DEV Community: AI Automation Playbook

I Ran the "Broken Template" Checklist on My Own n8n Workflows — Here's What I Found

AI Automation Playbook — Thu, 22 Oct 2026 04:02:36 +0000

My Own Workflows Failed the Broken-Template Test. I Fixed That.

This week two dev.to posts made the rounds: one author dissected why imported n8n AI workflows silently break, another claimed to have audited the 200 most popular n8n templates and found hundreds of HTTP Request steps with no retry logic — one API hiccup kills the whole run.

My first reaction was to feel smug. My second was to grep my own catalog.

The audit

I ran the same checklist on every workflow I sell — 8 product files, 13 HTTP Request nodes across the Workflow Pack, the free Starter Pack, and the AI Outbound Engine.

Result: zero retry configuration on all 13 nodes. Every single one would die silently on the first transient API failure. The same flaw I was nodding along to in someone else's audit.

Two lines I will not defend: my n8n.io templates (5 approved, 1 under review) were deliberately left untouched — changing them means re-entering the official review queue, and they passed review as-is. That's a scope decision, not a quality claim. What I sell on Gumroad is what I fixed.

The fix

Each of the 13 nodes now has:

retryOnFail: true
maxTries: 3
waitBetweenTries: 5000 (5 seconds between attempts)

Nothing else changed. Then I re-ran my full QA gate on all 8 files: structural lint (every connection references an existing node), real n8n import:workflow on a local instance, and syntax checks on every Code node. 8/8 PASS. The three affected deliverable ZIPs (Workflow Pack, Outbound Engine, free Starter Pack) were rebuilt from the fixed JSONs.

One line that is now true on my product pages: every HTTP request node in these workflows retries automatically (3 attempts, 5s apart) — one API hiccup won't silently kill your run. Before today, I couldn't honestly write that. Now I can.

Why I'm writing this

Selling automation is a trust business. My product pages already carry my n8n Verified Creator badge, but a badge earned last week doesn't cover a bug found this week. The audit happened because someone else published theirs. I'd rather be the seller who publishes the bad news first.

If you sell n8n templates, run the same grep. It's free, it takes minutes, and you might find the same thing I did.

✅ n8n Verified Creator — 5 workflows approved by n8n's official template review: workflow 1 · workflow 2 · workflow 3 · workflow 4 · workflow 5

Want more automation playbooks like this? The free newsletter breaks down one workflow a week: https://aiautomationplaybook.substack.com

P.S. The email + funnel tool I use is Systeme.io — 60% lifetime commissions, my affiliate link.

Disclosure: drafted with AI assistance. All audit numbers are from my own catalog, measured 2026-10-21 — no third-party stats cited beyond attributed dev.to posts.

Google's Nov 11 OAuth Deadline for AI Agents — What the Official Docs Actually Say

AI Automation Playbook — Thu, 22 Oct 2026 03:45:03 +0000

The Nov 11 OAuth Change, Checked Against Google's Own Documentation

A checklist about Google's new "agentic" OAuth flag made the rounds this week. It was useful — but its author openly said the details came from secondhand reporting, not from Google's announcement itself. So I read the official Google Workspace release notes and the "Manage OAuth Clients" documentation directly. Here's what checks out, what needs narrowing, and what n8n builders actually have to do.

What Google's official docs confirm

From the Google Workspace developer release notes and Google's Manage OAuth Clients page:

The checkbox is real. In Google Auth Platform > Clients, an OAuth 2.0 client can be designated for AI agents by selecting "Use this client for an AI-powered agent."
November 11, 2026 is the enforcement date. Starting then, Google Workspace MCP servers reject requests from any app or integration whose OAuth client is not marked as agentic.
November 12 is the lock-in. OAuth clients not marked agentic by November 12, 2026 remain non-agentic and can't call Workspace MCP servers. After that, the fix is a brand-new OAuth client marked as agentic, plus updating every config that referenced the old one.
The flag grants nothing extra. Per Google's own support doc: tagging a client does not grant additional scopes or privileges. It's a label — it lets Workspace admins apply agent-specific access controls without touching human-user traffic.

What the viral version got too broad

The headline framing ("Google Workspace will block AI agents") is wider than the actual enforcement. Google's release notes are explicit:

"Apps and integrations that are not AI agents, like those that only call Google Workspace REST APIs, are not affected by this change."

Translated for n8n builders:

Your built-in Google nodes (Gmail, Sheets, Drive, Calendar via REST) are untouched. Nothing to change, nothing to warn buyers about.
If you authenticate through n8n's hosted OAuth service instead of your own Google Cloud project client, the flag lives on a client you don't own — this change doesn't reach you.
What IS affected: your own OAuth clients used to call Google Workspace MCP servers (the Gmail/Drive/Docs/Sheets/Calendar MCP endpoints), and any agent tooling configured with your own client IDs.

One more detail the summaries missed: for newly created clients, Google's create-credentials doc says you can't change this setting after creation. Decide your client structure before you click Create.

