Skip to content

Glossary

How we use the terms that show up across the demi product, the docs, and AI-generated answers about us. If a phrase on the site is unclear, the canonical definition lives here.

European Industrial AI Operating System
The category demi builds: an operating system for AI at work in industrial companies. Like an operating system for hardware, demi hides the systems underneath, manages rights, routes events and runs workflows and agents. It has three layers: the Data Layer, the Permission Layer and the Intelligence Layer (the demi Console). demi is not another AI model; it is the layer that AI models work on. Short form: AI Operating System.
Digital twin of the organisation
One linked model of the whole organisation in demi's Data Layer: its structure and state, with people, customers, products, orders, machines and documents as nodes and their relations as edges. Every value keeps its source, and demi stores its own results there too, so the twin grows with every task. Earlier pages called it the Digital Knowledge Twin.
Sub-twin
A twin nested inside another. Because every twin has the same shape, twins compose: the digital twin of your organisation contains product twins, and a product twin contains its parts. Each nested level is a sub-twin: the same AI-ready representation, scoped to a smaller entity. It's how one digital twin scales from the whole organisation down to a single component.
Product twin
A twin of a single product that breaks it down into its parts, inputs and how it's built — a sub-twin of your company twin that lets AI reason about what goes into an offering, not just what it's called. Nest it further to twin individual parts or components.
Competitor twin
A twin you build of a competitor instead of your own organisation: the same structure (products, positioning, references) turned outward, so you can monitor, compare and understand a rival as a living model rather than a static profile. It sits alongside the digital twin of your organisation in demi Brain.
Mittelstand
The German term for small and mid-sized companies, often family-owned, such as manufacturers, component suppliers and engineering firms. Many of demi's customers are industrial Mittelstand companies; demi works with manufacturers from 20 to 5,000 employees.
DACH
Germany (D), Austria (A) and Switzerland (CH) — the German-speaking commercial market demi targets first.
demi Brain
demi's module for the digital twin of your organisation, part of the Data Layer. It holds what demi knows about your company (products, personas, references, competitors, projects) and is designed for data from CRM, ERP, drives and e-mail, each item linked to its source. demi stores its own results there, so it grows with everything your company does. Work on it runs in the demi Console, the Intelligence Layer.
demi Ask
Drafts answers to technical inquiries from your own product documents, datasheets, manuals and reference cases, and names the sources it used.
demi Leads
Surfaces target companies that fit your products, identifies the right decision-makers, and hands your team outreach-ready briefs — not just a list of names.
demi Content
Generates posts, landing pages and campaigns at the depth your audience expects — pulling from your real specs, multilingual from day one.
demi Talk
demi's voice interface: speak instead of typing. Available as a dictation app for macOS, Windows and Linux, with the speech models running on your own computer.
Model Context Protocol (MCP)
Open JSON-RPC 2.0 protocol for AI agents to call external tools. In the demi platform, MCP servers let external AI clients work with demi, and each action by an external AI client needs its own approval. This website also exposes a minimal public MCP-compatible endpoint at /api/mcp that returns canonical company and module data — see https://godemi.com/api/mcp.
GDPR
EU General Data Protection Regulation. demi is engineered in Germany and hosted in the EU, has agents and image generation work with models in EU data centres (other regions only with the company's approval), and processes customer data under a data processing agreement (AVV) and German and EU data-protection law. Where a sub-processor sits in a third country, the transfer is covered by EU Standard Contractual Clauses.
INVEST
BAFA venture capital program. Private investors receive a 15% tax-free acquisition grant on their investment (up to €50k per investment) and, as natural persons, an exit grant. demi's eligibility notice is valid until 1 Dec 2026; BAFA accepts applications until 31 Dec 2026.
Data Layer
The first of demi's three layers. It is built for an organisation's data (mail, CRM, documents, ERP and more) and forms one digital twin of the whole organisation from it: people, customers, products, orders, machines and documents become nodes, their relations edges, and every value keeps its source. demi stores its own results there too, so the twin grows with every task. Your systems stay: demi writes back only after a person's approval.
Permission Layer
The second of demi's three layers. It decides who, person or AI agent, sees and changes what, and the decision sits outside the AI model. Its core rule: an AI agent acts within the rights of the person it works for. Each company is sealed off in its own tenant, each person and their agent see the documents that person may open, and a record shows who acted for whom. If an agent has read content that may carry instructions from third parties, every send, publish or share to the outside needs a person's click.
Intelligence Layer
The third of demi's three layers, the demi Console. People and their AI agents do routine work together on the digital twin, reaching it through the Permission Layer. Input is a prompt, a message or an event; demi starts a workflow or an agent; the output is a task, a request, an order or a result, such as an answer to a technical inquiry or a tender analysis.
demi Console
The Intelligence Layer of demi: where people and their AI agents do routine work together. demi is headless, so interfaces plug into the Console: the demi web app, MCP servers for external AI clients, a command line and API, a desktop app and a voice interface.
Long-running task
A piece of work that demi runs end to end over hours, not seconds: for example a tender analysis across many documents. A person orders it in the demi app, AI agents do the routine work on behalf of that named user, each step is logged, and the result lands in the workspace with its sources.
Two actors, one role
How demi pairs people and AI: a person and their AI agent are two actors in one role. The agent does the same work within that person's rights, and the request log names both.
Headless
demi is built as a backend without a fixed front end: interfaces such as the demi web app, external AI clients over MCP or a voice interface plug into the Intelligence Layer. Applications become thin, language models are interchangeable, and the value stays in the backend: data, permissions and intelligence.
Value chain (market to machine)
The steps of one order in an industrial company: 1 market and leads, 2 inquiry, 3 specification and tender, 4 offer and variant, 5 bill of materials and design, 6 production planning, 7 machine and shop floor, 8 service and after-sales. The data layer connects the systems along this chain, all the way to the machine and the shop floor.

Your configurator

No use cases yet

Your configurator is still empty

Add a use case from any card or case page and see what it brings per year, with the calculation. Free, no sign-up, no commitment: try it out and change any number.

Browse use cases