BLOG
Comup Perspectives
Data and judgment for IT and customer-experience leaders: what AI is already achieving in real service operations.
Data and judgment for IT and customer-experience leaders: what AI is already achieving in real service operations.
Almost every organization has a knowledge base. Almost none has one that people actually use. The difference is not the platform: it is the process used to build it. A base that starts by uploading entire folders of documents ends up as a file cemetery — and a cemetery resolves no tickets.
Useful knowledge already exists, but it is scattered: documents on shared drives, manuals in PDF, repeated answers in email threads, and the experience of two or three people who always know. The first step is to inventory it and cross it with real demand: the questions that repeat most in tickets are the base's initial index. What nobody asks about is not urgent to document.
Source documents are never published as-is. Before becoming articles, they go through a sensitivity review: credentials and passwords that appear in manuals and screenshots, personal data, internal addresses, obsolete versions. Publishing a password inside a procedure manual is a security incident disguised as documentation — and it is more common than any CIO would like to admit.
articles published without approval: all content is born as a draft and is only published once the knowledge owner validates it.
Not all knowledge is for everyone. Requesters need short, safe answers; the resolving agent needs the full procedure, including the sensitive steps. The right structure separates both layers with folder-level visibility: what is public goes to the portal, what is internal is visible to agents only. Mixing them produces the worst of both worlds: either internal information leaks, or everything is hidden and nobody finds anything.
In operations serving more than one country, the base lives in two languages or dies in one. Real bilingual content means every article is born — or reviewed — in both languages, with the same owner and the same update cycle. Translating once and forgetting leaves two bases: one alive and one fossil.
Every article has three things: an owner, a review date and a status. It is born as a draft, validated by whoever knows the process, and published only with that approval. And the cycle continues: every ticket a person resolves without finding an article is a signal of what to document or fix.
A knowledge base is not a file-upload project: it is an editorial system with a security mindset. Built this way, people use it — and the AI that will answer for your organization tomorrow will finally have something solid to learn from. This is how we build them at Comup.