The 10-minute checklist (official-source version)

Inventory. Google Cloud console > Google Auth Platform > Clients. List every OAuth client ID you own.
Classify. For each: does anything use it to call a Workspace MCP server, or to run agent tooling with your own client? If neither, leave it alone.
Split before you tag. If one client serves both human-interactive features and agent/MCP traffic, Google's documented best practice is separate client IDs — one standard, one agentic. Don't just tag the shared one.
Tag the agent clients now. The toggle has been editable on existing clients since October 12. Do it before the November scramble, and before the November 12 lock makes it permanent.
Re-check your MCP configs. Any mcp_config.json or MCP client setup referencing your client IDs must point at the agentic one after the split.

That's it. No one's workflows break on November 11 unless they call Workspace MCP servers through their own untagged client — a narrower group than the headlines suggest, but a group with a hard, one-way deadline.

✅ n8n Verified Creator — 5 workflows approved by n8n's official template review: workflow 1 · workflow 2 · workflow 3 · workflow 4 · workflow 5

Want more automation playbooks like this? The free newsletter breaks down one workflow a week: https://aiautomationplaybook.substack.com

P.S. The email + funnel tool I use is Systeme.io — 60% lifetime commissions, my affiliate link.

Disclosure: drafted with AI assistance. Every factual claim above was checked against Google's official release notes and support documentation on 2026-10-21.

I built an n8n workflow that rebooks no-shows automatically — here's how it works

AI Automation Playbook — Wed, 14 Oct 2026 06:53:05 +0000

Every empty chair in an appointment book is money walking out the door. Missed appointments cost US healthcare an estimated ~$150B a year (WriteUpp, 2026) — and for a dental clinic, salon, or home-service business, the math is just as brutal at a smaller scale. Most no-shows never rebook on their own.

So I built a fix: an n8n workflow that watches the appointment book around the clock and rescues no-shows over SMS. It's now live as a ready-to-import workflow, and I'm an n8n Verified Creator — here are 3 of my approved templates: https://n8n.io/workflows/20250, https://n8n.io/workflows/20580, https://n8n.io/workflows/20743.

How it works

Three flows, one Google Sheet as the brain:

1. No-show detection (every 30 minutes). The workflow reads today's Appointments sheet and flags anyone whose time passed 15+ minutes ago without a check-in. Appointment times are interpreted in your business timezone (a timezone field in the CONFIG node) — not the server's — so a Vancouver clinic on n8n Cloud doesn't get false flags.

2. AI rescue SMS. For each no-show, it pulls up to 3 open slots from an OpenSlots sheet, and an AI agent drafts a personal text: "Hi Sarah, we missed you at 2pm — want Thu 10am or Fri 3pm instead?" Sent via Twilio, logged to a RescueLog sheet. The model is selectable — defaults to gpt-4o-mini, one CONFIG field switches it.

3. Reply handling + one gentle nudge. A Twilio webhook parses replies: YES (+ slot number) books the slot automatically, NO is logged as lost, anything unclear goes to human review. No reply after 2 hours gets exactly one nudge — then silence. The nudge is self-limiting by design: once a nudged row is logged, that number is never nudged again.

The details that matter

Reply parsing is defensive. "NO 2" is a no, not a yes-with-slot-2. Bare "2" counts as yes-with-slot-2. "STOP" is treated as lost (no more messages — correct behavior for an opt-out).
Phone matching strips formatting. +1 (555) 123-4567 and 15551234567 match the same rescue.
Zero slots doesn't hallucinate. If OpenSlots is empty, the AI is instructed never to invent times — it just asks the customer to reply YES and the owner finds a time.
Event-sourced log. RescueLog is append-only; the latest row per phone is always the current state. Simple to audit, impossible to corrupt with an edit.

What I learned

The timezone bug was the nasty one. My first version compared appointment times against the server clock — fine on a self-hosted box in the same city, completely broken on n8n Cloud (UTC) for a US/Canada business. A 2pm appointment would look 7 hours late. Always compute "now" in the business's timezone.

The second lesson: no-show detection is only half the product. The other half is restraint — one nudge, clear stop conditions, and every ambiguous reply routed to a human instead of guessed at. That's what makes it something you'd actually let text your customers.

Get it

The complete workflow (JSON + setup guide) is here: https://automateplaybook.gumroad.com/l/hhisyc

And if you want to poke at n8n first, my free starter pack is here: http://automateplaybook.gumroad.com/l/hhnxv

I'm documenting the build in public — follow for more n8n automation breakdowns. Questions about the workflow logic? Drop them in the comments.

✅ n8n Verified Creator — 5 workflows approved by n8n's official template review: workflow 1 · workflow 2 · workflow 3 · workflow 4 · workflow 5

P.S. I joined the Systeme.io affiliate program (60% lifetime commissions) — if you need email marketing + sales funnels alongside these automations, that's my link.

