Glossary
Digital Twin
A living virtual copy of something real. In engineering it mirrors a machine or a building through sensor data; in AI it is a conversational agent that answers the way a specific expert, brand or business would.
General definition
A Digital Twin is a virtual model that stands in for a real-world counterpart and is continuously updated with data from it. The idea comes from manufacturing and aerospace: a jet engine, a wind turbine or a factory line gets a software replica fed by its sensors, so engineers can monitor its state, predict failures and test changes on the twin before touching the physical asset. Building management, logistics and healthcare (organ and patient models) adopted the same pattern.
In the AI era the term gained a second meaning: a digital twin of a person, team or organisation, built as a conversational agent. Instead of sensor feeds, the twin is fed the entity’s own material: writing, talks, documentation, product catalogues, support history. A language model then answers questions in that voice and within that knowledge, so a customer can ask the twin of a consultancy what it would recommend, or an employee can query the twin of a retired expert.
- Knowledge: the twin’s answers are grounded in a curated corpus through retrieval-augmented generation, not the model’s general training
- Persona: a system prompt sets the tone, the boundaries and the point of view the twin represents
- Interface: a chat widget, a messaging channel or a voice front end where people talk to it
- Freshness: the corpus is re-indexed as the source material changes, which is the AI equivalent of a sensor feed
The two meanings share a principle: the twin is only as useful as the data link to the original. A conversational twin built on a stale or thin corpus will confidently misrepresent the person it stands for, which is why grounding, citations and an explicit persona matter more here than raw model quality. A digital twin is a kind of AI agent, but one whose identity is the point.
In the Ethora ecosystem
Ethora Agents (agents.ethora.com) is the Ethora showcase for conversational digital twins and bots built on the AI SDK. A twin is an agent with three ingredients: a knowledge base assembled from a website crawl, uploaded files and live web indexing; a persona system prompt that fixes who it speaks as and what it will not answer; and a surface, usually the embeddable assistant widget dropped onto a site with a script tag, or a room in a chat app.
Because twins run on the same platform as the chat product, they can sit in group rooms alongside people, reply to each other in multi-agent rooms with an autonomous-turn cap, remember the room transcript, and be pointed at any model, including a self-hosted one on a dedicated deployment. That last point matters when the twin represents a regulated business and its corpus contains material that cannot go to a third-party model API.
