles, watch the webinar: https://www.youtube.com/watch?v=tEwv5si7nY8