Documents tab: uploads, Microsoft 365 and connectors, the activity tray, redaction, spreadsheets and hidden content, large records, packages, rolling releases and portal delivery
Last updated: November 16, 2026 by Steve
Documents
The Documents tab is where all records related to a request are collected, reviewed, and prepared for response.

Uploading Documents
Click Browse or drag and drop files directly into the drop zone on the Documents tab. You can select or drop multiple files at once — they are uploaded sequentially with a combined progress indicator. Documents can be up to 2 GB each; video and audio recordings and large archives can be up to 50 GB (see Large Records). Files over 100 MB upload in resumable blocks through the Upload queue panel, which shows progress per file and survives a dropped connection (pause, resume, cancel) — keep the page open until it finishes. A file above the prompt threshold (2 GB by default) asks for confirmation first, stating its size and an estimated time; anything over the size limit is rejected with a message. Recordings then open in the media viewer, with transcripts and time-based redaction: see Video & Audio Records.
The same workspace runs on every case type — privacy assessments, incidents, and complaints carry a Documents tab with the same upload limits, previews, and redaction viewer.
The Activity Tray
Some work takes minutes or longer: a large upload, a scan for faces and on-screen text in a video, region tracking, a transcription, an AI redaction analysis, a case-wide Scan for suggestions, a response package, the accessible edition or the accessibility check of a package, or a collection from a repository. None of it needs the panel left open. Close the viewer or the case, or open another case, and a small tray in the bottom-right corner of the page follows the work:
- Uploads are grouped under their case, with pause, resume and cancel. A case whose Documents tab is open shows its uploads there instead. Leaving the page while a file is still uploading warns you first.
- Background jobs are listed one per row: what is running and on which record (for example Scanning for faces and on-screen text — interview.mp4), the case number and title, and a progress bar — the job's own percentage (with about N min left when it can estimate it) or, when it cannot say, a moving bar with how long it has been running. A finished job says what it produced (12 new proposal(s), 3 tracked face(s), Transcript ready, Ready in Export History) or why it failed. Open takes you back to the record's viewer, or to the response package with Export History open. Dismiss clears a finished row.
- A job stays out of the tray while you are looking at it, and appears when you close it. A job that finishes while you are looking at it is reported where you are.
- When a job finishes, screen readers hear one short line. The tray never takes keyboard focus; it moves to the other corner, or folds to its one-line header, when the control you are on would be under it.
Jobs that run on the server come back into the tray when you reload the page or open AccessPoint again. From 2.6.1 that includes every response package; before, a package without recordings was built while the page waited, and a reload lost track of it. A package that is still waiting or being generated cannot be deleted (This package is still being generated. Delete it when it has finished.); a package that failed stays in the tray with its reason until you dismiss it. You see only the jobs you started, and only on records you can still open. A scan or a package finishing still posts its notice in Notices. An AI redaction analysis keeps going if you close the viewer or reload: reopen the document and the viewer shows it running. If the month's AI budget is spent, or your subscription no longer includes AI Assist, the detection rules still run and the row says Only the detection rules ran and why.
Adding from Microsoft 365
Click Add from Microsoft 365 to search or browse records across your organization's Microsoft 365 environment, including SharePoint, OneDrive, and Outlook mailboxes. When your administrator has registered a repository connector that you may use, the button gains a chevron: see Adding from Repository Connectors.

Search
Type a keyword to search across all connected Microsoft 365 sources. Results are grouped into two sections:
- Files — Documents from SharePoint sites and OneDrive.
- Emails — Email messages from Outlook mailboxes.
Select the items you want to add to the request.
Browse
You can also browse Microsoft 365 sources directly:
- My Email — Browse Outlook folders and select individual emails (saved as .eml files).
- Shared mailboxes — Browse one of your tenant's configured intake mailboxes (Settings > Organization), or any shared mailbox you already hold Exchange Full Access to. Listing a mailbox in Settings grants nothing by itself — Graph enforces your existing Exchange permissions.
- My OneDrive — Browse your personal OneDrive files and folders.
- Followed Sites — Browse SharePoint sites you follow, navigating through document libraries and folders.
- Search for a site — Find a specific SharePoint site by name.
Files are copied into the request's document store. Emails are saved as .eml files preserving headers, body, and inline content. When an email has file attachments, each attachment is extracted as its own record — a paperclip marks it and its source reads Email attachment — so you can review, classify, redact, and include attachments independently. Attachments group under the parent email as a family with an expand/collapse chevron, inherit the email's task, folder, and scope, and selecting the parent cascades to its children. Inline signature images and S/MIME signature blobs are not extracted; a forwarded message becomes its own email record whose attachments are extracted in turn. Emails imported before extraction was enabled (or while the tenant toggle is off) can be processed later with Extract attachments on the row — safe to repeat, it only adds what is missing. An attachment that arrived separately can be joined to its email afterwards with Link to email… on the row. The original files and emails in Microsoft 365 are not modified.

More capture sources
Beyond files and email, Add from Microsoft 365 can also capture a custodian's own records from across Microsoft 365:
- Outlook Calendar — the custodian's meeting schedule. Private items are excluded by default, and a disclosure banner explains what was left out.
- Copilot interaction history — the custodian's Microsoft 365 Copilot interaction history. Copilot capture runs server-side and can take a moment: the picker closes immediately and the workspace shows a "Capturing Copilot sessions…" spinner until the transcripts land.
- Teams meeting recordings — recordings of meetings the custodian organized, with the Teams-generated transcript when one exists. Large recordings continue uploading in the background while the page stays open. See Video & Audio Records.
- Lists, OneNote, and Teams chats and channels — captured alongside Outlook email. A List is captured as a spreadsheet: see SharePoint lists.
- Suggested documents — records the picker surfaces automatically, derived from the request's title, scope, and keywords.
Microsoft 365 capture is a per-tenant feature toggle (one for requests, one for assignments and tasks). When it is turned off, these sources are hidden and the button reads Add from….


Adding from Repository Connectors
Responsive records rarely live only in Microsoft 365. Case-management systems, enterprise content management systems, network shares, ticketing and evidence systems all hold records an access office must search. A repository connector brings such a system into the same picker, so you browse, search and add records from it the way you already do, and every record you take carries an account of where it came from and what it looked like when it was taken. Your administrator registers connectors; see Repository Connectors for setup and Repository connectors for what they reach.
