The Vault: your data section

What the Vault is

Every business fact Verinode uses, an invoice, a certificate, an estimate, a carrier scorecard, an emailed statement, starts life as a document. The Vault is where you can see that pipeline directly: what has come in, what Verinode read from it, and what it turned into. It is not a filing cabinet you organize by hand. It is the receipt trail for the data that feeds every other section of the platform, vendors, jobs, clients, carriers, margin, benchmarks, all of it.

Open Vault from the sidebar at iq.verinode.ai/data. The page header reads "Vault" with a document-stack icon, and an Add Data button sits in the top right (keyboard shortcut Cmd+U / Ctrl+U). That button opens the same capture window used everywhere on the platform: drop files, snap a photo, paste text, tell it to you by voice, or forward an email. See Forwarding documents and Connecting your data for the ingestion channels in detail. This article covers what the Vault page itself shows you once data starts flowing.

Pinned in the top-right corner of the page is a small pill that reads "Encrypted by your Vault Key." It is a standing trust indicator, not a decoration: your operator data is encrypted at rest under a key scoped to you, never sold, and every access is logged. If you are the account holder, tapping it opens the reveal flow for your recovery key. If you are a teammate on the account, the pill stays informational (it is true for you too) but does not open the reveal, since only the account holder can recover the master key.

The page has three parts, stacked top to bottom: a hero panel, an Explore row of four tiles, and a Recent uploads row. There is no old-style tab bar across the top anymore, everything opens from a tile click.

The hero: "What Verinode Is Tracking"

At the top of the page sits one large panel with a single dominant number and two supporting metrics beside it.

The big number is your total document count, labeled "document(s) captured." It counts every row in your document log, processed or not, and animates upward when the page loads. Next to it, a colored pill shows your average extraction completeness as a percentage, for example "84% Extraction Completeness." This number answers "of the fields Verinode could plausibly pull from your documents, how many did it actually capture," averaged across every document that has a receipt. It is not the AI's self-reported confidence, it is a completeness ratio, so a low number means fields are genuinely missing, not that Verinode is unsure of itself. The pill's color tracks the same bands used throughout the Vault: green (emerald) at 80% and above, amber from 40 to 79%, red (rose) below 40. The pill only appears once you have at least one document; with zero documents there is nothing to grade yet.

Below the number, a line of sub-text adapts to your state:

• With no documents yet: "Connect a channel, email, uploads, voice, or photo, and your vault starts filling."
• If any documents need a quick assignment: "N document(s) need a quick review to unlock more signals."
• Else if any are still processing: "N document(s) still processing, signals fire as soon as they land."
• Otherwise: "Feeds every section of your business, vendors, clients, jobs, equipment, team, certifications, and more."

Two secondary tiles sit beside the hero number:

• Signals Fed, the count of documents that triggered platform intelligence, meaning at least one signal in your Signals feed traces back to that document. Its sub-label reads "documents that triggered platform intelligence."
• Needs Review, the count of documents Verinode auto-created an entity for but could not confidently match, so a person, vendor, or job link is still pending your confirmation. When this is above zero it highlights in Ember Red with the sub-label "entities to assign"; at zero it reads "all matched" in the neutral tone.

At the foot of the hero, under a hairline divider, sits the one-line trust statement described above (encrypted, scoped to you, never sold, access logged).

Explore: the four tiles

Below the hero, an Explore row holds four tiles side by side. Each opens the same overlay, a four-tab card slider, jumped straight to the tab matching the tile you clicked.

Uploads

Shows your total document count as the headline number. The sub-label prioritizes what needs your attention: if any documents need review it reads "N Need Review"; otherwise if any are still processing, "N Pending"; otherwise "Documents Captured." A small segmented bar underneath breaks that same total into three colored slices: documents that fed signals (good/green), documents needing review (warn/amber), and documents still pending (neutral/gray), so you can read the health of your vault at a glance, not just its size.

Clicking this tile opens the Uploads tab, the full document table. See "Inside Uploads" below.

Tracking

Shows the count of distinct entities, vendors, clients, or jobs, that Verinode has built a document trail for. The sub-label shows the average extraction completeness across those entities as a percentage ("N% Avg Completeness") when at least one entity has a receipt; otherwise it reads "Entities Tracked" if you have entities with no receipts yet, or "Add Data To Unlock" if you have none at all. A small gauge below the number reads that same completeness against the 40/80 thresholds, colored green, amber, or red the same way the hero pill is.

Clicking this tile opens the Tracking tab, the entity-level rollup. See "Inside Tracking" below.

Playlist, Reading, Saved

These three tiles surface the lists you build while browsing the Feed: things you have queued to watch, read, or simply bookmarked.

