Agent Knowledge Freshness Review
Every source behind an agent with its owner and review date, superseded versions still being cited, and a rebuild once an owner approves.
These images are illustrations of the concept, not screenshots of the actual product.
Overview
In AI and automation teams that run customer-facing agents, an agent is only as current as the documents behind it. Policies get a new version, help articles pass their review date, and an older copy that is still indexed keeps being cited, so two answers to the same question can quietly disagree. This Botlit concept shows a knowledge freshness review for one agent, designed so support and AI operations teams can see what the agent trusts and fix stale sources before a customer finds them.
The illustration shows a Knowledge freshness page for a sample support agent, reached from AI Operations in a sample agency workspace. Filter tabs switch between all sources, overdue reviews, conflicts and superseded documents, and four tiles count the sources, the reviews overdue, the conflicting passages and the answers in the last seven days that cited a retired source. A Sources table lists each document with its owner, when it was last reviewed, when the next review is due, a status and how many answers cited it in the last seven days. Most rows are marked Current and a help article is Overdue, while the older version of a refund and dispute policy is highlighted as superseded but still indexed, with answers still citing it.
A Conflict found card sets the two versions side by side: the newer one allows disputes within 14 days and the older one within 30 days, with a count of the answers that cited the older passage. Buttons retire the older version and rebuild the index in one action, or open those answers for review. The design envisions owners retiring, replacing or re-approving a source, with each approved change queuing an index rebuild.
Two other Burdenoff products appear in the design. A Live figures card marks seats in use per account as coming from a versioned SemanticFed metric, with a badge showing it replaces copied text, so a figure that changes often is read from the metric rather than pasted into a document. A Review reminders via FluidGrids card lists a reminder already sent to one owner and an upcoming review for another; a FluidGrids workflow is designed to send those reminders so reviews do not depend on anyone remembering. The concept is meant for AI operations leads, support knowledge managers and the document owners behind each source.
What this concept shows
- Every source behind an agent listed with its owner, last review, next review, status and seven-day citation count
- Filter tabs for all sources, overdue reviews, conflicts and superseded documents
- Tiles counting sources, overdue reviews, conflicting passages and answers that cited a retired source
- A superseded policy version flagged as still indexed, with the answers that cited it
- A conflict card comparing the dispute windows in two versions of one policy
- A one-step retire and rebuild index action, beside a button that opens the answers citing the older version
- Often-changing figures marked as read from a SemanticFed metric instead of copied text
- Owner review reminders sent through a FluidGrids workflow
How it works
- An AI operations lead opens the knowledge freshness review for one agent and reads the tiles for overdue reviews, conflicts and retired-source answers.
- The Sources table shows each document's owner, review dates, status and how often it was cited in the last seven days.
- A conflict card sets contradicting passages side by side, counts the answers that cited the older one and opens them for review.
- The owner retires, replaces or re-approves the source, and an approved change is designed to queue an index rebuild.
- A FluidGrids workflow is designed to remind owners when a review falls due.
- Figures that change often are pointed at a SemanticFed metric, so the agent reads them rather than quoting copied text.
Who it's for
- AI operations leads
- Support knowledge managers
- Policy and document owners
- Customer support team leads
Illustrations
1 illustration of this concept. Select one to view it full size.
Knowledge Freshness With a Policy Conflict
This desktop illustration shows the Knowledge freshness page for a sample support agent in Botlit, reached from AI Operations in a sample workspace. Filter tabs for all sources, overdue review, conflicts and superseded sit above four tiles counting sources, overdue reviews, conflicting passages and retired-source answers in the last seven days. A Sources table lists five sample documents with owner, last reviewed, next review, status and seven-day citations: three are Current, a help article is Overdue, and a highlighted older version of a refund and dispute policy is marked superseded and still indexed. On the right, a Conflict found card compares a 14-day and a 30-day dispute window across the two versions, with a cited-by count and buttons to retire the older version and rebuild the index or open the answers. A Live figures card marks seats in use as a SemanticFed metric that replaces copied text, and a Review reminders via FluidGrids card lists a sent reminder and an upcoming review.
Topics
- knowledge base freshness
- stale knowledge base content
- chatbot source review
- RAG document review dates
- conflicting knowledge base passages
- superseded document still indexed
- AI agent knowledge governance
- index rebuild after policy change
- knowledge owner review reminders
- support agent content audit
Related concepts
Grounded Answers with Cited Sources
Agents answer from your own knowledge bases, show the passages they used, and say so plainly when nothing matches.
1 illustration
AI Operations Findings and Bot Drafting
An operations copilot that flags regressions across your bots and drafts a new one from a plain-language description.
1 illustration
Content Policies and Guardrails
Workspace policies that block unsafe input and output, cap request rate and token spend, and list recent enforcement events.
1 illustration
Part of an industry solution
This concept appears in a cross-product solution on burdenoff.com — see how it works alongside other Burdenoff products to solve a problem in that industry.
