Help center gap board built from your Gladly conversations

Build me a help center content workspace for the person who owns our support content. They sit down once a week and decide what to write next, so the app opens on a ranked gap list: recurring customer questions that our published Gladly Answers do not cover.

The main view is a gap table sorted by how many separate conversations mention the question. Each row shows the gap title in the customer's own phrasing, the conversation count, the themes it clusters with, the closest existing published answer if there is one along with how close that match looks, and verbatim excerpts from the real conversations behind it. Clicking a row opens a detail panel with the full evidence list and the drafted article if one exists.

Build the picture from Gladly. Gladly has no org wide conversation list, so seed the mining from List Tasks to find which customers have recently been in touch, then call List Customer Conversations for each of those customers and List Conversation Items for each conversation to read the actual messages. Use Get Customer to label each excerpt with who said it. Also give the owner a text box to paste customer emails, resolved through Find Customers, so they can point a scan at a specific group. For the coverage side, call Get Organization first to resolve the org id, because the public answers endpoints live under it, then use List Public Answers and List Public Answer Titles to load what is published, Search Public Answers to test whether a given question is covered, and Get Public Answer to pull the full body of the closest match.

A second view is a browser over everything already published: a searchable, filterable list of our published answers backed by List Public Answers and Search Public Answers, with the full article body from Get Public Answer when a row is opened. The point of this view is that the owner can spot answers that are thin or out of date rather than missing, so show article length, when it was last updated, and which gaps point at that answer as their closest match.

Add a "Find this week's gaps" button that starts a background agent. The agent reads the recent conversations gathered from Gladly, clusters them into recurring questions using the customers' own phrasing, runs Search Public Answers for each cluster to judge whether we genuinely cover it, and for the uncovered clusters drafts a full help center article in our voice: a title in the customer's words, a short direct answer at the top, then the steps or explanation. It writes everything back into the app as reviewable drafts attached to their gap rows, together with the supporting excerpts and the coverage judgement it made. Show progress while it runs and let the owner keep working in the app.

The owner edits a draft inline in the app and then either approves or dismisses it. Approve files the article into our content backlog in Notion: Create a Page in the backlog database carrying the gap title, the conversation count and the theme, then Update Page Content as Markdown to write the draft body followed by a section of the supporting conversation excerpts. Store the resulting Notion page link on the gap row so the app shows at a glance what has already been filed. Dismiss marks the gap as not worth writing, with an optional reason.

Persist gaps, their evidence excerpts, drafts, approvals and dismissals inside the app so the workspace carries across sessions and reruns. Dismissed gaps must never resurface in a later scan, and a rerun should merge new evidence into an existing gap rather than creating a duplicate row. Rank strictly by the number of distinct conversations mentioning the gap rather than raw message count, so one chatty thread does not outrank a question ten different people asked. Keep the excerpts attached to every row at all times, because the writer needs the real wording. Note that Gladly's Answers API is read only, so approved drafts land in Notion as the publishing backlog and a person publishes the final article in Gladly.