Assessing an existing help center with poor self-service rates and restructuring it for modern AI-powered support — with a clear before/after comparison, AI readiness scorecard, and actionable restructuring framework.
A growing SaaS company has a help center that was built organically over several years. Articles are outdated, inconsistently structured, and hard to find. The support team is overwhelmed with tickets for questions that should be answered by self-service content. The company wants to deploy an AI agent (like Intercom Fin) but the content isn't ready — AI can't surface answers from poorly structured articles.
The first step in any audit is showing stakeholders exactly where things stand — and what "good" looks like. This side-by-side comparison makes the case for restructuring concrete and actionable.
Before deploying an AI agent, you need to know if your content is ready. This scorecard evaluates each dimension of AI readiness — giving teams a clear picture of what to fix first and how much work is involved.
| Dimension | What AI Needs | Common Problems |
|---|---|---|
| Content Structure | Consistent heading hierarchy (H1 → H2 → H3), lists for steps, tables for comparisons | Walls of text, no headings, inconsistent formatting between articles |
| Heading Quality | Headings that describe the answer, not just the topic | "Billing" vs. "How Monthly Billing Works" — AI needs the specific framing |
| Topic Separation | One distinct topic per article; no mega-articles covering 5 things | FAQ pages that bundle unrelated questions; articles that drift between topics |
| Content Accuracy | Up-to-date information that matches the current product state | Screenshots from old UI, pricing that's changed, features that were removed |
| Audience Targeting | Articles scoped to a specific audience; language matched to their context | Admin-level content shown to end users; mixed jargon levels in one article |
| Metadata & Tags | Descriptive titles, meta descriptions, category tags, audience tags | Generic titles ("Help"), no descriptions, no tagging system |
After the audit, the help center gets a new taxonomy — fewer, clearer categories with logical groupings that map to how users actually think about their problems.
Every article in the help center gets one of four dispositions. This gives the team a clear, prioritized action list — not a vague "things need improvement" recommendation.
| Disposition | Definition | Article Count | Action Required |
|---|---|---|---|
| Keep As-Is | Content is accurate, well-structured, and AI-ready | 31 | Add metadata/tags, assign to new category |
| Update & Restructure | Good topic, but needs rewriting for accuracy, structure, or AI-readiness | 54 | Rewrite with standard template, update screenshots, add headers |
| Merge | Content overlaps with another article; consolidate into one | 24 | Identify primary article, merge content, redirect old URLs |
| Archive | Content is obsolete, irrelevant, or about a deprecated feature | 38 | Unpublish, set up redirects where applicable |
A practical checklist for reviewing each article before enabling it for AI agents. Every article passes through this checklist before being flagged as "AI-ready."
A help center audit isn't just about cleaning up old content — it's about building a foundation for AI-powered support. Here's the process:
I can audit your existing content, build an AI readiness plan, and restructure your help center for modern support.
Start a Conversation