Earn 30% Promoting n8n Automation Products — Affiliate Partners Wanted

AI Automation Playbook — Wed, 14 Oct 2026 03:59:46 +0000

I build and sell n8n workflow packs for small businesses on Gumroad — and I'm opening up a 30% affiliate program for all six paid products.

Quick credibility: I'm an n8n Verified Creator — 3 workflows approved by n8n's official template review:

The offer

30% commission on every sale made through your affiliate link
Gumroad handles all tracking and payouts automatically (30-day cookie window)
You get paid only when sales happen — no fees, no minimums, no inventory

The products (per-sale commission at 30%):

n8n Workflow JSON Pack: 10 Client-Ready AI Automations — $39 → you earn $11.70
AI Automation Playbook: Complete Bundle — $79 → you earn $23.70
AI Automation Freelancer Client Acquisition Kit — $29 → you earn $8.70
AI Automation Playbook: 10 Client-Ready n8n Workflows — $29 → you earn $8.70
30 AI Prompts to Get Clients for Your Automation Business — $19 → you earn $5.70
5 Copy-Paste n8n Workflows for Local Businesses — $12 → you earn $3.60

Every workflow is QA-tested (imports cleanly into a real n8n instance) — I don't ship untested files.

Who this fits

You write, post, or make videos about automation, AI tools, freelancing, or small business operations — and your audience buys tools that save them time. That's it.

How to join

Drop a comment below (or DM me) with the email on your Gumroad account, and I'll send you your affiliate links. Takes a day at most.

Honest note: these products are new and we're early — which is exactly why the 30% is on the table. If your audience is into automation, let's talk.

🆕 New: No-Show Rescue — AI auto-rebooking for missed appointments — my latest n8n workflow ($29).

✅ n8n Verified Creator — 5 workflows approved by n8n's official template review: workflow 1 · workflow 2 · workflow 3 · workflow 4 · workflow 5

P.S. I joined the Systeme.io affiliate program (60% lifetime commissions) — if you need email marketing + sales funnels alongside these automations, that's my link.

Stop Typing Invoices by Hand: An AI + n8n Workflow That Files Them for You

AI Automation Playbook — Sat, 10 Oct 2026 02:01:43 +0000

Every supplier invoice that lands on a small business owner's desk costs two things: the minutes it takes to type it into the books, and the mistakes that slip in while doing it. Multiply that by a few dozen invoices a month and you get a chore nobody bills for and everybody hates.

This is one of the easiest automations to build — for yourself or as a service — because the pain is visible and the workflow is simple: a photo or PDF goes in, clean line items come out in your accounting system, and anything suspicious gets flagged for a human.

What the workflow does

Two lanes.

1. Invoice intake. A webhook receives the uploaded invoice file. The workflow validates the file type, extracts the text with PDF OCR, and hands it to an AI information extractor that pulls out vendor, date, line items, and totals as structured fields. Then two guardrails run before anything touches your books:

Duplicate check — the workflow looks the invoice up in a Google Sheets log. Already seen it? It goes to review instead of posting twice.
Totals validation — the extracted line items are summed and compared against the extracted total. Mismatch? Review queue.

Clean invoices post to your accounting software over HTTP (the QuickBooks/Xero endpoint you configure). Anything suspicious — duplicate, totals mismatch, unreadable file — lands in a human review queue and triggers an email alert to the owner. Every invoice is logged to Sheets either way, so there is a full audit trail.

2. Review-queue reminder. Every morning at 8 AM, the workflow emails the owner a list of invoices still sitting in review, so nothing gets forgotten.

What you need to run it

An OpenAI API key — the extractor runs on your key, metered per use
A Google account (Sheets for the log and duplicate check, Gmail for alerts)
A QuickBooks or Xero account — or any accounting API reachable over HTTP
n8n, self-hosted or Cloud

All credentials are placeholders you fill in yourself; nothing ships with real keys. Two CONFIG nodes hold the owner email and the log sheet ID, and the whole thing is one import.

Why this one is easy to sell

Bookkeeping is a cost center every small business already pays for, in staff hours or an accountant's bill. An automation that visibly removes a weekly chore is a short conversation: upload an invoice, watch the line items appear. And the review queue answers the first objection — "what if the AI gets it wrong?" — before anyone has to ask it.

Get the workflow free

The import-ready version of this workflow is in the free starter pack: https://automateplaybook.gumroad.com/l/hhnxv — three workflows plus a quickstart guide, no cost. Stated plainly, as it is in the pack itself: the OpenAI and Twilio steps in these workflows run on your own paid accounts. The workflows are free; the API calls are yours.

Want more automation playbooks like this? The free newsletter breaks down one workflow a week: https://aiautomationplaybook.substack.com

Disclosure: This article was written with AI assistance and reviewed by the author. The workflow described is a real, tested n8n template — the invoice-OCR workflow (workflow-04) from the AI Automation Playbook pack. Service pricing changes often; verify current plans before you commit.

✅ n8n Verified Creator — 5 workflows approved by n8n's official template review: workflow 1 · workflow 2 · workflow 3 · workflow 4 · workflow 5

🆕 New: No-Show Rescue — AI auto-rebooking for missed appointments — my latest n8n workflow ($29).

