The Minimal Automation Stack: What's Worth Keeping

You built the automations. The content pipeline, the email sequences, the payment notifications, the social distribution, the reporting dashboard, the monitoring stack, the CRM follow-ups, the backup workflows. You built all of it. And now you spend four hours a month fixing the things that break, checking the things that might have broken, and wondering whether some of these automations are saving you time or just converting idle anxiety into active maintenance.

This is the audit. After building and running every automation described in this series across a production stack of 15+ sites for over a year, some of them earn their keep. Some of them save minutes but cost hours. Some of them should never have been built. The honest accounting follows — what to keep, what to kill, and the minimal stack that delivers maximum return for minimum ongoing attention.

What The Docs Say

The automation platforms don't talk about killing automations. Their documentation, reasonably, describes building them. n8n's docs show you how to create workflows, not how to decide which ones aren't worth maintaining. Zapier's pricing page incentivizes more Zaps, not fewer. Make's marketing describes connecting "all your tools" — the word "all" doing heavy lifting for their revenue model.

The productivity literature offers frameworks for this — ROI calculations, time-tracking audits, the Eisenhower matrix applied to workflows. The formula is simple: setup time plus monthly maintenance time, divided by time saved per month. If the payoff period exceeds six months, reconsider. If the monthly maintenance exceeds the monthly time savings, kill it.

The docs also describe "set it and forget it" automation — the idea that well-built workflows run indefinitely without attention. This is technically true for some automations and functionally false for most. Any workflow that touches an external API requires periodic credential maintenance. Any workflow that depends on a platform's data format is vulnerable to API changes. The question isn't whether an automation can run forever — it's how much it costs per month to keep it running.

What Actually Happens

After a year of production operation, the automations sort cleanly into three tiers — and the tier boundaries are sharper than I expected.

Tier 1 — Always worth it. These automations save real time, prevent real problems, and require minimal maintenance. They've earned permanent slots in the stack.

Content publishing pipeline — the workflow that takes a Ghost publish event and distributes it to social media, email, and tracking. This is the highest-ROI automation in the entire stack. Without it, publishing a single article requires logging into 4-5 platforms and manually posting, adapting, scheduling. With it, I publish to Ghost and everything else happens. Maintenance cost is roughly 30 minutes per month — mostly credential checks and the occasional API format update. Time saved is 15-20 minutes per article across dozens of articles per month. The math is not close.

Email welcome sequences — the automation that sends a multi-step welcome series to new subscribers. Once built, this runs for months without intervention. Kit handles it natively with excellent reliability. Ghost's built-in email can handle basic newsletter sends but still needs external tooling for true sequences. The maintenance cost approaches zero — you review the sequence copy quarterly and update if your offering has changed. The time saved is real but harder to quantify; the value isn't time savings but consistent subscriber experience. Every new subscriber gets the same onboarding regardless of when they sign up or how busy you are that week.

Payment and membership notifications — Stripe webhooks triggering Slack alerts and membership provisioning. Stripe's API is rock-solid. The webhook delivery is reliable. The automation that says "new subscriber" or "payment failed" in Slack takes 30 seconds to read and keeps you informed without dashboard-checking. Maintenance cost is near zero — Stripe API keys don't expire, and the webhook format has been stable for years.

Uptime monitoring — UptimeRobot watching your sites and alerting on downtime. Not technically an n8n automation, but part of the stack. Free tier covers most needs. Maintenance cost is zero after initial setup. Value is binary — either it catches a downtime event before your audience does, or it sits there quietly costing nothing. There is no scenario where this automation isn't worth having.

Data backups — scheduled Ghost content exports to cloud storage. A weekly n8n workflow that hits the Ghost Admin API, exports content as JSON, and uploads to S3 or Google Drive. Maintenance cost is about 15 minutes per month — checking that the exports are actually running and that the files aren't empty. Storage cost is pennies. The value is insurance — you never appreciate it until you need it, and when you need it, nothing else matters.

Tier 2 — Worth it at scale. These automations justify themselves above a certain volume threshold. Below that threshold, the manual version is faster when you account for maintenance.

Social media scheduling with AI draft generation — the workflow that generates platform-specific social posts from article content using Claude, queues them for review, and schedules approved posts. This saves 5-10 minutes per article in copywriting time. If you publish 3-4 articles a week across multiple sites, that's 60-160 minutes per month. The maintenance cost is about 45 minutes per month — social platform API changes are frequent, token expirations are unpredictable, and the AI prompt needs occasional tuning as your voice evolves. The breakeven point is roughly 8-10 articles per month. Below that, write the social posts yourself.

