Skip to main content

Managing Information Requests

When users can't find an answer through the AI assistant, they can submit an information request. These requests surface knowledge gaps and give your team a direct channel to improve the knowledge base.

What are requests​

An information request is created when a user clicks the "Request Knowledge" or similar action in the assistant. Each request captures:

  • The user's question
  • The page URL and domain where they were
  • The page context (form fields, page structure) at the time
  • A timestamp and the user's identity

Requests are a feedback loop: they tell you exactly where users need help and what content is missing.

Who sees what is settled on the server, not in the interface. Within an organization an administrator sees the whole inbox and moderates it; anyone else sees only the requests they filed themselves. A caller whose account belongs to no organization is refused rather than shown an unfiltered list.

Stats dashboard​

At the top of the Requests page, four stat cards summarize the current state:

StatDescription
TotalAll requests ever submitted
PendingRequests that haven't been reviewed yet
In ProgressRequests currently being worked on
ResolvedRequests that have been answered or addressed
note

A high Pending count signals that your team should prioritize request triage. Aim to acknowledge new requests within 24-48 hours.

Filtering​

By domain​

Select a domain to see requests only from that application. Useful when different team members own different domains.

By URL​

Search for requests from a specific page or URL pattern to find clusters of related questions.

By status​

A request carries one of five statuses, and the same five words are used by the Backoffice, the backend and the asker's own view in My Account.

StatusMeaningWho sets it
PendingFiled and not yet reviewedThe system, on creation
In progressSomeone is working on itYou
ResolvedAnsweredYou
RejectedDeclined, with a reason if you gave oneYou
HarvestedA page-structure row written by the harvest endpoint, not a question anybody askedThe system

Request detail view​

Click on any request to open the detail view. You'll see:

  • User details: who submitted the request
  • Question: the exact question asked
  • Page context: the URL, domain, and captured page structure
  • Status: current status with history of changes
  • Notes: internal notes added by your team
  • AI Suggestion: an AI-generated draft response (if available)

AI Suggestion feature​

For each request, Appilot can generate an AI Suggestion: a draft answer based on existing knowledge base content and the page context. This helps you:

  • Quickly assess whether existing content partially answers the question
  • Use the suggestion as a starting point for a new knowledge article
  • Identify whether the gap is a missing article or an article that needs updating
tip

The AI Suggestion is a draft; always review and edit it before publishing as official knowledge content.

Workflow​

Changing status​

Update a request's status as you work through it:

  1. Pending → In Progress: When you start reviewing the request
  2. In Progress → Resolved: When the question has been answered (either by creating content or determining no action is needed)

Moving a request to Resolved or Rejected records who did it and when, and only on the transition, so revisiting a finished request does not rewrite the credit. That same transition writes a notification to the person who asked, in the language their request carried, linking to the request in their My Account.

What the asker sees​

A request carries two answer fields, and whatever is in them appears in the asker's own view under their question.

  • The published article: the knowledge entry written in answer to the request. The asker gets a link to read it. The Backoffice fills this in for you when you publish an article from a request.
  • A written answer: free text, and on a rejection the reason. The Backoffice has no field for it yet, so today it is set through the API. Until it is, a rejected request tells the asker plainly that no reason was given.

Creating content from requests​

The most valuable action is turning requests into knowledge:

  1. Review the request and AI Suggestion
  2. Click Create Knowledge (if available) to open the knowledge editor pre-filled with context
  3. Write or refine the knowledge article
  4. Publish the article

Publishing from a request closes the loop itself: the request moves to Resolved, the new article is attached to it, and the asker is notified. If that last step fails, the article is still saved and the request stays open with a warning, so you can set its status by hand.

Next steps​