📖 Detect KB gaps and auto-draft articles with GPT-4.1, Slack and Gmail
⚡ 39 views · 📖 Internal Wiki & Knowledge Base
Description
Who is this for?
Support teams, knowledge managers, and ops builders who are drowning in outdated KB articles and repeat tickets. If your team keeps answering the same questions because your knowledge base has gaps nobody has time to find — this template finds them automatically and writes the first draft for you.
What does it do?
This is a complete multi-agent AI pipeline — 31 nodes across 7 pipeline stages that automate the entire KB maintenance lifecycle:
- Gap analysis — An AI agent cross-references your recent support tickets against your existing KB articles and identifies what’s missing, what’s weak, and what’s stale.
- Automated article drafting — For every gap found, a second AI agent writes a complete KB article draft using your actual ticket data as context.
- Quality review — A third AI agent scores each draft on accuracy, completeness, and clarity — so you know what’s ready to publish and what needs work.
- Health scoring — A fourth AI agent generates a comprehensive health report with a 0-100 score, letter grade, executive summary, and prioritised action items.
- Dual notifications — Results are delivered via both Slack (rich blocks) and Gmail (polished HTML email) with color-coded health scores, gap tables, and action items.
- Staleness detection — Flags existing articles that haven’t been updated in configurable timeframes.
- Scheduled automation — Runs daily on a cron schedule (or manually on demand).
How it works
The Configuration node defines your helpdesk API endpoints, notification settings, and thresholds. The pipeline fetches recent tickets and current KB articles via HTTP, normalises the data, then sends it to the Gap Analysis Agent (GPT-4.1).
If gaps are found, each gap is split into an individual item and routed through the Article Drafter Agent → Quality Reviewer Agent pipeline. In parallel, a Staleness Check flags outdated articles.
All results merge into the Report Generator Agent, which produces a structured health report. The Format Outputs node transforms this into both Slack blocks and a polished HTML email, delivered simultaneously via webhook and Gmail.
If no gaps are found, the pipeline skips article drafting but still runs the staleness check and generates a health report.
How to use it
- Import the workflow JSON into n8n
- Open the Configuration node and set your helpdesk API URL, notification email, and Slack webhook
- Connect your OpenAI credentials to the 4 AI model nodes
- Connect your Gmail credential to the Send Email Report node
- Update the 2 HTTP Request fetch nodes with your helpdesk authentication
- Run manually to test, then activate the daily schedule
Required: OpenAI API key (GPT-4.1 and GPT-4.1-mini), Gmail account (for email reports) Optional: Slack incoming webhook (for Slack reports), any helpdesk with a REST API (Zendesk, Freshdesk, Intercom, etc.)
Tags
AI, AI Agent, Multi-Agent, LLM, OpenAI, GPT-4.1, Knowledge Base, Support, Helpdesk, Zendesk, Content, Automation, Gap Analysis, KB Management, Slack, Gmail, Report
Data Collection
> Ticket & KB Fetcher — Pulls recent support tickets and current knowledge base articles from your helpdesk API. Data is normalised into a consistent format for AI analysis. Update the HTTP Request nodes with your helpdesk URL and authentication.
AI Analysis Pipeline
> Multi-Agent Gap Analysis — 4 specialised AI agents work in sequence: (1) Gap Analysis identifies missing, weak, and stale content by cross-referencing tickets against KB articles. (2) Article Drafter writes complete KB article drafts for each gap. (3) Quality Reviewer scores each draft. (4) Report Generator produces a health score, executive summary, and prioritised actions.
Output & Notifications
> Dual Notification System — Results are formatted into both Slack blocks (with emoji health grades, key findings, and action items) and a polished HTML email (with color-coded score circle, gap tables with urgency badges, and action tables). Both are delivered simultaneously. Either channel can be disabled in the Configuration node.
Staleness Detection
> Article Freshness Check — Runs in parallel with the main pipeline. Flags any KB article not updated within the configurable threshold (default: 90 days). Results feed into the health report regardless of whether gaps were found.
Setup Note
> ⚠️ Credential Setup Required: After importing, you must configure: (1) OpenAI credentials on 4 AI model nodes (2x GPT-4.1, 2x GPT-4.1-mini), (2) Gmail OAuth credential on the Send Email Report node, (3) Your helpdesk API URL and auth on 2 HTTP Request nodes, and (4) Your Slack webhook URL and email address in the Configuration node. See the Setup Guide for a step-by-step checklist.
🔗 Nodes Used
HTTP Request, Gmail, Schedule Trigger, AI Agent, OpenAI Chat Model
📥 Import
Download workflow.json and import into n8n:
Workflow menu → Import from File