Where you find it. As soon as one connector is visible to you, Add from Microsoft 365 gains a chevron. The main half opens the picker as before; the chevron lists Microsoft 365 and then one entry per connector, with the connector's own icon, collapsing to More repositories… after six. A connector whose last health check failed is greyed, with the reason on hover, but is still clickable, because health is advisory. When Microsoft 365 capture is off for the workspace, the button reads Add from….
The picker's Browse home shows every source group you may use, Microsoft 365 and Repositories, each under its own heading. The panel is titled Add records when repositories exist and Add from Microsoft 365 when Microsoft 365 is the only source. Choosing a connector from the chevron opens the picker already inside that connector; Back to all sources returns to the home. You see a connector only when your administrator has made it visible to you (by default, people who can manage documents, such as coordinators) and you may add records to this case in the first place. A closed request refuses captures, as it refuses uploads.
Browsing. A connector opens on its top-level folders with a breadcrumb; click any crumb to go back up. Records show name, size, modified date and author, with the source path beneath. The filter box narrows what is already loaded and Load more fetches the next page. The (i) button on a record opens a details flyout with the standard fields plus whatever metadata the source publishes. A record the source says cannot be retrieved is marked Not retrievable and cannot be selected.
Searching the source. When a connector supports search, a Search here button appears beside the filter box. It sends your text to the source and replaces the listing with the results (with a Clear link back to browsing). The picker's Search tab has a Source selector, Microsoft 365 plus every connector that supports search, so one query can be run against your mail and SharePoint, then against the records system. The Suggested documents tab also asks each connector that supports search for records matching the request's own terms and period.
Adding records. Select records and click Add (n). The picker closes and AccessPoint fetches each record from the source on the server, checks its hash against the one the source declared (when it declared one), and creates the document through the same pipeline as an upload: PDF conversion, search indexing, redaction pattern and suggestion scans, duplicate detection and, where deployed, the malware scan. The record's Source reads Connector Export and it links back to the item in the source. Up to 10 records totalling 100 MB are captured at once; a larger selection becomes a background job that shows in the Upload queue and the activity tray. A record is skipped rather than duplicated when the same item, at the same version, is already on the case. The notice after a capture explains every skip: already captured, access denied, too large, not found or not retrievable.
Collecting a whole folder. On a folder row, the ⋯ menu offers Collect this folder…, the way to take a custodian's entire export or a case file's folder tree in one tracked operation. The dialog lets you narrow the collection by modified after / before dates, a name pattern (* and ? wildcards, such as 2024-* or *.pdf), an extensions list, a maximum size per record and a maximum total size for the job. Choose whether to mirror the source folder structure (on by default; each level becomes a nested workspace folder under the target folder you pick), then Start.
- The Upload queue row reads Discovering records… while the folder is walked, then X of Y processed · Z captured. Cancel stops it; a running job stops at the next record.
- A finished job posts a notice to whoever started it: Collection finished (with captured, skipped and failed counts; a collection with any skipped or failed record ends as Completed with issues) or Collection failed. Individual records that fail never fail the job; they are listed against it.
- Re-running a collection over the same folder is safe: records already on the case are skipped and existing mirrored folders are reused. A record without a modified date passes a date filter.
- Mirrored folders are ordinary workspace folders. Deleting one deletes its sub-folders too; the documents inside are unfiled, never deleted.
Knowing where a record came from. The facet sidebar has a Captured from facet (Uploaded, Microsoft 365, and each connector by name), so "everything we took from OnBase" is one click. On any captured record the row menu offers Source details (the source path and item id, the version taken, the hash with a Verified, Mismatch or Not declared pill, when the source last modified it, who captured it and when, and the collection job it arrived through) and Open in source, which opens the record in the source system in a new tab. You need your own access there; AccessPoint never grants it. Together with the audit ledger entry written for every capture, this is the chain of custody a court or a commissioner will ask for.

Ask AccessPoint. On a case, Ask AccessPoint can look in the connectors you can see and list the hits (names, paths and dates only, never content). When you ask it to bring something in, it offers a capture this record step that you confirm; the assistant never captures on its own.
Connector content stays in your tenant end to end, and every captured record is a frozen copy with a verified hash, never a live link that could change under the case.


Previewing Documents
Click a document name to open the preview viewer. Documents opened from All documents open on the Redline view. The viewer is AccessPoint's own PDF viewer, with thumbnails, an outline, find, zoom and page navigation; a large file is read in ranges, so its first page appears before the whole file arrives. The redaction workspace works as always: Original, Redline and Redacted preview, and drawing, moving and resizing marks. AccessPoint supports previewing:
- PDF files
- Microsoft Office documents (Word, Excel, PowerPoint)
- Images
- Text files
- Email messages (.eml and .msg files, rendered as formatted PDFs)
- Video and audio recordings, which open in the media viewer with playback, transcript, and time-based redaction — see Video & Audio Records
A multi-hundred-page scan opens in the same viewer with pages streamed on demand — jumping deep into the document renders that page directly rather than loading everything before it. The thumbnail pane steps aside past about 200 pages to keep paging fast. In the background AccessPoint also prepares a viewer-optimized copy of very large PDFs, used only for on-screen preview and never for downloads or the released package, so a large scan pages noticeably faster from the second open onward.

Classifying Documents
Each document can be classified as:
- Responsive — The document is relevant to the request and should be considered for the response.
- Not Responsive — The document is not relevant to the request.
- Duplicate — The document is a duplicate of another record already in the request.
Who can classify a given document depends on where it came from:
- Documents collected via an assignment are classified by the Custodian assigned to that assignment.
- Documents collected via a task are classified by the Contributor assigned to that task.
- Documents uploaded directly to the request — including documents attached to a correspondence entry or an external consultation — are classified by the Subject Access Officer or an Administrator.
This mirrors ownership: whoever collected the document decides whether it's responsive.
On assignments and tasks, documents are Responsive unless someone says otherwise. Every document filed on a custodian assignment or a contributor task (uploaded, added from Microsoft 365 or a connector, or extracted from an email) is marked Responsive when it arrives. The custodian or coordinator changes the ones that are not to Not Responsive. An untagged document never stops a custodian from submitting: one still without a relevance when the assignment is submitted counts as Responsive and is marked so, recorded in the History.
Classifying Multiple Documents at Once
To classify several documents in one step, use the multi-select icon in the Relevance column header. A menu lets you apply a value to all unclassified documents in one click, or to a row selection (tick 2 or more checkboxes first).
