Knowledge Gaps and Suggestions

Aigento watches how your tickets go and points at what your knowledge base is
missing. It does this in two ways, both on the Knowledge Base page:

• Knowledge Gaps are topics customers keep asking about that your knowledge
  base does not answer. A gap is a to-do: write the missing article.
• Suggestions are ready-made article drafts written by the AI, for you to
  review and accept or reject.

Knowledge Gaps, and Suggestions written from tickets, need AI turned on with
knowledge base search enabled (Enable AI Features and Knowledge Base
RAG under Settings → AI).

What Knowledge Gaps are

A knowledge gap is a group of at least three recent tickets that ask about the
same thing, where your knowledge base had no answer or only a weak match. Each
gap shows:

• A topic: a short description of what customers needed.
• A source badge: the main reason the tickets were flagged (see below).
• A count line, for example "4 escalations: not in the knowledge base · 2
  tickets with no matching doc". When the gap was first found from resolved
  tickets, it also shows their average resolution time.
• A priority: low, medium, high or critical, set by the AI. The
  list shows the highest priority first, then the gaps with the most tickets.

Where Knowledge Gaps come from

Every night, Aigento looks at the last 7 days of tickets and collects the ones
that point at missing knowledge. In a busy workspace the AI groups the 60 most
recent of them. Each ticket counts once, under the first of these that
applies:

A resolved ticket nobody opened in Aigento has no knowledge base match on
record, so it is not counted. Spam is left out, and so is any ticket that is
already part of a gap, whether that gap is open, dismissed or addressed.

The AI then groups the tickets that ask about the same thing. A group needs at
least three tickets to become a new gap, so one odd question does not create
noise, and a single scan reports at most eight new or updated gaps. When new
tickets match a gap that is already open, Aigento adds them to that gap
instead of creating a second one; even a single new ticket can join it.

When gaps appear

The nightly scan runs on its own. A new gap shows up after the first nightly
scan that finds at least three similar tickets within the last 7 days. Tickets
older than 7 days are not counted.

Owners and admins can also click Scan now to run the scan immediately. It
can run once an hour per workspace, and every click counts, even one that finds
nothing. The message after a scan tells you what happened:

• "Scanned 25 tickets: 1 new gap, 2 updated." means the scan found something.
• "Scanned 25 tickets: no new gaps." means the tickets formed no new group of
  three on one topic and added nothing new to an open gap.
• "Not enough escalations or unmatched tickets in the last 7 days to find a
  gap." means no new ticket qualified, or fewer than three did and there is
  no open gap they could join.
• "Turn on AI and knowledge retrieval to scan for gaps." means AI or knowledge
  base search is switched off for the workspace.
• "The AI could not finish the scan. Try again later." means the AI call
  failed.
• "Gaps were scanned recently. Try again in N min." means the hourly scan was
  already used.

What "No open gaps" means

"No open gaps" means nothing is waiting for you: either the scan found no
group of three similar unanswered tickets, or every gap it found has already
been handled. It is not an error. Owners and admins see this message with the
Scan now button; other team members see the panel only when there is a gap
to show.

What to do with a gap

Any team member can act on a gap:

• Create Article opens a new draft article titled from the gap, with a
  short template to fill in, and removes the gap from the list. The draft is
  customer-facing. The AI does not use it until you write the content and set
  its status to Active.
• Dismiss (the X button) removes a gap that is noise or not worth an
  article.

The tickets behind a gap are never counted again, so a dismissed or handled
gap does not come back from the same tickets. The topic can return as a new
gap only if at least three new tickets ask about it and your knowledge base
still has no active article that answers them.

A No tool gap shows "The AI had no tool for this" and offers only Dismiss.
An article cannot fix it: Smart Reply needed to do something, such as look up
or change an order, and had no way to. The fix is to connect the integration
or set up the action it needed.

Owners and admins can also answer a gap by importing content with Smart
Ingest, for example a page from your website, and then dismissing the gap.

Suggestions

Suggestions work the other way round: instead of naming a missing topic, the
AI writes the article for you.

About once an hour, Aigento picks up to three of your tickets resolved in the
last 24 hours that had at least two outbound messages (replies from your team
or the AI, or internal notes). If the knowledge base does not already cover
the ticket's topic well, the AI drafts a new article from how the ticket was
handled. Each ticket is looked at once and produces at most one suggestion,
and most resolved tickets do not get one.

Suggestions also come from content you import. Open ones appear in the
Suggestions panel at the top of the Knowledge Base page with a Ready,
Pending or Processing badge. View all opens the full list on the
Smart Ingest Suggestions page, where you can filter by status.

Open a ready suggestion to review it. You can edit the title, type, category,
visibility and content, then:

• Accept & Create adds it to your knowledge base as an active article
  with the visibility you chose. A suggestion from a page on your website is
  Customer-facing by default, so the AI can quote it to customers. One
  written from a ticket or from pasted text is Internal by default:
  internal articles guide the AI but are never quoted to customers.
• Accept & Update appears instead when the suggestion updates an existing
  article. The article keeps its visibility.
• Reject marks the suggestion rejected and changes nothing in your
  knowledge base.

A suggestion written from a ticket is based on the ticket's subject and every
reply and internal note on it, including the AI's. Read it for customer names,
order details and anything private before you make it customer-facing.

Gaps and Suggestions side by side