P.S. I joined the Systeme.io affiliate program (60% lifetime commissions) — if you need email marketing + sales funnels alongside these automations, that's my link.

n8n vs Zapier vs Make: Which One Should You Actually Use in 2026?

AI Automation Playbook — Thu, 08 Oct 2026 17:27:54 +0000

If you're building automations for clients, the tool you pick matters. It affects your costs, what you can build, and how much you can charge. Here's an honest breakdown of the three big options.

The Short Version

n8n: Best for freelancers and agencies. Self-hostable, fair pricing, no per-task gouging.
Zapier: Best for non-technical teams who want the biggest app library. Expensive at scale.
Make (formerly Integromat): Best visual builder. Good middle ground on price.

If you're selling automation services, n8n is usually the right call. Here's why.

Pricing: Where It Really Hurts

Plan prices below are as of September 2026 — these companies change plans often, so check their current pricing pages before you commit.

Zapier charges per task. A "task" is every single action — not every workflow run. A 5-step workflow that runs 1,000 times a month is 5,000 tasks. Zapier's Professional plan starts at $19.99/month, with task tiers (100 / 2K / 10K / 100K / 1M / 2M) scaling with volume. Do the math: real client workloads blow past the entry tiers fast, and clients end up on far more expensive plans.

Make charges per operation, similar concept. Their pricing is more generous — the Core plan ($9/month) includes 10,000 operations. Better, but complex workflows still add up.

n8n charges per workflow execution on cloud (€20/month Starter, billed annually, for 2,500 executions), or you can self-host for free and pay only for infrastructure (roughly $5–$10/month on a small VPS). An execution is the whole workflow run, not each step. A 5-step workflow running 1,000 times is 1,000 executions, not 5,000.

For client work, this difference is massive. Your margin on a $200/month retainer disappears if the tool eats a big chunk of it.

What You Can Actually Build

Zapier has 9,000+ app integrations — the biggest library by far. But complex logic (loops, branching, custom code) feels bolted on. It's designed for simple "when this happens, do that" flows.

Make has an excellent visual builder and 2,000+ app integrations. Branching, loops, and error handling are first-class. The learning curve is steeper than Zapier but the ceiling is higher.

n8n is the most powerful of the three for technical users. Full JavaScript/Python code nodes, self-hosted AI models, webhooks, and no artificial limits on logic complexity. If you can think it, you can build it. The trade-off: the UI is less polished, and you'll need to understand APIs.

AI Capabilities

This is where n8n pulls ahead in 2026:

n8n has native AI agent nodes, LangChain integration, and can run against any LLM API (or self-hosted models). Building an AI chatbot or document processor is straightforward.
Zapier has AI add-ons, but they're limited and pricey.
Make has decent AI modules but nothing as deep as n8n's agent framework.

If your offer includes "AI automation" — and client demand is clearly moving that way — n8n is the most natural fit of the three.

The Freelancer Math

Say you land 5 clients, each with 3 active workflows:

Monthly tool cost: Zapier $300–$600 · Make $50–$100 · n8n Cloud $20–$50 · n8n self-hosted $5–$10
Cost per client: Zapier $60–$120 · Make $10–$20 · n8n Cloud $4–$10 · n8n self-hosted $1–$2

On a $200/month retainer per client ($1,000 total), Zapier eats 30–60% of your revenue. n8n self-hosted costs a few dollars a month — the tool barely touches your margin.

My Recommendation

Selling automation services? Use n8n. Self-host when you're comfortable, use their cloud when you're starting.
Building for a non-technical team that manages it themselves? Zapier or Make, because the client can understand the UI.
Just need a quick personal automation? Whichever you're fastest in.

The AI Automation Playbook includes 10 import-ready n8n workflows with setup guides — a shortcut if you're building an automation service business on n8n.

Want more automation playbooks like this? The free newsletter breaks down one workflow a week: https://aiautomationplaybook.substack.com

Disclosure: This article was written with AI assistance and reviewed by the author. Pricing figures were verified against the official pricing pages in September 2026 — these companies change plans often, so check the current pages before you commit.

✅ n8n Verified Creator — 5 workflows approved by n8n's official template review: workflow 1 · workflow 2 · workflow 3 · workflow 4 · workflow 5

🆕 New: No-Show Rescue — AI auto-rebooking for missed appointments — my latest n8n workflow ($29).

P.S. I joined the Systeme.io affiliate program (60% lifetime commissions) — if you need email marketing + sales funnels alongside these automations, that's my link.

5 n8n Automations Every Local Business Needs (And What to Charge for Them)

AI Automation Playbook — Tue, 06 Oct 2026 16:41:35 +0000

If you're selling automation services to local businesses, you need offers that solve obvious, painful problems. Not "AI transformation" — a plumber doesn't care about that. They care about missed calls, no-shows, and drowning in paperwork.

