Definitions
The terms that come up when we talk about sovereign, purpose-built AI, defined plainly.
- Sovereign AI
- AI where you hold jurisdiction and ownership, not merely a location. You own the system and it runs on infrastructure you control, and inference runs under European jurisdiction, so no foreign authority can compel a provider to hand over your data. It does not mean nothing ever leaves your network. It means no US entity is ever in a position to be compelled.
- Purpose-built AI agent
- An AI agent built and tuned for one specific job in a company’s workflow, on that company’s own data. The opposite of a configurable platform: instead of a toolkit you model your work into, you deploy the specific agents your workflow needs, where each agent is both the logic and the interface.
- Agentic AI
- A system that takes actions to finish multi-step work rather than only answering questions. It can plan, use tools, read and write data, and carry a task through to a result. The shift from a chatbot to an agentic system is the shift from "it tells you" to "it does it".
- Digital colleague
- A plain-language name for a purpose-built AI agent: a specialist for one workflow that has read the relevant systems and does the digital work, while a person steers and decides. Not a website chatbot, and not a replacement for the people who do the work.
- Forward deployed engineer (FDE)
- An engineer who works embedded inside a customer’s company rather than at arm’s length: sitting with the people who do the work, learning how it really runs, then building and deploying the software on the customer’s own infrastructure and staying until it holds up in production. The role grew at data and AI companies, where a product can demo well and still be hard to make work inside a real business. At Cultivator AI it is how we build: forward deployed engineers on the factory floor, turning one real process into a working swarm.
- On-premise AI
- AI that runs on infrastructure the customer owns or controls rather than a vendor’s SaaS cloud. At Cultivator AI the agents, containers and data run on the customer’s own infrastructure; model inference runs on European providers by default, or on the customer’s own hardware where they have it.
- US CLOUD Act
- A US law under which US authorities can compel a US provider to produce data it holds anywhere, including in an "EU region". It is the reason an EU region on a US cloud is not the same as European jurisdiction: the question turns on the provider’s nationality, not the data’s postcode.
A term we did not cover?
The FAQ answers the questions manufacturers actually ask, and we are happy to answer anything else directly.