[Sharing] 3 SMB workflows built on an approval-first pattern (news digest → drafts, lead scoring, blog → social queue) — free workflow JSON included

Built with n8n
1

Hi everyone,

Sharing a set of three workflows I built for small-business use, all following one architectural pattern I haven’t seen discussed much here: an explicit human approval gate between the LLM output and any outward action.

Instead of letting the workflow post, send, or reply, every outward-facing artifact lands in a Google Sheet with a status field, and a human moves it forward. Boring, but it’s the version non-technical owners actually keep switched on.

Workflow 1 — AI News Digest → Blog Drafts (free sample)
Schedule (08:00) → RSS Read → Limit (10) → HTTP Request (OpenAI chat completions, gpt-4o-mini) → Google Sheets (append) with status DRAFT - needs human review → Slack notification. The prompt asks for headline, summary, and a “why it matters to your customers” line per item.
Free JSON: Free Sample — AI News Digest n8n Workflow

Workflow 2 — Lead Capture → AI Qualification
Webhook → HTTP Request (LLM returns score 0–100, reason, suggested next step as JSON) → Google Sheets (append, all leads) → IF (score >= 70) → Google Sheets (hot leads) + Slack alert. Logging every lead unconditionally was a deliberate choice — scoring should prioritize, not filter out.

Workflow 3 — Blog → Social Approval Queue
Schedule (Mon 09:00) → fetch latest post → one LLM call producing three formats (X thread / LinkedIn post / newsletter blurb) → Google Sheets approval queue, status WAITING_APPROVAL. No posting nodes exist in this workflow at all.

Implementation notes:

  • Standard HTTP node for the LLM call — any OpenAI-compatible endpoint works by swapping URL + key.
  • All credentials, sheet IDs, and feed URLs are placeholders (REPLACE_, YOUR_) so the JSON imports cleanly on n8n 1.x.
  • Sheets double as both the log and the review UI — for the target user, that’s a feature, not a compromise.

These three, plus a full setup guide and a 30-prompt library for SMB work (content, support, sales, ops, hiring), are also bundled as a paid starter pack here: SMB AI Automation Starter Pack — 3 n8n Workflows + 30 Prompts — but the first workflow is free above, and I’m happy to answer structural questions in the thread regardless.

Curious how others here implement approval gates — Sheets, n8n’s Wait node, Telegram buttons? What’s held up best for non-technical clients?

2

Logging every lead and letting the score prioritize instead of filter is the right call. One addition from a review queue I built for a client (a supplier price-list matcher): store the rejections as data too. When the reviewer says “not this one”, that pair is never suggested again. In the public demo of that project, the next month’s review queue had less than half the questions.

With your Sheets setup, a REJECTED status plus a reason column, and a lookup against it before a new draft or lead goes into the queue, does the same. Otherwise the owner answers the same question every week and eventually switches the workflow off.