• Playlist shows the count of saved audio/video items, with "N Unlistened" when some are unplayed, "Queued For Later" when all are listened to, or "Save Audio From The Feed" when the list is empty. A small dot grid shows total items with the still-unlistened ones flagged amber.
• Reading shows the count of saved articles, with "N Unread," "Queued To Read," or "Save Articles From The Feed" when empty. A small progress ring shows the share of the reading list already read.
• Saved shows your bookmark count, with "Bookmarked Items" or "Bookmark From Any Card" when empty. It has no read/unread split, so it is a plain count with no chart.

All three open the Lists tab pre-filtered to the matching list.

Notifications

Shows your unread notification count as the headline (or the total count if nothing is unread). The sub-label reads "Unread" (plus "· N Scheduled" if any are queued for later) when you have unread items, "All Caught Up" when you don't but have notifications on file, or "Alerts Land Here" when you have none at all. A segmented bar breaks the total into unread (amber), scheduled (info blue), and read (neutral).

Clicking this tile opens the Notifications tab, jumped to unread if you have any.

Recent uploads

The bottom row of the Vault page, "Most recent," shows up to eight of your latest documents as tiles, newest first. Each tile shows:

• The document type as a small label (for example "invoice," "certificate," "estimate"), humanized from the underlying type.
• The source it was matched to, a vendor name, carrier or client name, or a job reference, as the headline.
• The extracted dollar amount underneath, when one was captured, formatted as $1.2M / $340k / $1,250 depending on size, or an em-dash equivalent when no amount applies.
• How long ago it was processed ("3h ago," "2d ago," etc.) and, if it triggered any signals, a bold count like "2 signals" beside it.
• A small completeness badge in the corner, colored green (≥80%), amber (≥40%), or red (below), when the document has an extraction receipt.
• Each tile is color-accented by which bucket it belongs to: vendor documents in IQ teal, client/carrier documents in steel blue, job documents in copper.

Clicking any tile opens the Uploads tab with that specific document's row already expanded to its extraction detail.

Empty state. With no documents at all, this row reads: "Recent captures appear here as documents flow in. Forward an email, upload a file, or paste a job export, the latest 8 land in this row."

Inside Uploads: the full document table

This is the working list of every document Verinode has on file for you, most recent first, up to 200 at a time. A search box filters by vendor, carrier, job, or document type. Two dropdowns narrow further: All buckets (All / Vendors / Clients / Jobs, each showing its own count) and All types (every document type present in your data, listed by name).

If you have fewer than five documents on file, a Connect prompt appears above the table nudging you toward the ingestion channels that fill it fastest.

Each row shows:

• Source, the vendor, client, or job the document is linked to (a link into that section), with the bucket name underneath. Unlinked documents fall back to the section you filed them under, or "Unknown" only as a last resort. If a document is flagged for review, an amber "Assign vendor / Assign carrier / Assign job" pill sits beside the source, click it to open the assignment panel inline.
• Type, a rounded badge with the humanized document type.
• Amount, the extracted dollar total, right-aligned, or a dash if none was captured. If the document was in a foreign currency, a small blue "Confirm" chip shows the converted USD figure and lets you confirm the conversion with one click.
• Processed, the date Verinode finished reading it.
• Signals, the number of signals this document triggered, as a clickable copper pill linking to your Signals feed, or a dash if it produced none.
• Status, a colored dot and label: Processed (green), Pending (amber), or Failed (red). Beside it, when a receipt exists, a percentage chip shows the extraction completeness, click it to expand "What we read from this document."

Expanding a row shows a plain-language receipt: what fields Verinode captured (a comma-separated list, humanized), and, if anything did not come through cleanly or the completeness was low, a note saying so with two actions: Edit fields (opens an inline form to manually correct document type, amount, period start/end, document date, and reference number) and Re-run extraction (re-processes the source file, useful after you have replaced or fixed the original).

Row actions on the right: a preview-eye icon opens the original document in a new tab (when the source file is on hand), a pencil opens the edit form directly, and a trash icon asks you to confirm before removing the document, which also removes any signals sourced from it. You can select multiple rows with the checkboxes; a floating bar at the bottom of the screen then lets you remove them together.

Empty state. With no documents matching your filters, or none at all, the table is replaced by an ambient panel: "Your document intelligence builds as files flow in," with five bullets describing what happens automatically once documents arrive, classification, entity matching, data extraction, signal detection, and a full audit trail.

Inside Tracking: what Verinode knows, entity by entity

Where Uploads lists documents, Tracking rolls them up by the vendor, client, or job they belong to, one row per entity. This is the "what does Verinode actually know about this vendor, and how confident is it" view.

A search box filters by entity name. Two dropdowns narrow the list: All entities (All / Vendors / Clients / Jobs, each with a count) and All completeness (High ≥ 80%, Medium 40 to 79%, Low under 40%, or No receipt).

Each row shows:

• Entity, the name (linking into its section), with an amber "Needs Review" label if any of its documents are still unmatched.
• Type, Vendor, Client, or Job.
• Docs, how many documents have been filed against this entity.
• Completeness, a colored pill with the average extraction completeness across that entity's documents, or "no receipt" if none of them have one yet.
• Fields, the count of distinct data fields Verinode has captured across every document tied to this entity.
• Signals, how many signals this entity's documents have collectively fed, in copper, or a dash if none.
• Last update, how long ago the most recent document for this entity was processed.