Redacting Documents
The Redactions sub-tab is where records are marked up for release. Alongside manual redaction, three tools speed the work — and all of them only ever propose a redaction. A person accepts or dismisses every machine-produced redaction, each one carries an origin badge (Pattern, Suggestion, or Propagated) and its matched text, and proposed redactions render as dashed amber boxes that can never reach an export until they are accepted.
- Find & redact — type a text (a phone number, a third party's name, an account number), see per-document match counts, pick the exemption(s), and redact every occurrence in one action, placed word-accurately on each page. Choose Propose for my review so hits await your accept/dismiss, or Apply immediately so hits are applied but flagged for verification.
- Live pattern matching — keep a Find & redact term as a pattern and it runs automatically on every new document added to the request, so the same third party never needs hunting twice. The Patterns panel shows each pattern's applied, awaiting-review, and dismissed counts, and lets you re-run, pause, or retire it.
- Suggested redactions — Scan for suggestions runs your organization's detection patterns (phone numbers, email addresses, SIN/SSN, payment cards, and jurisdiction-specific detectors) across the request and proposes redactions with a confidence score. New documents are scanned automatically after conversion, and — where your jurisdiction pack wires them — accepted suggestions arrive pre-tagged with the right exemption citation.
Filter the sub-tab by Origin (Manual, Pattern, Suggestion, Propagated) and by the Proposed status to review everything the tools did in one place; accept or dismiss inline, per document, or in bulk. These tools work with or without the optional Azure AI Search component — matching runs against the documents themselves.
Nothing proposed slips through. The response-package pre-flight counts unreviewed proposed redactions (pattern, rule, AI, and prior-request suggestions still awaiting accept/dismiss) across the included documents. By default it warns with a Review suggestions shortcut back to this sub-tab; a tenant setting under Settings → Redaction appearance can escalate that to Block, enforced server-side until every proposed mark is accepted or dismissed.
If enrichment partially fails. After a document converts, background enrichment runs — search indexing, redaction pattern and suggestion scans, AI analysis, duplicate detection. When part of that pipeline fails, the document's row shows an amber pill naming what failed, with its own Retry to re-run just the enrichment.
Scanned documents (OCR). When the optional Document Intelligence component is deployed, scanned, image-only documents get their text read automatically, so they participate in content search, duplicate detection and AI summaries like any other document. The AI redaction analysis reads that text too, but text read from a scan carries no word positions, so neither it nor Find & redact or the redaction rules can place a mark on a scanned page: the analysis lists what it found as items to check by eye, and the page is marked by hand. A scan is read once and its text kept. Without it, scanned documents still match by name, type, and tags.
Working with a redaction mark
- Approval. If your organization requires redaction approval, a new redaction enters Pending approval until an approver confirms it — and any edit of an approved redaction, a bulk Change exemption included, sends it back for approval. Administrators can reopen a locked redaction for re-disclosure.
- Recently used exemptions first. Every exemption picker — Redact… on a selection, the side panel (a media mark's card too), Find & redact, parts and grid review — lists your three most recent exemptions under Recently used at the top, shared across all of them.
- A multi-line selection is one redaction, not many. Selecting a passage that wraps across several lines paints a tight box on each line, but they stay a single entry — one exemption, one set of alternates, one comment thread, one status, one line in the redaction schedule. This mirrors how Acrobat models the same mark, so exporting to XFDF and re-importing round-trips without splitting the passage.
- Adjust a mark in place. Click a redaction to select it, then drag its body to move it or any handle to resize — no need to withdraw and redraw (which would lose the exemption, comments, and history). Geometry changes save to the existing mark; its citation and audit trail carry through.
- Alternate exemptions can be captioned on the mark. When a redaction cites alternates, the overlay can read
s.19 (s.22, s.23)so a reviewer sees the fallback citations without opening the panel; alternates that don't fit collapse to a count (s.19 (+2)), and if even that is too wide only the primary shows. It's a tenant setting under Settings → Redaction appearance, defaulting to redline copies only — the released copy stays clean. - Choose what each mark displays. The Label row in a mark's details controls what paints on the box: Exemption (the default — the citation, with alternates per the setting above), Custom text (exactly the text you type, such as a schedule reference or "Third party"), or No label (a clean box). The label is presentation only — a mark set to Custom or No label keeps its exemption, and the redaction index, exemption summary, schedule, and log always cite it. Select several marks and use Set label to apply a mode in bulk. The same three modes apply to marks on recordings.
- Switching views keeps your place. The Original, Redline, and Redacted preview tabs each load once and stay live, so flipping between them is instant and lands on the page you were reading. Original shows the clean source, Redline shows applied marks, proposals, and reviewer markup, and Redacted preview shows exactly the marks a release would burn.
- Downloads name what they give. The viewer header's Download ▾ menu offers the original file in its uploaded format, the PDF without marks, and — on a record with redactions — the burned Redline PDF and the Redacted preview PDF. The redacted preview download physically removes the content under every applied mark before drawing the box, so it is safe to hand out; the redline keeps the content readable under a translucent overlay because reviewers need to verify it. There is no ambiguous download or print button on the viewer toolbar: pick the view you want, and print from that PDF. The Documents-tab row menu also offers Download original, so you can always get the uploaded
.docx,.msg, or image back out.
On a phone or a narrow Teams column
On a screen narrower than 640 pixels, the viewer's side panel starts closed, so the record has the screen. It comes up over the lower part of the record when you select text or draw a box — the exemption form is in it — or when you select Show panel; Hide panel closes it again. The viewer's header wraps instead of scrolling sideways, and a hidden panel keeps the record's redactions loaded and painted on every screen size.
Advisors and reviewers: discussing redactions
An advisor or a reviewer on a request opens its records in the read-only preview, not the redaction workspace. When their role lets them discuss that request's redactions, the preview's header offers Discuss redactions. It opens a panel beside the record listing its redactions by page, with their exemption and a flag icon on the flagged ones.
- Open a redaction to read its flag and its discussion, then Flag for discussion (a reason is required — it starts the discussion) or reply in the discussion.
- A reviewer can also Resolve flag. Resolving needs the Resolve discussion permission, which reviewers hold and advisors do not.
- Every control works from the keyboard, and the panel announces each outcome.
Nothing else is offered there: advisors and reviewers never draw, change, withdraw or approve a redaction, and a redaction on a closed request is locked. On a request they are not on — or for a custodian, contributor or reader — the button does not appear.
