Business Epic – Concept and Development of a dialogue-oriented AI mentor
The AMMANN Components division of AMMANN AG Schmiede- und Bearbeitungstechnik offers visitors, customers, and prospective clients innovative and easy access to manufacturing expertise, performance capabilities, and the company’s context.
Edison combines a web-based chat with curated corporate knowledge and deterministic context control. The goal is not to create a general-purpose chatbot, but rather a subject-matter-expert guidance partner for industrial manufacturing.
Systems and technologies used: WordPress, n8n, Mistral, Caddy, Cloudflare, HTML/JavaScript, and Context-World/DBRS principles.
Customer Satisfaction, Feedback
- In the testing and approval process
- Edison is already being used in practice to prepare for preXcon .
- Tests conducted so far show that even extensive information on manufacturing techniques, machinery, quality, and system suppliers’ expertise can be made easily accessible through dialogue.
- Further development will proceed iteratively in collaboration with AMMANN Components.
Customer Benefits and Objectives
AMMANN Components possesses extensive technical expertise and many years of experience in the manufacture of ready-to-install mechanical components, assemblies, and subsystems. However, a significant portion of this expertise is difficult to present fully and clearly on traditional websites.
Edison is designed to enable customers and prospective customers to
- To ask questions about manufacturing techniques and capabilities in natural language,
- to understand and contextualize technical concepts and performance limits,
- to go into specific details as needed,
- Combining information on CNC machining, forging, welding, quality, full-service capabilities, and expertise as a system supplier, and
- to be able to more quickly assess whether AMMANN Components might be a suitable partner for a specific project.
Edison is intended not only to provide information, but also to improve access to existing experience and corporate knowledge.
Another goal is to make the actual capabilities of AMMANN Components transparent: the knowledge that exists within the company should also be understandable and accessible to both people and AI systems.
Background
The current AMMANN Components website reflects only a portion of the company's actual knowledge and manufacturing expertise.
Technical information, in particular, is scattered across different pages, documents, and practical knowledge. This makes it difficult for visitors to piece together a complete picture of the possibilities offered by AMMANN Components from individual pieces of information.
At the same time, the project revealed that a traditional AI chatbot based on the current website does not receive enough reliable context.
To ensure implementation in the short term -particularly with regard to preXcon- a comprehensive, curated context for Edison was therefore established first. This includes, among other things, information on
- Positioning and expertise in system suppliers,
- Manufacturing techniques,
- CNC capabilities and technical specifications,
- Machinery and measuring equipment,
- Quality and process reliability,
- Full-Service Offerings,
- Supply Chain as well as
- relevant contacts, and additional sources of information.
This initially resulted in a deliberate, pragmatic interim solution while AMMANN Components continues to develop its public Context World.
Solution Concept
Edison will be integrated directly into AMMANN Components' digital communications as an AI mentor.
Users can ask Edison any questions they like, such as:
“What CNC capabilities does AMMANN Components have?”
or explore the topic further through discussion afterward:
“Explain that to me in more detail.”
To ensure that even extensive business context does not unnecessarily slow down every simple query, a two-tier context architecture was developed.
A deterministic resolver decides between a “slim” and a “fat” context before the actual AI call.
The “Slim” context is designed for simple conversational situations and general orientation. When technical terms, technical specifications, or requests for further information -such as “explain,” “more,” “in detail,” or ‘thoroughly’ -are encountered, the comprehensive “Specialized” context is automatically activated.
A shared dialogue cache also allows Edison to refer back to previous questions. This enables Edison to provide a brief answer initially and then gradually expand on the information, rather than providing all available details with every query.
The architecture thus deliberately separates
Dialogue → Contextual Decision → Domain Knowledge Base → Response Generation.
In the medium term, the context—which is currently still partially stored in the Edison Playbook—is expected to increasingly come from AMMANN Components’ structured Context World / DBRS.
Edison thus serves as a collaborative gateway to an independently maintained knowledge base that can also be used for other purposes.
The Technical Foundation: Samy @ Trusted Context World
Edison is based, both technically and methodologically, on Samy @ Trusted Context World. Samy provides the reusable foundation for dialogue-oriented AI applications, while Trusted Context World provides the structured, curated, and traceable business context.
For AMMANN Components, this framework was tailored to the specific manufacturing and business context and implemented as Edison.
Applied Methods and Tools
- SAMY
- Context World / DBRS
- Trusted-Context-Principles
- Collaborative AI Supported Engineering (CAISE)
- iterative Prototyping and Testing
- deterministic Context Resolver
- Slim-/Fat-Context-Concept
- Conversation Memory
- n8n Workflow Automation
- Mistral as LLM
- HTML/CSS/JavaScript Web-Client
- WordPress Integration
- Caddy and Cloudflare
- Joint technical review with AMMANN Components
The development process was deliberately iterative. Edison's responses were tested using real questions, and then the context, playbook, resolver, and dialogue behavior were adjusted accordingly.
Project Contribution and Role
- Design of the AI Mentor Edison
- Analysis and Organization of Available Business and Manufacturing Knowledge
- Establishing the Subject-Matter Context
- Design of the Context-World-/DBRS-Integration
- Development of the Slim/Fat Context Model
- Development of the Deterministic Resolver
- Design and Implementation of n8n Workflows
- Design of the Web-Based Edison Client
- Integration into the Web Infrastructure
- Test Design and Quality Assurance of Responses
- Innovation Engineering
- Technical Architecture
- Project Management
Final result
Edison provides AMMANN Components with a fully functional AI Mentor that makes the company's technical information accessible through interactive dialogue.
The first production-ready phase was deliberately designed so that it can already be used for preXcon, even though AMMANN Components’ future Context World has not yet been fully established.
The architecture developed for this purpose is particularly important: Edison does not have to process the entire extensive knowledge context with every query. A deterministic resolver determines when the extensive domain context is actually required. As a result, the solution combines response speed, conversational capability, and domain depth.
Edison also serves as a practical test case for the broader Context-World architecture: In the future, the information curated today could be provided from a structured knowledge base maintained independently of the chatbot and used not only for Edison but also for other applications.
Contact Information for the Reference
- Available on Request.
Related Links
- AMMANN Components: https://www.ammann-components.com/ki
- Edison – AI-Mentor: https://www.ammann-components.com/edison
- preXcon – AMMANN Components: https://prexcon.ch/de/ausstellende/16833