Keeping your knowledge base healthy takes more than writing good SOPs once — content drifts out of date, gaps appear, and it's hard to know where to focus. Knowledge Health is the area of Whale that surfaces suggested improvements across your workspace, so you always know what to fix next.
Knowledge Health has two tabs: Recommendations (covered in this article) and Actions, which is coming soon and will let you act on recommendations automatically instead of one at a time.
*Knowledge Health is available on the Advance and Enterprise plans only.
📍 Where to Find It
Go to Knowledge Health in your Whale workspace navigation, then select the Recommendations tab.
🩺 What Recommendations Shows You
Alice reviews your workspace and surfaces things worth your attention, one at a time. Each recommendation shows a category (what kind of issue was found), a priority (High, Medium, or Low), and the specific content it applies to. Below it, an Up next list previews what's queued up after the current one.
Recommendations can come from a few different sources:
Health score — a published card or checklist whose content quality has dropped low enough to need attention (this includes content that's gone stale or hasn't been viewed in a long time)
Search gaps — questions people are searching for that your content doesn't answer well
Quiz failures — quiz questions that a lot of people are getting wrong, which may mean the underlying content needs work
Unanswered chats — questions Alice couldn't answer from your existing content
Note: workspace-level recommendations like search gaps, quiz failures, and unanswered chats aren't tied to a single board, so they're hidden whenever you filter by a specific board (see Filtering below).
📊 What Drives a Health Score
Opening a health-score recommendation shows the specific dimensions behind that score:
Quality — clarity, structure, and writing quality of the content
Coverage — how completely the content addresses the topic
Knowledge integration — how well the content connects to and reflects the rest of your knowledge base
Each dimension finding has its own Fix action, so you can address specific issues individually instead of only working from the overall score.
✅ Taking Action on a Recommendation
From the feed, you can:
Improve content — opens Alice with a focused prompt that loads the flagged content and asks her to review and fix the open findings
Skip — dismiss a recommendation that isn't relevant right now and move to the next one
Some recommendation types have their own dedicated fix path:
For content flagged as stale or rarely viewed, you can assign it to the people who need to review it — and optionally require an Electronic Signature, so the same action both gets the content re-reviewed and gives you confirmation that specific people have seen it. See Signing off on assignments with Electronic Signature for how signing works.
For an unanswered chat recommendation, you can answer the question directly from the Recommendations page — that answer is fed back to Alice, so the next person who asks gets a real answer instead of another miss.
Once you've handled a recommendation, mark it Fixed or Archived to remove it from your open list — Fixed for ones you resolved, Archived for ones you're deliberately setting aside. Either way, it's just a status change: switch the Status filter to see them again later (see Filtering below).
Published cards and checklists with open health-score findings also appear automatically in the Knowledge Health backlog, so nothing gets lost even if you skip it in the moment.
🔎 Filtering the List
Status — Open, Fixed, or Archived (defaults to Open)
Type — Card or Checklist
Board — narrow to one or more specific boards (listed alphabetically, with a "Show all" toggle once you have more than five)
What's wrong — filter by recommendation category (Health score, Search gaps, Quiz failures, Unanswered chats)
Priority — High, Medium, or Low (defaults to High)
Use Recommendations regularly to keep your knowledge base accurate and complete — it's the fastest way to see where your content needs attention without having to audit everything manually.