Here are five automations that sell well to local businesses, what they do, and rough pricing to anchor your quotes.

1. Missed-Call Text-Back

The problem: The business misses calls during jobs. Callers don't leave voicemail — they call the next business on Google.

The automation: When a call goes unanswered, n8n fires a webhook (via Twilio), sends an SMS almost immediately, and an AI agent qualifies the caller and books the appointment.

Typical pricing: $800–$1,500 setup + $100–$200/month for monitoring and SMS costs.

Why it sells: The pitch writes itself: a single recovered job can pay for the system.

2. Google Review Auto-Responder

The problem: The business gets reviews but never responds. Unanswered negative reviews kill trust.

The automation: n8n watches for new Google reviews via API. Positive reviews get a warm thank-you. Negative reviews trigger an instant Slack/email alert to the owner with a suggested response draft.

Typical pricing: $500–$900 setup + $50–$100/month.

Why it sells: Every business owner knows reviews matter. This is a visible, easy-to-understand win.

3. Lead Follow-Up Sequence

The problem: Quotes go out, then nothing. The business is too busy to chase.

The automation: When a quote is sent (tracked in a Google Sheet or CRM), n8n starts a timed follow-up sequence — email on day 2, SMS on day 5, final check-in on day 10. Replies route to the owner.

Typical pricing: $700–$1,200 setup + $75–$150/month.

Why it sells: "How many quotes did you send last month that went nowhere?" That question closes deals.

4. Daily Sales Summary

The problem: The owner has no idea what happened today without logging into five systems.

The automation: Every evening, n8n pulls data from Stripe, the booking system, and the POS, then sends one clean email: revenue, bookings, top services, flagged issues.

Typical pricing: $600–$1,000 setup + $50–$100/month.

Why it sells: Business owners love a single number in their inbox. It's the kind of thing they didn't know they needed.

5. Abandoned Booking Recovery

The problem: Customers start booking online but don't finish. The business never knows.

The automation: When a booking is abandoned (tracked via webhook from the booking tool), n8n waits 30 minutes, then sends a friendly SMS: "Still need that appointment? Here's your link." A second nudge goes out the next morning.

Typical pricing: $700–$1,200 setup + $75–$150/month.

Why it sells: Same logic as cart abandonment in e-commerce. Every recovered booking is revenue the business would otherwise have lost.

How to Package These

Don't sell "automation." Sell outcomes:

"Never miss another after-hours lead" (missed-call text-back)
"Every review answered within an hour" (review responder)
"Your quotes follow up on themselves" (lead follow-up)

Price on value, not hours. A system that recovers meaningful revenue in lost jobs is worth a four-figure setup without blinking.

Want the actual workflow files? The Free n8n Workflow Starter Pack includes the missed-call text-back workflow, ready to import. The full set of ten is in the Workflow JSON Pack.

Want more automation playbooks like this? The free newsletter breaks down one workflow a week: https://aiautomationplaybook.substack.com

Disclosure: This article was written with AI assistance and reviewed by the author. Pricing ranges are rough anchors for quotes, not market data; platform details reflect September 2026 and may change — verify on the official sites before you build or quote.

✅ n8n Verified Creator — 5 workflows approved by n8n's official template review: workflow 1 · workflow 2 · workflow 3 · workflow 4 · workflow 5

🆕 New: No-Show Rescue — AI auto-rebooking for missed appointments — my latest n8n workflow ($29).

P.S. I joined the Systeme.io affiliate program (60% lifetime commissions) — if you need email marketing + sales funnels alongside these automations, that's my link.

10 n8n Workflows You Can Sell to Small Businesses Right Now

AI Automation Playbook — Sun, 04 Oct 2026 18:26:17 +0000

Small businesses know they should automate. Most haven't, because they don't know where to start and don't have anyone in-house to build it. That's the gap you sell into.

Below are ten n8n workflows that solve real, specific problems for small businesses. For each one: what it does, who pays for it, and what to charge. These aren't theoretical — they're the types of automations that agencies and freelancers sell and deploy for clients today.

Pricing is for the US/Canadian market, and the ranges below are rough anchors for your quotes — not market data. Adjust for your region, but don't undercharge: the value is in the outcome (hours saved, leads captured), not the hours you spent building.

1. Missed-Call Text-Back with AI Booking

What it does: When a call goes unanswered, the caller gets an SMS within a minute. An AI assistant qualifies them, answers questions, and books an appointment directly into the business's calendar. The owner gets a one-line summary.

Who buys it: Any business where the phone is the main way customers reach them — plumbers, HVAC, electricians, dental clinics, auto repair shops, law firms doing intake.

Typical pricing: $1,500–$3,000 setup + $200–$400/month for monitoring and AI costs. The monthly retainer is the real business here — the workflow runs itself, and you're paid to keep it running.

2. AI Lead-Qualification Chatbot

What it does: A chat widget on the business's website that greets visitors, asks qualifying questions (budget, timeline, project type), scores the lead, and routes hot leads to a human immediately while nurturing the rest by email.