The same rule everywhere. Every redaction action is checked against the redaction's own case: a role granted on one case (a case-matched role, or a role limited to a person's own requests) reaches it, an organization-wide permission still works everywhere, and anyone acting on a redaction must be able to open its record.



Spreadsheets and Hidden Content
When a spreadsheet (.xlsx, .xlsm, .xls, .ods, .csv) converts to PDF, AccessPoint lays it out for reading and checks what it holds beyond what prints. Conversion, previewing and redaction all run on AccessPoint's own document engine.
Readable spreadsheet PDFs
Each sheet goes on the smallest paper on which it stays readable: Letter (or A4), then landscape, then Legal and Tabloid (A3), up to the largest paper your administrator allows. A sheet still too wide at the smallest scale (65 % by default) is split into pages across, then down. The header row (or the rows you froze) repeats on every page, and so does column A (or the frozen columns) when a sheet is split across. Row and column headings, gridlines and a sheet · page n / N footer are printed. A cell whose text would be cut off is widened, then wrapped, so every character is in the PDF.
Page layout… (the row menu, and the viewer's toolbar on a spreadsheet) shows how the document is laid out and, for each choice, how many pages it would print and on what paper, before anything changes:
- Readable width and As authored are the two organization defaults.
- Whole sheet puts each sheet on one page of whatever size it needs (up to 200 in and 1,000 rows; a sheet beyond either is laid out as Readable width, and the dialog says which and why).
- Custom sets the paper, orientation, scale or Fit to pages, page order, the rows repeated at the top, the columns repeated at the left, headings and gridlines. Show the pages for these settings works out the pages first.
- Use the organization default returns to the setting under Settings → Document conversion.
Changing the layout needs the redaction-edit permission on the document. A new layout moves the text to other pages, so the document's marks are withdrawn (reason Layout changed, kept in the history) and the dialog says how many first. When the new layout is ready, Find & redact opens with their text so you can propose them again. Once any redaction on the document has been applied, the layout can no longer change. An Office record's page count is the converted PDF's; when a layout used paper larger than Tabloid, the Pages column also shows its standard-page equivalents (12 (40)), and fees and the complexity and statistical reports count those.
Hidden content
Rows, columns and sheets the author hid, content outside a print area, and comments are revealed by default: hidden rows and columns are printed in place, hidden sheets after the visible ones with a Hidden in source footer, and comments in one Comments and notes appendix at the end. Your administrator can choose Author's view instead, which prints only what the author left visible.
- The inventory. Every converted workbook also gets an inventory of what it holds beyond what prints: hidden rows, columns and sheets, comments, text coloured like its background, data far from the rest of the sheet, pivot-table caches, links to other files, data connections, macros, embedded files, hidden named ranges and the document properties. A row with findings shows a Hidden content revealed (or Hidden content found) chip, and Review hidden content… in the row menu lists the findings. Metadata alone (document properties, an attached template, the metadata SharePoint and OneDrive add) is listed but never raises the chip or holds a release.
- Acknowledgement before release. A record whose inventory has findings cannot leave the tenant, in a response package, a secure delivery or the reading room, until someone who manages the document ticks I have reviewed the hidden content this record contained and gives a reason. The acknowledgement covers that inventory: if the record is converted again and the inventory changes, it is asked for again. An export that includes unacknowledged records is refused, naming them.
- Parts. The viewer's side column lists what the conversion added after the sheets, each revealed hidden sheet and the Comments and notes appendix, with its pages and whether it will be released. Go to shows it; Withhold all pages… withholds every page of the part under the exemption you choose; Undo withholding releases it again.
- Hidden in source. Rows and columns the author hid are hatched in the viewer, and a revealed hidden sheet's pages are outlined with a Hidden in source banner. The bands are for review only: they are not marks and never appear in a release. Propose marks over everything hidden in the source adds a proposal over each, which you accept or dismiss like any other.
- Cells. On a laid-out spreadsheet, the redaction list shows the cells each mark covers (
Budget!B4:C9, or part of when it covers only part of them), the response package's Redaction Index cites them the same way, and the viewer shows the cell under the pointer.
Settings → Document conversion (administrators) sets Reveal or Author's view, whether records with hidden content are held until acknowledged (on by default), Readable width or As authored, the largest paper, the smallest scale, and headings and gridlines. Changes apply to documents converted from then on.
Word and PowerPoint records
Word and PowerPoint records get the same treatment: an inventory, the Reveal or Author's view setting, parts that can be withheld as blocks, and the acknowledgement before release.
- Reveal (the default). In a Word record, tracked changes print as markup where they were made (insertions underlined, deletions struck through), hidden text and sections print, and the comments are in a Comments appendix at the end. In a PowerPoint record, the visible slides come first, then the hidden slides as their own block (with a Hidden in source banner in the viewer), then a Comments appendix and a Speaker notes appendix. Objects placed outside the slide never print; the inventory lists them.
- Author's view. The record prints as it reads: changes without markup, hidden text and slides hidden, no appendices. The inventory still lists everything.
- Embedded files become records. A workbook, document, deck or PDF embedded in a Word, Excel or PowerPoint file is added as a record of its own under it (source Embedded object), converted and reviewed like any record. At most 50 per record, and objects embedded more than three levels deep are not extracted.
- Records already converted keep their PDF until they are converted again (Re-convert).
SharePoint lists
Add from Microsoft 365 → Lists captures a SharePoint or Microsoft Lists list as a spreadsheet (source List Export), which then converts like any workbook, so its ID and Title columns and header row repeat on every page.
- What you capture. Capture the entire list as a spreadsheet, or tick items and choose Add (n): either way it is one record. The picker loads up to 5,000 items, most recently modified first; a larger list is flagged, so filter it in SharePoint to capture the rest.
- Columns. Every column the list has, and each item's created, created by, modified and modified by, are ticked by default. Link to item is offered but not ticked, because it puts an address only your staff can open in a record you may release. Hidden and system columns are listed unticked; tick one to include it. Populated only drops empty columns.
- Values. Numbers stay numbers, dates stay dates, date-times are shown in your time zone, and people, lookups and terms appear by name.
- About this capture is written for the requestor: it names the list and its site and the day it was exported, then the selection, the item counts and the notes. It carries no internal address, link or staff name; the record's Source details keep those for your staff. A Columns sheet lists the columns people see in SharePoint.
