Workflow · Customer insight to content
From what customers ask to the article they need
Real work Rebuilt from a Claude plugin I built for a content team
The article search tells you whether coverage exists. The support history tells you what customers ask, in their own words, and where the chat bot already fails. This workflow reads the conversations, groups them by question, finds the proven gaps, and drafts the missing article with its evidence beside every line.
Try it
- Select Mine the support history.
- Pick a question to read the evidence.
- Approve the plan, and the workflow drafts the article.
Not run yet
What this workflow cannot tell you
- How many customers had the problem and never asked. Count conversations as a floor, never a total.
- Questions raised later in a conversation. Most search tools only match the opening message.
- Whether a support answer is right. A reply is evidence of the question, and only sometimes of the answer.
Safeguards
What keeps the workflow safe to run on a real knowledge base.
- Personal data removed firstNames, emails, card numbers and order numbers are masked before any conversation is read or quoted.
- The search is read onlyIt reads conversations and writes nothing back to the support tool.
- You approve the planNothing is written until a person approves the title, the scope and the evidence.
- Facts are checked in the productAnything a support reply says about how the product works is confirmed on screen before it goes in.
- A draft, never a publishThe workflow saves a draft for review. A person publishes it.
- Evidence stays with every lineEach sentence shows where it came from, so a reviewer can check it in seconds.
Take it to the next level
How to get results that match the best teams, and what to build next.
- Read every channelAdd call transcripts, reviews, community posts and failed searches, so the picture is not only from people who wrote in.
- Group with careLet artificial intelligence (AI) suggest the groups, then have a person name and merge them, and keep that labelled set to check future runs.
- Measure the resultAfter the article goes live, track how often the chat bot hands that question to a person. It should fall.
- Close the loopTell the support team when the article is live, and give them the link to use in replies.
- Feed the auditSend every proven gap to the audit report as a task, so the backlog shows real reader needs.
- Spot trends earlyRun it weekly and flag any question whose volume doubles, which is often the first sign of a product problem.
Rebuilt from a Claude plugin I built for a real content team. Keystone is an invented payments product, and every name, figure and conversation here is made up.