Who buys it: Businesses with website traffic but no one staffing live chat — contractors, real estate agents, SaaS companies, marketing agencies, private clinics.

Typical pricing: $1,000–$2,500 setup + $150–$300/month. The pitch is simple: "You're paying for traffic. This makes sure none of it goes to waste after 5 PM."

3. Automated Review Request and Response

What it does: After a job is marked complete (in the CRM, booking system, or a spreadsheet), the customer automatically gets a text asking for a Google review. Positive responses get a thank-you; negative ones get routed to the owner before they post publicly. AI drafts suggested replies to every review.

Who buys it: Any local business that lives on Google reviews — restaurants, dentists, contractors, auto shops, salons. If they have fewer than 50 reviews, they need this.

Typical pricing: $800–$1,500 setup + $100–$200/month. Easy to sell because the results are easy to see: review counts climb as the requests go out.

4. Invoice Data Extraction to Accounting

What it does: Watches an email inbox or upload folder for incoming invoices (PDFs, photos). AI extracts the vendor, line items, totals, and tax, then pushes structured data into QuickBooks, Xero, or a spreadsheet. Flags anything that doesn't match a purchase order.

Who buys it: Small businesses drowning in paperwork — construction companies, distributors, property managers, any business with 50+ invoices a month and a bookkeeper doing it by hand.

Typical pricing: $1,200–$2,500 setup + $150–$300/month. Sell it on hours saved: if the bookkeeper spends 10 hours a month on data entry, the time savings alone can justify the cost.

5. Abandoned Quote Follow-Up Sequences

What it does: When a quote or estimate is sent but not accepted within a set timeframe, the prospect automatically enters a follow-up sequence — a check-in text at day 3, a value-add email at day 7, a limited-time incentive at day 14. Stops automatically if they reply or accept.

Who buys it: Businesses that send quotes but are bad at following up — contractors, landscapers, wedding vendors, B2B service providers, anyone whose sales process has a "we sent the proposal and never heard back" problem.

Typical pricing: $1,000–$2,000 setup + $100–$200/month. A single recovered quote can cover the entire cost.

6. CRM Lead Enrichment and Routing

What it does: When a new lead enters the CRM (from a web form, spreadsheet import, or manual entry), the workflow enriches it — company size, industry, social profiles, technographics — scores it against the business's ideal customer profile, and routes it to the right salesperson with a Slack notification and context summary.

Who buys it: B2B companies with a sales team of 2+, marketing agencies, business brokers, commercial real estate, any business where speed-to-lead and correct routing matter.

Typical pricing: $1,500–$3,000 setup + $200–$400/month. The enrichment APIs cost money, so bake that into the retainer.

7. AI Content Pipeline (Blog + Social)

What it does: Takes a topic or rough outline, generates a draft blog post, creates 3–5 social media variations, formats everything for the business's CMS and social scheduler, and drops it in a review queue. A human approves before anything publishes.

Who buys it: Businesses that know they should publish content but never do — local service companies, consultants, coaches, small e-commerce brands. The owner is the bottleneck; this removes them from the drafting process.

Typical pricing: $1,000–$2,000 setup + $200–$400/month. Position it against the cost of a freelance writer — and point out the pipeline drafts around the clock and never misses a deadline.

8. Meeting Notes Summarizer to CRM

What it does: After a Zoom/Teams/Google Meet call, the transcript is automatically pulled, AI summarizes it into key points, action items, and next steps, then pushes the summary into the CRM under the correct contact and deal. The salesperson gets a Slack message with the summary.

Who buys it: Sales teams, consultants, agencies, real estate teams — anyone doing 10+ client calls a week and updating the CRM from memory (badly, or not at all).

Typical pricing: $800–$1,500 setup + $100–$200/month. The sell: "Your CRM is only as good as the data in it. This makes the data automatic."

9. Multilingual Customer FAQ Bot

What it does: An AI chatbot trained on the business's FAQs, policies, and service info that responds in the customer's language — Spanish, Mandarin, French, whatever the local market needs. Handles the 80% of repetitive questions; escalates the rest to a human with full context.

Who buys it: Businesses in multilingual markets — clinics, property managers, restaurants, retail, any business where staff spend hours answering the same questions in multiple languages.

Typical pricing: $1,200–$2,500 setup + $150–$300/month. The alternative — staffing for every language your customers speak — costs far more.

10. Applicant Screening Pipeline

What it does: When job applications come in (email, form, or job board), AI screens resumes against the job requirements, scores candidates, sends knock-out questions to borderline applicants, and presents the hiring manager with a ranked shortlist. Rejected applicants get a polite automated response.

Who buys it: Businesses that hire frequently but have no HR department — restaurants, retail chains, home-services companies, warehouses, franchises.

Typical pricing: $1,000–$2,000 setup + $100–$200/month. Hiring managers will pay to never read another 200 unqualified resumes.

