Glossary

The terms you’ll run into around the app, in plain English.

Artefact

A single classified item extracted from a communication — a risk, issue, decision, action, and so on. Artefacts are what fill the registers.

Register

A list of artefacts of one kind: the risk register, issue register, decision register, action register, stakeholder register, and so on.

Register state

A decision or risk’s effective state (for example current, superseded, duplicate, retired) is derived from how it relates to other artefacts — it can differ from the raw status field. A retired item was archived as stale history (see Born-retired), which reads differently from one archived by hand.

RAG / weakest-link

The red / amber / green health colour. A project’s colour is set by its worst dimension — its weakest link — not by an average. See How health works.

Score

The 0–100 number shown next to a health colour, giving the magnitude behind the colour so two same-coloured items can still be compared.

Sentiment (trust-weighted)

The mood of communications, scored per message and rolled up to a project mood. Messages are weighted by trust rather than averaged flat, and the written summary is generated from the rollup and top rationales — never from raw message bodies.

Flow

A process signal derived from your tracker’s status history (not a copy of your tasks): rework rate, ageing tasks, cycle time, lead time.

Commitment gap

Where what’s talked about (actions), what’s worked on (tracker flow), and how people feel (sentiment) diverge for a topic — a sign that talk and reality have drifted apart.

Topic / Theme

A topic groups related artefacts; a theme groups related topics. They emerge from the content automatically rather than being tagged by hand.

Stakeholder

A person or organisation involved in the project. Stakeholders are created automatically from the senders of communications and the owners of actions.

Stakeholder–user link

The connection between a stakeholder (someone in the project record) and a user (someone who signs in). This link is what makes action items show up in that person’s My work — items are assigned via the link.

Evidence chain

On an artefact’s detail page, the trail back to the source communications it came from, including the exact excerpts that support each claim — so you can verify where a fact originated.

Held mail

Email that’s been stored but not yet classified because a budget or rate limit was hit. It’s never dropped — it’s released when the window resets or when you release it manually.

Trusted sender

A sender allowed to bypass classification limits — but only on a message that passes email authentication (SPF/DKIM), to prevent spoofing.

Module

A feature area you can enable or disable per project (under Settings → Modules). Disabling stops new processing and hides it from the sidebar but keeps existing data.

Backfill vs reconcile

A backfill replays all data to catch up (for example after connecting a source or enabling a module). A reconcile corrects drift between systems without reprocessing everything.

Born-retired

An import option that archives historical items on arrival, so old data enriches the knowledge graph without cluttering your live work surfaces.

Dedup (de-duplication)

Finding and merging artefacts that describe the same thing. Candidates can be found by vector similarity or by an LLM pass, and you confirm a merge or keep items separate.

Relation

A typed link between artefacts — supersedes, duplicate, relates, contradicts. Relations drive register state and show up as badges on items.

Confidentiality tier

A visibility level on data that controls which roles can see it. Tiers are managed under Roles & permissions.

PMI area

A grouping used in the knowledge Feed; one area can span more than one kind (for example the risk area covers both risks and issues).