Decide who creates, maintains and reviews help articles before launch, starting from the knowledge base you have today.
- Time to set up
- One planning session, then a weekly slot
- Cost
- £0 to start
- Driver
- Process
The problem
Most knowledge bases are written at launch and updated only when a customer complains. An AI agent then repeats whatever the articles say, accurate or not. Gartner found that the most common reason self-service fails, in 43% of cases, is that customers cannot find content relevant to their issue.
The right set-up depends on how your teams work together and the resources you have. Some organisations have a knowledge manager or a technical author. In others, the work is shared across the support team. In either case, the decisions about ownership need to be made before launch, not after it.
The fix
- Start from where you are. If there is no help centre, make a team leader responsible and write articles for the five to ten most frequent issues that have a known solution and are not caused by bugs or defects. If the help centre holds only product documentation, add troubleshooting content and explain what has changed in each release. If it serves onboarding only, add technical support sections and share the writing across the support team, using a standard template.
- Decide who maintains the content, not only who writes it. Name one owner, even where the writing is shared, and protect time for the work in the rota.
- Agree what prompts a new article. Escalations to senior team members, regular reviews of ticket data, gaps flagged by agents, product releases and customer ratings on articles should all feed the list.
- Review every article before it is published. A team leader or senior engineer checks it against the live product.
- Set a review date for every article, at least annually, and review the most-used articles first and most often.
- Agree the lifecycle before you buy a tool. If you are selecting a knowledge platform, make the ability to manage creation, review and retirement a technical requirement.
Tools to try
- Your helpdesk reporting to find the most frequent issues with a known solution. A spreadsheet also works if tickets are tagged, although it takes longer.
- An AI assistant such as Claude or Copilot to review exported ticket data and suggest which articles would have the greatest impact.
- The AI features in most knowledge tools, which can flag articles that need updating and frequent questions that have no article yet.
How to measure it
Track the share of articles with a named owner and a review date (aim for all of them), and ticket volume on the topics your new articles cover, before and after. At one high-volume B2B software and hardware company where I led support, ticket volume fell by 20% once this approach was in place, measured against the baseline.
Why this matters: You don’t have an AI problem. You have a content problem
Want help setting up a knowledge base that stays current?
Book an intro call