- Attachments (on by default) become their own records under the list. Include version history (opt-in) adds a Versions sheet.
Releasing a spreadsheet as a file
A workbook (.xlsx, .xlsm) can go out as a spreadsheet instead of, or beside, its PDF, for requestors who ask for data they can sort and total.
- Choosing the format. Release as (beside an included workbook's inclusion switch) offers PDF, Spreadsheet (XLSX), CSV (one file per sheet) or PDF and spreadsheet. As the request prefers follows the request's Dataset format card, which records what the requestor asked for; with neither, a record goes out as PDF. Only a response package releases files; a redline review copy and the originals are always PDF.
- What the file is. It is rebuilt from the original's values, never a copy of the original: each released sheet in order, every cell's value (a formula's last result, never the formula), number formats, column widths, bold and italic, and merges; rows and columns the author hid are shown. A withheld cell reads [Withheld — s.19(1)] (or your custom text, or nothing, following the mark's label); a mark over part of a cell withholds the whole cell. A withheld hidden sheet becomes a one-line statement under a neutral name. Charts, pivot caches, macros, embedded files and the document properties are left out.
- Nothing you did not see. Before anything is released the file is checked independently, and the package is refused, naming the record and why, when a mark is not on cells, when a cell with content did not print on any page of the PDF, or when a cell's number format prints nothing of its value.
- Computed from withheld cells. A total under withheld salaries gives them away. The viewer's side panel lists every released cell whose formula reads a withheld cell (or whose references cannot be checked): Propose marks over them proposes a mark over each, and Release these values as they are… accepts the list as it stands (recorded in the audit ledger). A spreadsheet release waits until each is marked or the list is accepted.
- Review cells in the grid. In the viewer's side panel, Review cells in the grid opens the original workbook, read-only: hidden sheets are listed, hidden rows and columns tinted, and withheld cells hatched. Select cells, choose the exemption, and press M (or Mark selection); the mark is drawn on every page where the cells printed.
- Delivery. In a ZIP the file is under
Files/with the other records released as files; beside a merged PDF it is stored with the export and offered in Export History (Redacted spreadsheet, with its SHA-256), sent by the secure portal delivery and published in the reading room beside the package. The record's original never goes out.
Cross-Request Document History
When a document arriving on a request is byte-identical to one you already processed in an earlier request you can access, AccessPoint recognizes it. The workspace shows a banner, and the redaction viewer shows a Processed before panel — the earlier request, its outcome, and its redactions by exemption — with one click to import the prior redactions as suggestions into the current document. As with every suggestion, you review each one before it is applied. History from requests you can't access is never shown.

AI Assist for Documents
When AI Assist is enabled for your tenant, three optional features help with document review. Each one produces a suggestion or an editable draft — a person always decides what is applied.
- AI redaction analysis — in addition to the pattern-based suggestions above, the AI reads document text in context and proposes redactions for personal information about individuals in their personal capacity (officials acting officially are not flagged), advice and recommendation passages, solicitor-client privileged communications, and third-party confidential commercial information — each with a confidence score and a short reason. Proposals flow into the same review pipeline as every other suggestion, so you accept or dismiss them the same way.
- Document summaries — the redaction viewer's side panel offers an on-demand AI summary per document, collapsed until you expand it — generated in your language when you ask for it and cached until the document's content changes.
- Semantic search and duplicate detection — with both Azure AI Search and AI Assist deployed, document search understands meaning as well as exact words, and duplicate detection additionally catches near-duplicates that plain text matching misses, such as a translation of the same memo or a heavily rewritten draft.
AI Assist is optional. These features appear only when your organization has deployed AI Assist; if you don't see them, ask your administrator.
Privacy and responsible AI. In Hosted AccessPoint, Realizer runs AI Assist in Canada, and an Administrator decides whether AI models that Microsoft may process outside Canada can be used (see Features). In AccessPoint for Microsoft 365, AI Assist runs on an Azure OpenAI resource in your organization's own Azure subscription and region — data at rest never leaves your tenancy, and where inference is processed is your organization's deployment-time choice (global, or bounded to the EU/US data zone). Microsoft does not train its models on your content, and prompts and responses are not stored — only usage metadata is kept, visible per record on the Activity tab's AI assist filter. Every AI output is a suggestion or an editable draft that a person reviews and decides on. For how these features are configured tenant-wide, see the Features administrator page.

Applying Exemptions
SAOs can apply exemption codes to documents or portions of documents. Exemption codes are drawn from the configured list set up by your administrator and correspond to the specific legislative provisions that justify withholding information.
Response Packaging
SAOs manage which documents are included in the final response package. From the Documents tab you can:
- Toggle individual documents for inclusion or exclusion from the response.
- Reorder documents to control the sequence in the response package.
A package that includes recordings generates in the background and delivers them as a ZIP with the rendered media files and a hash manifest — see Video & Audio Records and Response & Fees.
Assembling the response package
The Releases sub-tab (formerly Response Package) lists the records marked for inclusion; drag to reorder (auto-saved). A response package to the requestor is built through a release — Build the next release (see Releases below). The menu keeps Redline review, Originals and Adobe Acrobat XFDF exports for review copies. The pre-flight dialog summarizes redaction counts (active, flagged, withdrawn, untagged) and lets you set the format (Merged PDF or ZIP), the output language, Bates numbering, a watermark (Redline review defaults to a translated "REVIEW COPY — NOT FOR RELEASE"), and a cover letter with the exemption summary, redaction index and exemption descriptions pages. For a release, the Bates prefix starts as the request number (A-2026-0147-) and there is no start number: a release carries on from the request's last number. Generated packages appear in Export History to download or deliver.
The accessibility check. Every response package is checked when it is built. Its row in Export History carries one accessibility chip that sums up the check and the accessible edition, coloured only when someone has to act:
- Accessibility: checking… or Accessible edition: building… while the work runs;
- Accessible (green) when the package passed, or Accessible edition ready (green) when its edition is ready and nothing holds the package back;
- in amber, Accessible edition: n figure(s) to describe, Accessible format owed (the request's Accessible format requested is set and no edition is ready) or Accessibility: action needed (your organization requires a pass or an acknowledgement before any secure delivery or publication, and the package has neither);
- Accessible edition failed in red;
- otherwise a neutral Accessibility details, Accessibility: acknowledged or Accessibility: not checked.
