PERKS provides a chatbot solution that helps stakeholders in manufacturing exploit procedural knowledge in an efficient and traceable way. Instead of relying on static documents or scattered instructions, users can interact with a conversational assistant that guides them through procedures in a structured, explainable, and reliable manner.

The chatbot brings two worlds together: on the one hand, it enables structured, transparent conversations that support trustworthy execution of procedures; on the other, it leverages large language models (LLMs) to offer flexible, natural-language interaction. This combination allows users to follow well-defined workflows while still being able to ask questions, clarify individual steps, or explore related information in their own words.

The user may ask a question during the execution and then continue the execution seamlessly.

The user may ask a question during the execution and then continue the execution seamlessly.

.

At its core, the bot uses an agentic approach. A central LLM-based Conductor Agent handles user requests and reasons over the current conversation context to determine which tools to use in order to deliver the best possible result. These tools may include querying data from the knowledge graph, calling external APIs, answering questions based on company documentation, or generating a response directly from the ongoing dialogue.

Use case owners can enable and configure available tools from the agent workbench.
Use case owners can enable and configure available tools from the agent workbench.
.
This agentic design enables a seamless transition between structured procedure execution and open-ended question answering, with multilingual support available out of the box. As a result, PERKS can be deployed across different languages and regions without additional engineering effort.

Each use case can customize the bot’s behaviour using natural language alone—both at a high level (for example, defining its goals, answering style, or degree of formality) and at the level of individual tools (such as specifying constraints or follow-up actions). This gives each deployment the opportunity to create its own tailored user experience without complex technical configuration.

Follow-up actions can be configured with natural language by the use case owners.
Follow-up actions can be configured with natural language by the use case owners.
.
After a procedure is completed, the bot selects the correct follow-up action based on the tool configuration on the fly.
After a procedure is completed, the bot selects the correct follow-up action based on the tool configuration on the fly.
.

The bot is closely integrated with the overall PERKS solution. It connects to the Procedural Knowledge Management System (PKMS) to obtain procedures represented in RDF and described with the Procedural Knowledge Ontology and its extensions. At the same time, it integrates with the auditing module by providing execution traces after every interaction with a procedure. In this way, all actions are not only guided and explainable, but also transparently recorded for traceability and compliance.