Skip to main content

Open in AthenaHQ

app.athenahq.ai/knowledge-base/review

​
Purpose

This page allows customers to manually review, approve, or reject knowledge base claims that Athena’s AI has extracted from various sources (such as uploaded documents, research runs, or AI responses). By default, claims that the AI flags as “high” or “medium” confidence are automatically approved and added to the customer’s knowledge base. The pending queue primarily exists to catch “low” or unknown confidence claims. Users must approve these claims before they officially count toward published knowledge base pillars, ensuring accuracy and brand safety.

​
What’s on the page

  • Page Header: Displays the page title and a “Back to Knowledge Base” link.
  • Description / Pagination Summary: Explains how the queue works (noting that high and medium confidence claims normally auto-approve). If there are more than 100 claims, it shows the current viewing range (e.g., “Showing 1–100 of 342”).
  • Confidence Filter Tabs: Lets users filter the queue to show “All”, “High”, “Medium”, or “Low” confidence claims.
  • Bulk Action Bar: Appears dynamically when one or more claims are selected. It shows a count of selected items alongside “Approve” and “Reject” buttons for bulk processing.
  • Pagination Controls: Located at the bottom of the list, providing “Previous” and “Next” buttons along with a page count (e.g., “Page 1 of 4”).

​
Pending Claims List

This main list displays every claim awaiting review.
  • Checkbox: Selects the claim for bulk approve or reject operations.
  • Claim Text: The full text of the extracted fact or claim.
  • Pillar: The knowledge base topic this claim belongs to. It appears as a clickable link. If the AI couldn’t categorize it, this shows as “No pillar assigned” in italics.
  • Confidence: A badge showing the AI’s confidence level (high, medium, or low).
  • Source Type: A badge indicating where the claim came from (e.g., research, response, upload, or content_hub).
  • Source: An external link to the original webpage or document. Only visible if a source URL exists.
  • Source Quote: A blockquote excerpt from the original document that supports the claim. Only shown when available.
  • Approve (row action): A checkmark button to approve the single claim.
    • Tooltip (when assigned): “Approve”
    • Tooltip (when unassigned): “Assign to a pillar before approving”
  • Reject (row action): An X button to reject the single claim.
    • Tooltip: “Reject”

​
Drilldowns

  • Pillar Link: Clicking a blue pillar name navigates the user to the detail page for that specific knowledge base pillar.
  • Source Link: Clicking the “Source” link opens the original external URL in a new browser tab.

​
What you can do here

  • Filter claims: Click the Confidence Filter tabs to narrow the list by AI confidence level.
  • Select all visible claims: Use the master checkbox at the top of the list (next to the “reviewable” count) to select or deselect all claims currently on the page.
  • Bulk Approve: After selecting claims, click the “Approve” button in the top action bar. Note: This only approves claims that have a pillar assigned.
    • Tooltip (success path): “Approve {N} claim(s) across {N} pillar(s)”
    • Tooltip (if items lack pillars): “{N} selected claim(s) can’t be approved: no pillar assigned”
  • Bulk Reject: After selecting claims, click the “Reject” button to discard them entirely. This works even for claims without a pillar.
  • Approve/Reject Individually: Use the checkmark or X buttons on the far right of any specific row to process a single claim.
  • Navigate Pages: Click “Previous” or “Next” at the bottom to page through large queues. (Navigating clears your current selection).

​
Data shown

This page displays unreviewed, pending facts and claims that Athena’s AI has recently extracted from the customer’s uploaded documents, generated AI responses, and connected websites. It also references the names of the customer’s existing Knowledge Base pillars.

​
Common workflows

Bulk approve high-confidence claims:
  1. Open the page and click the “High” confidence tab to filter the queue.
  2. Click the header checkbox to select all visible high-confidence claims.
  3. Click the “Approve” button in the bulk action bar.
  4. A success toast will confirm how many claims were approved and assigned to their pillars.
Reject a single low-quality claim:
  1. Locate a poorly extracted claim in the list (often found using the “Low” confidence filter).
  2. Click the X (Reject) button on the right side of the claim’s row.
  3. The claim is instantly removed from the queue, confirmed by a success message.
Review claims page by page:
  1. Start on page 1 using the “All” filter.
  2. Read through the claims, approving or rejecting them individually, or selecting a batch to approve via the top menu.
  3. Click “Next” to advance to the next page (note that any lingering checkboxes are cleared when changing pages).
  4. Repeat until the queue is empty.
Filter and bulk reject unassigned claims:
  1. Leave the filter on “All” or pick a specific confidence level.
  2. Identify claims showing “No pillar assigned”.
  3. Check their checkboxes (you can select them even though they lack a pillar).
  4. Click “Reject” in the bulk action bar to permanently discard them.
Navigate to a pillar from a claim:
  1. Find a claim that has a blue pillar name link.
  2. Click the pillar name.
  3. The browser navigates away from the review queue and opens that pillar’s detail page.

​
Empty, loading, and error states

  • Loading: While fetching claims, the page displays six skeleton placeholder rows.
  • Empty (All Filter): If there are no claims pending at all, a dashed-border card appears stating “Nothing to review” and “All extracted claims have been reviewed.”
  • Empty (Specific Confidence Filter): If there are claims, but none match the selected tab, the card says “No [confidence]-confidence claims pending” and “Try a different confidence filter.”
  • Errors: Standard error toasts appear if an approval or rejection action fails due to network issues. If the page completely fails to load data, standard retry logic applies.
  • Linked from: The main Knowledge Base page (usually via a prominent “Review N pending” button in the header), and the Knowledge Base sidebar navigation.
  • Links to: Individual Knowledge Base pillar pages and external source URLs.

​
Common support questions

Why is the “Approve” button grayed out for some claims? A claim must be categorized into a Knowledge Base pillar before it can be approved. If the AI couldn’t figure out where it belongs, it will say “No pillar assigned.” You cannot approve these from the bulk review screen; you must either reject them or manually add them to a pillar elsewhere. Why did the page jump back to page 1? Whenever you switch a confidence filter or perform a bulk action (like approving a dozen claims), the queue recalculates and resets you to the first page. This prevents you from landing on an empty or out-of-range page if you just cleared out the page you were on. Where do these claims come from? They are auto-extracted by Athena when you upload documents, connect URLs, or when Athena generates new AI responses. The system tries to catalog useful facts automatically to keep your Knowledge Base up to date. I just uploaded a document, why don’t I see the high confidence claims here? High and medium confidence claims are usually auto-approved by the system. This queue is primarily a safety net for claims where the AI had low confidence or couldn’t safely auto-categorize the data.