How to Actually Sell These

Building the workflow is the easy part. Here's what matters for selling:

Sell the outcome, not the tech. Nobody wants to buy "an n8n workflow with 18 nodes." They want "never miss another $500 job because you were on a ladder." Lead every conversation with the problem and the dollar value of solving it.

Start with one workflow, not ten. Pick the one that matches the business's most obvious pain point. Land the first project, deliver it well, then upsell the rest. A client who trusts you with one automation will buy three more.

Price on value, charge monthly. The setup fee covers your build time. The monthly retainer — $100 to $400 depending on the workflow — is where the real money is. You're not selling software; you're selling "this keeps working and someone watches it." Frame the retainer as insurance, not a subscription.

Use the free workflows as your foot in the door. Give away something useful (like the missed-call text-back starter) to start the conversation. Once they see it work, selling the full build is straightforward.

The Full Playbook

If you want all ten of these as importable n8n workflows — with exact build steps, node configurations, pricing scripts, and the outreach templates to land clients — that's what the AI Automation Playbook covers:

Get the AI Automation Playbook

Ten import-ready workflows, the pricing calculator, proposal templates, and everything you need to start selling automation services this month. Available in English and Chinese.

Want more automation playbooks like this? The free newsletter breaks down one workflow a week: https://aiautomationplaybook.substack.com

✅ n8n Verified Creator — 5 workflows approved by n8n's official template review: workflow 1 · workflow 2 · workflow 3 · workflow 4 · workflow 5

🆕 New: No-Show Rescue — AI auto-rebooking for missed appointments — my latest n8n workflow ($29).

P.S. I joined the Systeme.io affiliate program (60% lifetime commissions) — if you need email marketing + sales funnels alongside these automations, that's my link.

How to Build a Missed-Call Text-Back System with n8n (Free Workflow Inside)

AI Automation Playbook — Fri, 02 Oct 2026 18:45:14 +0000

A missed call after hours isn't just a missed call. For a plumber, a dental clinic, or an HVAC contractor, it's a job that just went to whoever picked up the phone first.

Most small businesses handle this with voicemail. Voicemail doesn't book jobs. An automated text-back does.

This guide walks through building a missed-call text-back system in n8n: when a call goes unanswered, the caller gets an SMS right away, an AI assistant qualifies them, and a booking gets scheduled — all without anyone at the business touching a phone.

The full importable workflow is free. I'll link it at the end.

Why Missed Calls Cost More Than You Think

Think about the math for a typical home-services business. They get 20 calls a day and miss 5 because the techs are on roofs or under sinks. If each job is worth $300 and half of those missed callers go elsewhere, that's $750 a day walking out the door.

The standard response is "call them back later." The problem: later usually means an hour or three. By then the caller has already booked with someone else. Businesses that respond within minutes tend to convert far more leads than those that take an hour — speed is the whole game here.

A missed-call text-back fixes the speed problem. The caller gets a message almost immediately: "Sorry we missed you — how can we help?" That small acknowledgment keeps them from calling the next business on Google Maps.

What the System Does

Here's the flow, start to finish:

Call comes in and goes unanswered. Your phone provider (Twilio in this guide) fires a webhook to n8n.
n8n checks if it's a new number. Returning customers get a different message than first-time callers.
An SMS goes out immediately. Short, human-sounding, asking what they need.
The caller's reply comes back through the same webhook. An AI agent (OpenAI) reads it, figures out intent — new quote, existing job, emergency, wrong number — and responds accordingly.
If it's a booking request, the AI collects details. Service needed, preferred time, address.
The appointment gets created in Google Calendar (or Cal.com, or whatever the business uses) and the caller gets a confirmation text.
The business owner gets a summary. One text or Slack message: new booking, Thursday 9 AM, burst pipe, address.

Total time from missed call to first text: as fast as the Twilio-to-n8n round trip allows — no human in the loop. Total human involvement: zero, until the tech shows up at the door.

What You Need

n8n — self-hosted or n8n Cloud. Either works.
Twilio — a phone number with SMS and voice. You need the Account SID, Auth Token, and the phone number. Twilio's pay-as-you-go is fine for starting.
OpenAI API key — for the AI agent that reads and responds to messages.
Google Calendar (or Cal.com) — for booking. The workflow uses Google Calendar by default; swapping in Cal.com is a single-node change.

That's it. No custom app, no server to maintain beyond n8n itself.

Building the Workflow: Node by Node

1. Webhook Trigger

The workflow starts with an n8n Webhook node set to receive POST requests. In Twilio, you configure two things:

Voice URL: points to a TwiML bin or a small endpoint that plays a message and hangs up (or routes to voicemail). The key part is that Twilio also fires on the "call completed" event.
Status callback / missed call webhook: Twilio sends call status events to your n8n webhook URL. You filter for calls where CallStatus is no-answer or busy.

The webhook payload includes the caller's number (From), your Twilio number (To), and the call SID. That's all the workflow needs to kick off.

2. Check for Existing Contact

Before sending anything, the workflow checks whether this number has called before. The simplest approach: a Google Sheet or Airtable with two columns — phone number and name. An n8n node queries it.

