In industrial settings, procedural knowledge is too often locked away in static manuals, scattered notes, or the heads of experienced colleagues. When an operator needs to carry out a maintenance intervention or a safety procedure, finding the right step at the right moment can be slow, error-prone, and hard to trace.

Solution 6A (S6A) addresses this challenge by providing a conversational chatbot that lets users consume procedural knowledge through natural language. Instead of paging through documentation, operators, technicians, and engineers can interact with an assistant that guides them through procedures step by step, answers questions along the way, and records what was done for later traceability.

Developed by Onlim on its enterprise Conversational AI platform and deployed in the cloud, S6A is the user-facing entry point of the PERKS exploitation layer. It connects directly to the procedural knowledge stored and validated by the rest of the PERKS solutions so that every answer it gives is grounded in approved, structured knowledge rather than in a free-floating language model.

How S6A Works

At its core, S6A combines structured, transparent procedure execution with the flexibility of large language models (LLMs). This lets users follow well-defined workflows while still being able to ask questions, clarify a step, or explore related information in their own words.

  • An agentic design. A central, LLM-based Conductor Agent handles each user request and reasons over the current conversation context to decide which tools to use. These tools may include querying procedures from the knowledge graph, calling external APIs, answering questions from company documentation, or responding directly from the ongoing dialogue. This allows a seamless transition between structured procedure execution and open-ended question answering within a single conversation.
  • Grounded, explainable answers. S6A retrieves procedural knowledge through a knowledge-graph-aware Retrieval-Augmented Generation (RAG) approach, combining access to the procedural knowledge graph with document sources such as user manuals. Because answers are anchored in retrieved, approved procedures, the assistant stays useful and reliable while avoiding hallucinations.
  • Step-by-step execution with context. The assistant drives the user through a procedure, displaying the steps as structured cards, while keeping track of the workflow state. Users can pause to ask “why is this necessary?”, explore alternatives, and then continue exactly where they left off — which matches how procedures are really followed in industrial environments, where interruptions are common.
  • Configurable in natural language. Each use case can tailor the assistant’s behaviour without complex technical setup. Use case owners enable and configure tools, follow-up actions, and high-level behaviour (goals, answering style, degree of formality) using natural language alone, giving every deployment its own tailored experience.
  • Beyond the chat window. The same agentic reasoning layer is independent of the interface. With Agentic Voice, a Realtime Agent lets users interact by speaking — asking “What do I do next?” out loud — while the Conductor Agent continues to orchestrate the procedural logic in the background. This lowers friction for technicians and field workers who cannot easily type, and it supports multiple languages out of the box.

Why S6A Is a Key Building Block

S6A is where procedural knowledge created, managed, and validated across PERKS finally reaches the people who need it. It closes the loop between the back-end solutions and day-to-day operational work:

  • S5 (PKMS) stores procedures as RDF described with the Procedural Knowledge Ontology (PKO) → S6A retrieves and executes them in conversation.
  • S6B generates questions and answers from documents and knowledge graphs → these feed S6A’s question-answering and intent configuration.
  • S6A logs every conversation and produces execution traces → S4 records execution, feedback, and error reports for auditing and traceability.
  • Feedback collected during execution via S6A → reviewed by authorised users through S3, which can trigger updates to the underlying procedures.

In this way, S6A makes procedural knowledge not only accessible and intuitive, but also explainable and transparently recorded — supporting trustworthy execution and continuous improvement.

Learn More on S6A

Development Team

The PERKS Solution S6A has been designed and developed by the Onlim team: Umutcan Serles, Ioan Toma, and colleagues.