PromenautGlossary
The words we use, defined
Clean definitions of the terms behind a governed digital workforce. Where a definition says what the platform does, it rests on a capability in our catalogue.
- Governed digital workforce
- An estate of AI agents an organisation runs as accountable workers: composed into teams and workflows, acting on people’s authority, and governed as they run. The category Promenaut defines.
- Digital workforce
- The agents, teams and workflows an organisation puts to work alongside its people.
- Agent
- Software that does one defined job, using a model where judgement is needed and plain code where it is not; versioned, with a declared contract for what it reads and writes.
- Team
- A group of specialised agents working together on one stage of the work.
- Workflow
- Teams composed, with declared inputs and outputs, to do a whole piece of work that a person starts and follows.
- Module
- An installable package of agents, teams and workflows for one kind of work, which an organisation installs, changes or builds itself; the workforce ships as a growing library of them, and clients add their own.
- Deputy
- An agent acting on the authority of the person who launched the work, within that person’s permissions. On Promenaut, it is how an agent reaches knowledge.
- Governed by construction
- Designed so that control and evidence are produced by the work itself, rather than reviewed or reported after the fact.
- Maker and checker
- The person who launched a change cannot approve it: in a domain that requires approval, an agent’s change to knowledge waits for a different person.
- Approval gate
- A point in a workflow where work is held for a person’s decision, approved or rejected, with the decision kept.
- Containment
- Stopping or pausing work cleanly: a running workflow can be paused, stopped or cancelled, and an agent paused so it is refused at its next dispatch.
- Model-risk evidence
- What a run leaves for the second line: a verdict on whether the agents that ran matched those it launched with, and a check of what each agent declared against what it wrote.
- No lock-in
- Neutral to the model and to where it runs: the workforce can run in your own cloud account, on the model providers you choose, and your module code stays yours.