Known number: "Hey Mike, sorry we missed your call — what's going on?"
Unknown number: "Hi, this is Acme Plumbing — sorry we missed you! What can we help with today?"

This one check makes the difference between a message that feels automated and one that feels personal.

3. Send the Initial SMS

A Twilio node sends the first text. Keep it under 160 characters. Keep it short and human. (One caution: SMS rules such as US A2P 10DLC generally require prior consent and opt-out language — I'm not a lawyer, so check your local regulations and your provider's requirements before you send.)

Good: "Sorry we missed your call — what can we help you with today?"

Bad: "Thank you for contacting Acme Plumbing LLC. An agent will respond shortly. Reply STOP to opt out."

The first one sounds like a person. The second sounds like a phone tree.

4. The AI Responder

This is the core of the system. When the caller replies, the message comes back into the same webhook. An n8n AI Agent node (or an OpenAI node with a well-crafted system prompt) processes it.

The system prompt needs to cover who the agent is, what it can do (answer common questions, collect booking details, escalate emergencies), what it can't do (quote exact prices for unseen work), and the tone: short texts, one question at a time.

Here's a skeleton of the logic the agent follows:

IF message contains emergency keywords (flood, burst, gas, leak)
  → mark URGENT, notify owner immediately, tell caller help is on the way
ELSE IF message asks about booking
  → collect: service type, preferred day/time, address
  → check calendar availability
  → confirm booking
ELSE IF message is a question (hours, pricing, service area)
  → answer from knowledge base
ELSE
  → ask a clarifying question, one at a time

5. Calendar Booking

Once the AI has the details, an n8n Google Calendar node creates the event. The event title includes the service type and caller name: "Burst pipe repair — Sarah M."

The node also needs a buffer. Don't let the AI book back-to-back with zero travel time. Set a 30-minute buffer before and after in the calendar settings, or have the workflow check for conflicts within a 1-hour window.

After the event is created, a Twilio node sends the confirmation:

"You're booked for Thursday 9:00 AM. We'll text when the tech is on the way. Address on file: 42 Oak Street — is that right?"

That last question matters. Wrong addresses are a common cause of wasted truck rolls.

6. Owner Notification

Every booking triggers a summary to the business owner — SMS or Slack, their choice:

"New booking: Sarah M., burst pipe repair, Thu 9:00 AM, 42 Oak St. AI handled 6 messages. No action needed."

The "no action needed" line is deliberate. Owners ignore notifications they have to decode. Make it skimmable.

Handling Edge Cases

A few things that will come up in production:

Spam/robocalls: If the initial SMS gets no reply, don't follow up more than once. One nudge after 10 minutes ("Still there? Happy to help when you're ready."), then stop.
After-hours vs. business hours: Use an n8n IF node with the current time. During business hours, the text can say "We're with a customer — we'll call you right back." After hours, it routes to the full AI booking flow.
Multiple messages in quick succession: Callers sometimes send three texts in a row. Add a 30-second wait node before the AI processes, so it reads all three as one conversation instead of replying three times.
Wrong numbers: If the caller says "wrong number," log it and stop. Don't keep texting.

Testing Before You Go Live

Use your own phone. Call the Twilio number, hang up, and walk through the full flow as a customer.
Test the emergency path. Text "my basement is flooding" and confirm the owner gets pinged promptly.
Test the calendar conflict. Book a time that's already taken and make sure the AI offers alternatives instead of double-booking.
Test with a real business owner watching. They'll think of scenarios you won't — "what if someone asks about a warranty?" or "what if they want to reschedule?"

Run it in parallel with the old system (voicemail) for a week. Compare: how many missed calls turned into bookings under each system? That comparison is what sells this to the next client.

Get the Free Workflow

I've packaged this exact system — the full n8n JSON, the Twilio configuration steps, and a setup guide — as a free download:

Download the Free n8n Starter Pack

It includes this missed-call text-back workflow plus an invoice-extraction workflow, both ready to import into n8n. Import the JSON, add your Twilio and OpenAI credentials, point the webhook at your Twilio number, and you're live.

Prefer the official route? This exact workflow is now published as a free template in n8n's template library — one-click import, no download needed:

Reply to missed calls and book appointments with Twilio, Google, and GPT-4o

If you want the full playbook — 10 import-ready workflows with build steps, pricing guidance, and the scripts to sell them — that's in the AI Automation Playbook.

Disclosure: This article was written with AI assistance and reviewed by the author. The workflow steps and links were checked; pricing and platform details reflect September 2026 and may change — verify on the official sites before you build.

✅ n8n Verified Creator — 5 workflows approved by n8n's official template review: workflow 1 · workflow 2 · workflow 3 · workflow 4 · workflow 5

🆕 New: No-Show Rescue — AI auto-rebooking for missed appointments — my latest n8n workflow ($29).

P.S. I joined the Systeme.io affiliate program (60% lifetime commissions) — if you need email marketing + sales funnels alongside these automations, that's my link.