When the requirement covers only the reading room, the chip never asks for action. A package that did not pass is not an error in itself: a scanned record often cannot keep its tags, which is what the accessible edition is for. A package that was never checked says so, with Check now in its report. The chip opens the package's report, which starts with the same summary and what to do about it:
- What was found, failures first: a record that is not tagged for screen readers (a scan, or a record whose tags were removed when its redactions were burned in), a merged PDF that is untagged because some of its records are, redactions drawn without a label (a screen reader is not told that text was withheld), figures with no text alternative, and anything the checker reported. Recordings and other files released beside the package are listed as not checked.
- The PDF/UA-1 check. When your organization deployed the accessibility checker, it checks the package in your own subscription and lists each file with the rules it fails and the pages concerned. Check again runs a fresh check at any time, even while one is still waiting; a package built before 2.5.0 has Check now. When the checker is configured but does not answer — it is not deployed, or it stopped — the package reads No answer from the accessibility checker after three hours instead of Accessibility: checking…: AccessPoint's own checks still show, no PDF/UA-1 claim is made, and a release rule that requires a passing check holds the package as needing review, which an acknowledgement releases. An answer that arrives later still counts. Without the checker no package can pass, and AccessPoint never describes one as meeting PDF/UA-1.
- Human checks. No machine can tell whether the reading order makes sense or whether a text alternative describes its image. Open the package with a screen reader, then select I checked the reading order and the text alternatives; the confirmation names you and the date, is written to the audit ledger, and can be withdrawn.
- Release anyway. Someone holding the permission Release packages that did not pass the accessibility check can record why a package may be released although it did not pass. The reason is kept in the audit ledger and the acknowledgement can be withdrawn.
A package passes only when the checker found nothing, AccessPoint's own checks found nothing and a person confirmed the human checks. Rebuilding a package clears its confirmation and acknowledgement. Settings → Redaction appearance → Accessibility check before release decides what the verdict does: Advisory (the default; it never blocks), or Require a pass, or an acknowledgement, before publishing to the reading room, or before any secure delivery or publication.
The accessible edition. A redacted, scanned or untagged record cannot keep the tags a screen reader needs, so AccessPoint can build an accessible edition beside a response package that holds one. Azure AI Document Intelligence reads the redacted pages (never the originals), and each record is rebuilt as a tagged PDF with its text, headings, tables and figures in reading order and every redaction announced where it sits ([Withheld — s.19(1) Personal information]). The package itself is never changed. By default an edition is built only when it will be used: at once when the request's Accessible format requested is set; otherwise when the package is delivered through the portal or proposed for the reading room, or when you choose Build accessible edition in the accessibility report. Until then the edition is Not built, and the chip stays neutral unless something else needs a person. The report's Accessible edition section shows its status, how many records were rebuilt, any record whose pages could not be read, a Download for each file, and Build again when a build failed. Once the edition is built, the accessibility check checks the edition, since that is what a requestor using assistive technology reads. Settings → Redaction appearance → Accessible edition of response packages chooses When it is needed (the default), For every response package that needs one, or Off. Editions need Document Intelligence, billed per page read.
Figures need a person. When AI Assist is deployed and Figure description drafts is on, each figure arrives with a draft description; the report shows every figure beside its draft. Choose Use this description (edit it first if needed), Decorative (a screen reader skips it) or Draft again. A figure nobody described is announced as Figure not described. Choose Rebuild with these descriptions after describing figures.
Releases
A request is often answered in instalments: the records that are ready go out first — Release 1 — and the rest follow in Release 2, Release 3 and the final release. Later, a record can go out again with corrected redactions (a corrected release), or records found after the final release can follow it (a supplemental release). Since 2.6.0 every response package to a requestor is built through a release, on the Releases sub-tab of Documents (formerly Response package). Review copies (Redline review, Originals) and the redaction log stay in the tab's menu as before.
AccessPoint records each release once, when it is built, and never changes it: which records went out, at which redaction state, with which page and Bates numbers, in which package, with which letter, delivered how and when, and paid how. To change what went out, build a corrected release.
Building a release
- Select Build the next release. The picker lists the request's records that are In the response and ticks, by default, the ones ready and not yet released: no proposed or unapproved marks, no untagged mark when your organization requires exemptions, a clean malware verdict (or an acknowledgement), hidden content acknowledged where that check is on, a usable PDF, and the request's required reviews done.
- Untick a record to keep it for a later release, or choose Withhold in full to decide it in this release with its exemptions — each release's index then accounts for every record it decided. A record already released and unchanged can go again (the picker warns); a record whose redactions or file changed since it was released goes only in a corrected release.
- Choose the kind (Interim or Final), an optional title and the cuts to build (the default cut and any variants), then Build release.
AccessPoint numbers the release, reserves its page numbers and builds one package per cut. From 2.6.1 the build answers at once: the packages are generated on the server, the activity tray shows their progress, and you can keep working, close the case or reload the page. A release with several cuts queues all of its packages, then generates them one after another. A package that cannot be generated returns its release to preparation with the reason, and the failed package stays in the tray with that reason; a package that waited a day without starting is marked failed and can be generated again. If the build fails, nothing is numbered and the draft stays open. A release that only withholds records is built without a package. A request still in Draft takes no release: the tab says The request is still a draft. Activate it before building or sending a release.
Page numbers
Bates numbers run on across releases: Release 2 starts after the last number of Release 1, every cut of one release stamps the same number on the same page, and a corrected record gets new numbers — a number always names one page. The build dialog's Prefix starts as the request number and a hyphen (A-2026-0147-000086), so every page says which request it belongs to; change it, or clear it for the numbers alone. A release built earlier keeps the numbers it was stamped with. Recordings and files that go out as files take REC- labels that also run on. Review copies number from their own INTERNAL- counter, so a printed review copy can be cited without using the requestor's numbers.
Where each record went
The Documents tab's Release column and filter show each record's state: Not yet released, Release 2, Withheld (Release 1), Changed since Release 2 (its marks or file changed after it went out) and Corrected in Release 3. The redaction viewer's document details say the same. A record decided in a built or sent release cannot be deleted, and a release's packages cannot be deleted or cleaned up by housekeeping.
Sending a release
On a built release, select Send release… — one action with three parts you confirm together:
- Delivery — through the secure portal (the package of the cut you choose, its expiry and whether the requestor is told), or another way (Email, Mail, In person, Courier, Fax or Other) with the day it went and a note. The portal checks the package's own records again (malware and hidden content) before anything leaves.