Click any row to expand it and see the actual list of fields captured, each shown as a green pill with its humanized name, for example "Total Amount," "Period Start," "Reference Number." This is the transparency point of the tab: no black box, you can see exactly what data Verinode extracted for this vendor or job, not just a score.

Empty state. With no tracked entities at all: "Tracking fills as documents flow in and Verinode extracts fields," with three bullets explaining that each invoice, certificate, and statement is parsed by capability, that fields stack up per vendor, client, or job, and that clicking a row shows exactly which fields are tracked. If entities exist but your filters match none, it instead reads "No entities match these filters."

Inside Lists: Playlist, Reading, Saved

One combined table for everything you have queued from the Feed. A search box filters by title or source. Two dropdowns narrow further: All lists (All / Playlist / Reading / Saved, each with a count) and a read-state filter (All / Unread / Read).

Each row shows a small thumbnail, the title (linking out to the source when one exists) with the source name underneath, a colored "Playlist / Reading / Saved" pill showing which list it came from, a media-type badge (video, article, etc.), and how long ago it was added. Row actions: a checkmark to mark an item read, an external-link icon to open its source, and a trash icon to remove it from the list. Read items dim slightly once marked.

Empty state. With nothing saved yet: "Lists fill as you save items from the feed," with three bullets, watch a video and it lands in Playlist, save an article and it lands in Reading, bookmark anything and it lands in Saved. If items exist but none match your filters, it reads "No items match these filters."

Inside Notifications: your full alert inbox

Every alert Verinode has sent you, read or not, live in the inbox here. A search box filters by title or body text. Two dropdowns narrow further: a state filter (All states / Unread / Read / Scheduled / Dismissed, with live counts) and a category filter listing every category present in your notifications.

Each row shows the notification title (linking to the relevant page when applicable) with its body text underneath, a category badge, a colored state pill (Unread in copper, Scheduled in amber, Dismissed in muted gray, Read in green), the delivery channel ("In-app + Email" or a single channel), and a timestamp, either how long ago it fired or, for scheduled items, how long until it does. Row actions: a checkmark to mark read, an arrow to open the linked page, and an X to dismiss.

Empty state. With no notifications on file: "Notifications appear as Verinode flags things worth your attention," with three bullets, decision reminders fire when a step is due, survey responses ping you when a teammate replies, scheduled emails queue here before they go out. If notifications exist but none match your filters, it reads "No notifications match these filters."

How a document becomes a signal

The Vault is the visible half of a pipeline that runs the same way for every document, regardless of channel:

1. 1A document arrives. You forward an email, drop a file, snap a photo, paste text, or say it out loud through Add Data. Voice and paste get transcribed; files and photos get read directly.
2. 2Verinode classifies it. The document type (invoice, certificate, estimate, scorecard, statement, and more) and its likely bucket, vendor, client, or job, are determined automatically.
3. 3It gets matched to an entity. Verinode tries to link the document to an existing vendor, carrier, or job. When the match is confident, it links automatically. When it isn't, the document lands with Needs Review so you can confirm the link yourself from the Uploads table.
4. 4Fields get extracted. Amounts, dates, reference numbers, and other structured fields are pulled out and written to your record. The extraction receipt records exactly what was captured, this is what powers the completeness percentage everywhere in the Vault.
5. 5Signals fire. Where the extracted data is significant, an anomaly, a benchmark-worthy pattern, a renewal date approaching, it becomes a signal in your Signals feed, traceable back to the exact source document.

This is also why the Vault's numbers move even when you haven't touched anything: a scheduled cron reprocesses queued documents, and QuickBooks/Xero or other connected sources can add rows without a manual upload.

Best-practice example

Say you forward a batch of vendor invoices by email. Within minutes, the Uploads tile's total climbs and its status bar shows a slice of pending documents processing. As they clear, the Recent uploads row starts showing tiles with completeness badges, mostly green. One invoice comes back with an amber "Assign vendor" pill, the vendor name on the invoice did not closely match anyone on file. You click it, confirm it is a new vendor, and it links. Now open Tracking: that vendor has one document, a completeness score, and a small set of captured fields, total amount, period dates, reference number. As you forward more invoices from the same vendor over the following weeks, its document count in Tracking climbs, its field list grows, and its completeness score stabilizes, all without you doing anything except confirming that one initial match.

Related reading

• Forwarding documents, the email and channel mechanics behind Add Data
• Connecting your data, automated ingestion sources like QuickBooks and Xero
• The Feed, where Playlist, Reading, and Saved items are created
• The decision workspace, where signals fed from your documents turn into decisions
• Understanding your margin, one of the sections your extracted documents feed directly

Data sources