Knowledge base and
support documentation skills

A knowledge base is the library of answers a support organisation runs on. External help centres let customers solve problems themselves; internal knowledge bases help agents give fast, consistent answers. The work includes finding which questions to answer from tickets and search data; writing clear, task-based articles; structuring them with titles, steps, tables and screenshots; following a consistent style; tagging and linking them; reviewing and retiring outdated content; measuring use and helpfulness; and working with product teams when features change.

The writing must suit how people read online. Nielsen Norman Group's research found most web users scan pages rather than reading word by word,¹ so help articles need answers first and short, scannable steps. Knowledge-Centered Service (KCS), developed by the Consortium for Service Innovation, makes knowledge part of everyday support: agents search first, reuse and improve articles, and capture new answers as they solve cases.² AI raises the stakes. Retrieval-augmented generation — the technique behind many support chatbots — answers questions by retrieving documents first,³ so wrong or stale articles become wrong bot answers. Style guides such as Google's developer documentation guide set useful standards.⁴ In India, AmbitionBox lists technical writers at a typical ₹8.1–8.9 lakh⁵ and knowledge managers at ₹11.7–12.9 lakh.⁶

Typical technical writer pay in India (AmbitionBox)

₹8.1–8.9 lakh

Typical knowledge manager pay in India (AmbitionBox)

₹11.7–12.9 lakh

Technical writer salaries reported in India (AmbitionBox)

10,000+

How most people read web pages (Nielsen Norman Group)

Scan

Table: How knowledge base candidates are assessed (editorial summary)

Table: Rewriting a help article (illustrative)

Table: How knowledge bases developed (editorial summary)

• Three sample help articles (how-to, troubleshooting, FAQ)
• A before-and-after rewrite
• A simple style guide for your articles
• A content review plan
• A list of top failed searches and your fixes
• A verified knowledge base certificate

Show impact: "Rewrote 60 help articles using ticket data; added customer search terms; tickets on top three topics fell after launch; set up quarterly reviews." Use your real results.

"How do you decide what to write next?"

Use ticket volumes, failed searches and low-rated articles to find the biggest gaps.

"What makes an article good?"

A task-based title, the answer first, short numbered steps, plain words and a clear next step.

"How do you keep articles up to date?"

Owners, review dates, alerts from product releases and agent feedback buttons.

"What's KCS?"

A method where agents capture and improve knowledge while solving cases, so content stays current.

"How does AI change knowledge work?"

Chatbots answer from articles, so content must be accurate, structured and free of contradictions.

• "Who owns knowledge content?" — accountability.
• "Do agents contribute articles?" — KCS maturity.
• "Is AI answering from the help centre?" — stakes.
• "What analytics are available?" — measurement.

Table: Knowledge-related roles and typical pay in India (AmbitionBox, October 2026; indicative)

Table: Common knowledge base tools (editorial guidance)

• Support agent contributor — writing first articles.
• Knowledge specialist — owning a help centre.
• Technical writer — product documentation.
• Knowledge manager — programmes and governance.
• Head of content or support operations — strategy.

1. Week 1: Write — Turn three common questions into articles.
2. Week 2: Prove it — Rewrite a weak public article; take the Knowledge Base Certification.

Related: Help Desk, Technical Writing and Knowledge Management. Guides: help desk, technical writing and knowledge management.

• Ticket review — finding gaps.
• Writing — new articles.
• Updating — after product changes.
• Analytics — searches and ratings.
• Agent feedback — fixing flagged articles.
• AI checks — testing bot answers.

"What's ticket deflection?"

When a customer solves the problem through self-service instead of contacting support.

"How do you write for search?"

Use customers' own words in titles and first lines, and add common synonyms.

Table: A first-90-days plan (editorial guidance)

Start with an audit: which articles get the most views, which get poor ratings and which searches return nothing. Rewrite the highest-impact articles first. By the third month, set up owners, review dates and an easy way for agents to flag or improve articles.

Support chatbots and agent assistants now answer questions by retrieving knowledge articles and generating replies. That makes knowledge content the foundation of AI service: contradictions, outdated steps or vague wording turn into wrong answers at scale. Knowledge specialists now test bot answers, structure articles so AI can use them, remove duplicates and track questions the bot can't answer. AI also helps draft articles from resolved tickets — useful, as long as a human checks accuracy.

Start with a title that matches the customer's task or question. Answer it in the first sentence. Use numbered steps for actions, one action per step. Add a table for options or timelines. Use screenshots only where they help, and keep them current. End with what to do if it doesn't work. Test every step yourself before publishing, and review the article whenever the product changes.

"Nobody reads help articles." People use them when articles are easy to find and scan.

"Write it once and you're done." Content decays with every product change.

"AI chatbots replace the knowledge base." Chatbots depend on it.

List every article with its views, rating, last update date and owner. Mark each one keep, update, merge or retire. Merge duplicates that answer the same question, retire articles for discontinued features and update anything older than your last major release. Doing this once a year — and before launching an AI chatbot — removes the contradictions that confuse both customers and bots.

• Company jargon in titles.
• Answer buried at the end.
• Outdated screenshots.
• No review process.
• Ignoring failed searches.

KCS comes from the Consortium for Service Innovation; reading behaviour from Nielsen Norman Group; retrieval-augmented generation from Lewis et al. (2020); style guidance from Google; and India pay from AmbitionBox (read October 2026). The refund scene, rewrite example and history are editorial. Science of Apps publishes the free test linked here. Last reviewed October 2026.

• Knowledge base — a searchable library of help articles.
• KCS — Knowledge-Centered Service, a method for capturing knowledge while solving cases.
• Self-service — customers solving problems without contacting support.
• Deflection — a contact avoided through self-service.
• RAG — retrieval-augmented generation, how many bots answer from documents.