- The letter — none, a letter already on the request, the letter held for the senior official's sign-off, or one you Compose… now. The composer offers your templates of the release's purpose (Interim response, Final response, Corrected response, Supplemental response) with the release's details filled in — the request's own jurisdiction pack's letter first, otherwise the Universal baseline's generic Interim release, Corrected release and Supplemental release letters. When the letter needs sign-off, the whole send waits: only the letter goes for approval, and once it is approved you select Send release… again with it.
- The response date — the dialog says what sending will do to the request's response date (see Response dates and statistics).
A release sent without a letter offers Link a letter… afterwards. Each built or sent release shows its Index: every record it decided, with pages, disposition, Bates numbers or label, and exemption codes. The index says what a corrected record replaces ("replaces ATIP-000001 – ATIP-000002") and which later release replaced an earlier one.
Letters for a release. Release letters can quote the release's number, kind, date, records, pages, Bates range and index, the index of every release so far, and its instalment; any letter on the request can quote the next release date and the remaining estimate from the release schedule. Any letter to the requestor can also name a release: in Send correspondence, pick it under Release. See Sending Correspondence and, for the tokens, Correspondence Templates. Letters to a requestor use everyday words for a release in their language — Spanish entrega, Danish udlevering, German Lieferung — while AccessPoint's own screens keep their terms.
On the portal
The requestor's case page lists deliveries grouped by release, newest first ("Release 2 — sent 14 November 2026"), each with its figures and its index. A record withheld in full is listed without its name (its title can itself be what is withheld), with exemption codes only.
The requestor sees the cut they received. When a release went out as a cut that leaves records out, its index, figures and Bates range on the portal — and the release details in a letter for it — list only the records in that cut and what was decided about them; nothing names a record the cut left out. Your own view of the release (its Index on the Releases sub-tab) always shows every record it decided. A corrected release says which release it corrects and why, in one of four categories — after an appeal decision, after a complaint or a regulator's order, because an error was found, after a third party was consulted — never your own words; an earlier release says which release replaced its records.
Response dates and statistics
Sending releases writes the request's First release sent and Final release sent dates (Key dates) and its response date — the date the on-time statistics read: the day the final release is sent, or the day the first release is sent where the request type's legal authority says so (Settings → Legal authorities → Response date counts from). A corrected or supplemental release moves none of them. Once a request has a built or sent release, its response date on the Response tab is read-only.
Pages disclosed is the sum of the pages released by sent releases, unless someone entered the figure by hand. Each record counts once, by its latest release: a record released again unchanged, re-issued by a corrected release, or released after an earlier release withheld it is counted in Pages disclosed, the to-date letter figures, the status report's running totals and the Report Builder subject Releases only as its latest release decided it. Each release's own figures still say everything it contained.
Release schedules
The Releases sub-tab opens with a Schedule panel. Select Set schedule… and choose:
- Fixed dates — each with a label, an estimate of pages and whether the requestor was told;
- Every N days from the first release; or
- A number of pages per period — AccessPoint dates a tranche of the remaining pages for each period.
Say where the plan comes from (Our estimate, Agreed with the requestor, Ordered by a court, with the order's reference) and add a note. The dialog previews the dates as you type. A schedule estimates; it never blocks:
- The next release due shows on the panel, in Key dates and in My Day ("Prepare Release 2 … due in 3 days", "Send Release 2 … overdue since …"), and a follow-up rule can open work on it (the case date Next release due). A date names the release that will answer it: a release built and not sent yet (held until paid, say) comes first, then the next new number.
- Each sent release answers the earliest open date; a release sent after its date is recorded as missed, and a final release records every earlier date it leaves unanswered.
- Change schedule… and Delete schedule… need a reason, keep the dates already answered and record the missed ones. Telling the requestor is a letter someone sends.
Status report downloads, as CSV or Excel, each release and schedule date with its figures — records and pages reviewed, released in full, in part and withheld in full — the running totals, and the schedule against actuals: the figures a court status report asks for. Across requests, the Report Builder subject Releases gives the same rows for every request you can see, with net figures you can add up across releases without counting a record twice (see Custom Reports & Report Studio).
Instalments and hold until paid
A built or sent release can carry its own instalment of the request's fee. Select Request installment…: AccessPoint suggests the fee schedule's charge for the release's pages — after the free pages the earlier releases did not use — and caps it at what the fee leaves (instalments plus the deposit never exceed the fee). Accept the amount or change it.
- A record the requestor already received is free when it is sent again. Each record's pages are charged once across the request's releases, so a record released again unchanged, or re-issued by a corrected release, is not charged in the later release, and the dialog says so (3 pages already sent in Release 1 — not charged). Received means sent: a record whose earlier release is built but not sent yet is charged in the release that goes out first. A record an earlier release withheld in full was never received, so it is charged when it is released.
- A record another release's instalment already asks for is free too. Once an instalment is asked for (open or paid), its pages are not charged again in another release, even before either release goes out; whichever instalment was asked for first charges them. If that instalment is cancelled or waived, its pages are charged again in the next suggestion.
- The copies must cost the same. A record sent before as electronic copies and now sent on paper (or the other way round) is free when the fee schedule charges both at the same rate, and charged when it charges them differently — the dialog says so.
- How the copies go. Choose Electronic copies (portal or email), Paper copies or Inspection (no copies). Paper copies are charged at the schedule's duplication rate; electronic copies at the schedule's own electronic rate (free when the schedule makes them free — then nothing is suggested); inspection is never charged per page. The choice starts from how the release went (mail, courier, fax or in person: paper; the portal or e-mail: electronic) and, for a release not sent yet, from the request's response method.
- Getting paid. With electronic payments on, the requestor can be e-mailed the link to pay on the portal. With them off, ask for it in a letter — the Universal baseline's Fee for a release letter, sent with the release named — and record the payment when it arrives.
A request type can hold releases until paid (Settings → Request types → Hold releases until paid; each release can be switched on, off or back to the type's setting). A held release with an open instalment cannot be sent — by Send release or by the Export History's Deliver — until the payment is recorded: only a payment the payment provider confirmed (or one recorded by hand) counts, never the requestor's return from the payment page. To let it go anyway, Waive installment… (a fee waiver, with Waive fees) or Allow sending before payment… (the instalment stays owed; needs Act as senior official and a reason). Cancelling a held release's open instalment needs the same permission and a reason. A release with no instalment is not held.
