Until yesterday, artificial intelligence in the factory was a silent spectator: it analyzed data, generated charts, and suggested optimizations. Today, artificial intelligence no longer just observes, it takes action
Zerynth
AI MATTERS
Zerynth is an Italian deep-tech company that develops an IIoT platform for the rapid and scalable digitalization of industrial production. The company is currently evolving its solution into an Industrial AI Copilot platform, equipped with a conversational agent that enables natural interaction with production data and processes. The development of this artificial intelligence component is considered strategic for the evolution of the product and the company’s future positioning.
Configuring an industrial monitoring system or changing production parameters often requires specific technical skills and time spent navigating complex software. This “barrier to entry” slows down operations and increases the risk of manual errors.
The challenge of our experimentation was clear: to make the factory not only reactive, but self-configurable through natural language.
For this project, we developed a modular architecture based on Agentic AI. Unlike a standard chatbot, this system is equipped with “digital hands” (through APIs and MCP protocols) that allow it to interact directly with the Zerynth platform.
The solution is built on three technological pillars:
Agent orchestration: Use of the Agno framework to manage specialized agents handling different tasks, from cost analysis to device configuration
Cloud-to-edge integration: Through Zerynth Jobs APIs, text input is translated into a physical command sent directly to the IoT gateway on the machine.
Rules engine automation: The AI can autonomously generate alarm and automation rules, translating an operator’s request into the correct technical
Interacting with heavy machinery requires extreme caution. For this reason, the system incorporates AI Safety and Explainable AI (XAI) principles. The agent never executes a critical command autonomously: the system preconfigures the action and always requires explicit user confirmation (human-in-the-loop).
In addition, every action is accompanied by an explanation of why the system is suggesting it, significantly reducing operational risk.
Validation in demo environments and at the MADE competence center produced encouraging results, but also highlighted several challenges for the future:
● Context precision: The AI performs well in understanding temporal requests (“this week”) and complex filters (“only the presses on line A”).
● Manual digitalization: Manuals rich in images require advanced preprocessing to avoid losing critical details during conversion into text for the RAG system.
● Name mapping: It is essential that the AI understands that “the large press” corresponds to the technical ID “PR-094”. This semantic mapping work is an area where we continue to invest.
The success of this phase of the AI-MATTERS project opens the door to a factory where:
Efficiency is just a voice command away: Less time spent on configuration, more time dedicated to strategy.
Technology becomes inclusive: Even personnel with limited software expertise can safely manage complex configurations.
Interaction becomes proactive: The system does not simply wait for commands—it suggests parameter changes based on energy costs or production workloads.
The collaboration between AI-MATTERS and Zerynth shows that conversational control is the key to safe and democratic digitalization. The factory of the future will not be controlled with code, but through dialogue.