Reporting dashboards — the weekly data pull from Ghost, Stripe, GA4, and Search Console into a Google Sheet. Saves about 30-45 minutes of manual data collection per week. Maintenance cost is about 30-60 minutes per month — Google API credential refreshes are the main culprit. Worth it if you're running 5+ sites or reporting to someone who expects consistent formatted numbers. Not worth it for a single site where you can just open Ghost and Stripe once a week.

CRM follow-up reminders — the automation that nudges you when a contact or deal has gone stale. Useful once you're managing more than 20-30 active relationships. Below that, your memory and a simple to-do list work fine. The automation's value is preventing things from falling through cracks — but if you only have five cracks, you can see them all without a system.

Tier 3 — Usually not worth it. These automations sound valuable in the planning stage and underdeliver in production.

AI customer support for small operations — if you receive fewer than 50 support inquiries per month, answering them yourself is faster than building and maintaining the AI support layer. The knowledge base needs ongoing updates, the AI occasionally confabulates answers that create more support load than they resolve, and the escalation logic needs tuning. [VERIFY] The 50 inquiry/month threshold is based on personal experience; industry data may suggest a different breakeven point. For larger operations, AI support can work. For a solopreneur or small publisher, it's a solution looking for a problem.

Complex multi-step workflows that span more than 5-6 services — the "connect everything" architecture from article 16.10 is powerful, but every additional connection is a maintenance liability. When one API in a six-service chain changes, the whole pipeline needs debugging. The workflows that touch 2-3 services are robust. The ones that touch 6-7 are fragile. If you find yourself spending more time maintaining a workflow than the workflow saves, it has crossed the line from automation to hobby.

Anything requiring weekly manual intervention to keep running — if a workflow needs you to manually refresh a token, manually check an output, or manually restart after a failure more than twice a month, it hasn't actually automated the task. It has added an intermediary between you and the task. This sounds obvious, but in practice several of my early workflows fell into this trap — technically automated but practically requiring more attention than the manual version.

When To Use This

The automation audit should happen quarterly. Block two hours on your calendar, open your n8n dashboard, and review every active workflow. For each one, answer three questions: when did it last fail? How long did the fix take? How much time has it saved since the last audit? If a workflow failed more than twice and the total fix time exceeded the total time saved — flag it for simplification or removal.

The minimal stack recommendation for a typical publisher or solopreneur is this: n8n self-hosted on a $10-20/month VPS. Five to seven core workflows — content publishing pipeline, email welcome sequence, payment notifications, uptime monitoring, data backup, and optionally social scheduling and reporting if your volume justifies them. UptimeRobot free tier for site monitoring. A Slack channel for all automation notifications. Total monthly cost: $10-20 for the server, $0 for UptimeRobot, whatever your Ghost/email/Stripe subscriptions already cost. Total monthly maintenance: 2-3 hours.

That's it. Everything beyond this is optional, and "optional" means "add it only when the manual version of the task is costing you more than the automation will."

The "just do it manually" test is the most useful filter for automation decisions. If the manual version takes 5 minutes and happens twice a week, that's 40 minutes per month. If the automation takes 4 hours to build and 30 minutes per month to maintain, the payoff period is roughly five months — and that assumes zero debugging surprises. For anything that happens less than weekly or takes less than 5 minutes manually, the automation probably isn't worth building. Write it down, do it by hand, move on.

When To Skip This

Skip the audit if you have fewer than three automations running. The overhead of a formal review process exceeds the value when your stack is small. Just keep a mental note of which workflows are running and check them monthly.

Skip killing automations that protect against catastrophic but rare events — backups and monitoring — even if they technically don't "save time" in the monthly accounting. These aren't productivity automations. They're insurance automations. The ROI calculation doesn't apply because the value is asymmetric: the downside of not having them is disproportionate to their maintenance cost. Your backup workflow might run for two years without you needing it. The one time you need it, it saves your entire operation.

Skip the temptation to rebuild everything from scratch. When the audit reveals that a workflow is too complex or too fragile, the fix is usually simplification — removing nodes, reducing service connections, splitting a mega-workflow into smaller independent workflows — not starting over. The knowledge embedded in a working-but-messy workflow has value. Clean it up. Don't throw it away.

And — at the risk of ending a series about automation with a deeply analog observation — skip automating the things that benefit from your attention. Writing the article, replying to the reader's email, deciding what to build next. The automations handle distribution, notification, tracking, and protection. The work that matters is still yours to do. The goal of the minimal stack is to give you back the time to do it.

This is part of CustomClanker's Automation Recipes series — workflows that actually run.