Corrected and supplemental releases
- A corrected release names a sent release and the reason the records go again — Appeal decision, Complaint or regulator order (with the complaint, when one is linked to the request), Error found or Third-party consultation outcome — and lists only that release's records (a record changed since it was released is offered here). Each corrected record gets new page numbers; the earlier release and its packages are kept as evidence.
- A supplemental release adds records no release decided yet, after the final release.
Both can be built and sent on a closed request without reopening it (Build a corrected or supplemental release, with a reason in words). Marks on a closed request stay locked except through the existing unlock for re-disclosure.
When the last release was the final one
When you close a request whose releases include no final release, the Close request dialog offers Mark Release n as the final release (the last release sent): tick it, say why, and the release is marked before the request closes, so the on-time statistics get their response date. You can also open the last release sent and select Mark as the final release…, give the reason and select Mark as final. Its day becomes the final release date and, where your jurisdiction counts the final release, the response date; the release schedule expects nothing more. A release is marked once, and only the last one sent while nothing else is in preparation or waiting to be sent.
A release whose secure-portal delivery is still waiting for its scan cannot be sent again — it is sent when the delivery opens; revoke the delivery to send it another way.
In Lab mode, releases are training case data: a release is sent "another way" (the portal is not used), its letter is simulated, and no instalment can be asked for.
Settings that shape releases
- Legal authorities → Response date counts from — The final release (the default) or The first release: which sent release sets the response date of requests under that authority. Jurisdiction packs set it from their statute (California: the first release). See Legal Authorities.
- Requestor fee categories → Electronic copies — Charged like paper copies (the default), Free or Their own rate per page: the request's estimate and final fee and each release's instalment read it. California's pack makes them free. See Requestor Categories.
- Request types → Hold releases until paid — off by default; California's public-records types switch it on. See Request Types.
Settings that jurisdiction packs bring are a starting point built from public information about the legislation, not legal advice — your legal counsel validates and adjusts them.
Delivering Records Through the Portal
With the public portal deployed and enabled, a generated response package can be handed to the requestor through a secure delivery instead of an e-mail attachment or a courier. The usual way is Send release… with delivery through the secure portal (see Sending a release); in Export History, a Ready export also offers Deliver via secure portal (the requestor must have an e-mail address on file). Give the delivery a title the requestor will see, the days until it expires (30 by default, 1 to 365) and an optional per-file download limit. AccessPoint copies the package, with the rendered recordings and native files released beside it, into a delivery area the portal can reach, records each file's size and checksum, and, when Defender for Storage is deployed, waits for the copies to be scanned before the delivery turns Available. A copy Defender flags revokes the whole delivery. The delivery letter carries the portal link, or copy it from the request's Portal tab.
The requestor opens the link, enters the e-mail address the release was sent to, and receives a fresh 8-digit code on every visit. A requestor who has just checked the request on the portal is not asked again for a delivery of that request sent to that address; a delivery sent to anyone else, or opened from the link without such a session, asks for its own code. The page lists the files with sizes and hashes (the accessible edition beside the package, marked as such), says in About these files what they are and how to open them, and under Accessibility says in plain words what the accessibility check found and how to ask for another format. Each Download is a short-lived link to that one file. The requestor may Confirm I have received these records, which stamps the acknowledgement on the delivery and e-mails them a confirmation. Every code, page view, download and acknowledgement is written to the audit ledger.
Revoke (Response & fees → Secure deliveries) or expiry deletes the copied files, which ends every outstanding download link immediately. While the portal is switched off, its delivery pages are closed; to revoke a delivery, switch the portal back on and revoke it. Before a delivery or any export can leave the tenant, every included record must carry a clean malware verdict or an explicit acknowledgement (see Large Records).
Large Records
Two size classes apply in every workspace: documents, up to 2 GB per file, and recordings and large archives (video, audio, and a configurable list such as .zip .7z .pst .iso .tar .gz .bak), up to 50 GB per file, the largest object Microsoft Defender for Storage can scan. Both ceilings are settings an operator can raise, and 0 removes a ceiling. Files over 100 MB always travel through the resumable Upload queue.
- The confirmation dialog. Above the prompt threshold (2 GB by default) an upload or a picker capture is not refused; it asks first, stating the size and an estimated time and, when Defender is deployed and the object exceeds 50 GB, that it will not be malware-scanned and will need an acknowledgement before release.
- Not scanned. A record Defender could not cover shows a Not scanned pill with the reason on hover. Such a record cannot be included in a response package or a secure delivery until someone who manages the document acknowledges the risk with a reason; the acknowledgement can be withdrawn. Malicious still blocks and Pending still holds an export. Tenants without Defender are unaffected.
- Collection jobs have a budget. The Collect dialog's Maximum total size caps the bytes one job may capture (500 GB by default;
0is unlimited); further records are skipped as Budget reached and the job completes with issues. A fetch fails when no bytes arrive for two minutes. - Recordings above 2 GB are prepared only by the media processing job in your Azure subscription; without it such a recording is marked Failed with a message naming the setting. A package has no 2 GB per-document ceiling.
- Files in archive storage. Where your organization switched case archiving on, a file of a closed case can carry a chip on its row, Archived, Restoring — ready about …, Preview removed or Disposed of. Opening an archived file offers to restore it, with the time and the cost; a removed preview is rebuilt on request.
Working Files
The Working files sub-tab is a scratch area for drafts, such as a response letter in progress, with check-out and versioning, notes and, for requests, a Finalize action that seals a draft into correspondence. Word and spreadsheet working files open in editors built into AccessPoint. The Word editor handles text, paragraph styles (Normal, Heading 1 to 3, Title), character formatting, alignment, bullets and numbering, links, tables and page breaks, and shows the letterhead read-only on every page. The spreadsheet editor has formulas, number formats, fonts, fills, borders, merge, wrap, freeze panes, several sheets, row and column insertion and deletion, undo and find. Both save into the original file, so whatever they do not edit (images, fields, tracked changes, comments, pivots) is kept exactly as it was.
Related Pages
- Video & Audio Records — marking, tracking and releasing recordings.
- Duplicate Documents — grouping and handling duplicates.
- Repository Details — repositories bound to a connector folder or Microsoft 365 location.
- Response & Fees — the response package pre-flight and fees.
- Mark Package as Sent — which package reached the requestor, and when.
- Request Collaboration — advisors, readers and discussions.
- Repository Connectors — registering and managing connectors (administrators).
- Repository connectors — which systems connectors reach.