Document Sets
A Document Set is a named group of documents inside a project — for example, "Relied Upon Documents", "Structural Certificates", or "Fire Safety Reports". Document sets let you:
- Organise received documents into buckets instead of one undifferentiated pile.
- Drive per-document metadata (via fields linked to a document set — see below).
- Populate tables in Document Templates that repeat once per document in a set.
- Assemble Approval Packs grouped by category.
How document sets fit the three tiers
Like everything else in Project Creation Setup, document sets exist in three tiers.
| Tier | Where you configure it |
|---|---|
| Library | Project Creation Setup → Document Sets (/project-creation-setup/document-sets). Define the name and, optionally, an AI classification prompt. |
| Project Type | Add library document sets to a Project Type to ship them as a bundle. |
| Project Instance | A snapshot of the document set appears inside a project as soon as the project type is applied. From that point on, its membership is independent of the library. |
Populating a document set
Once a project has been created from a project type, each of its document sets starts empty. There are two ways to populate it, and both are available on every set — you can mix them freely on the same set.
AI classification via a prompt
Each library document set has an optional AI Prompt field. If you write a prompt on the library definition, ApprovIQ uses it to classify received documents into the set automatically. The prompt is what you use to control what the AI picks:
"Select documents the applicant cites in their declaration as relied upon. Do not include the applicant's own covering letter."
The prompt is carried into the project instance and runs in two situations:
- At project creation, as part of the extraction and classification pipelines that run over any documents already uploaded when the project is created.
- On demand, when you press Classify on the project's document-set page after new documents are uploaded.
Behind the scenes, the AI reads the available documents and returns those that match the prompt. The result replaces the document set's current membership; you can edit it manually afterwards. Writing a good prompt is the same skill as writing a good document check — be specific about what belongs and what doesn't.
Manual assignment
Open the project's document-set page and add received documents to the set directly. This is the simplest option and gives you exact control over membership — useful when the rule is subjective, or when you'd rather not write a prompt for a small handful of documents. You can also use it to tidy up the AI's output after classification runs.
Who can populate a document set
Both classification and manual assignment on a project are available only to approvers and co-approvers of that project. Applicants and collaborators can upload documents but cannot edit document-set membership. The library itself is per-user: each approver maintains their own library of document-set definitions and prompts.
Membership only, not a copy
A document set doesn't store copies of documents. It stores a list of references to documents that live in the project. Moving a document in or out of a set never deletes the underlying document — it only changes the set's membership.
Fields linked to a document set
A project information field can be linked to a document set. Linking a field turns it into a per-document field: every document in the set gets its own value for that field.
For example, if you link a Purpose field to the Relied Upon Documents set, each document in that set can carry its own purpose ("Structural calculations", "Signed builder's declaration", etc.). This is how you capture document-level metadata in a structured way that's available to document templates and approval packs.
The workflow Generating a Project Document from Relied-Upon Documents walks through this end-to-end.