AboutWorkServicesToolsBlogContact Get in touch โ†’
Workflow

My AI Agent Portfolio: Every Automation I've Shipped, Ranked

Every automation, routine, and pipeline I actually run in production across my sites, ranked by hours saved, traffic impact, and what breaks if it stops. One place, not a dozen scattered posts.

I've written about a dozen different automations across this site over the past few months, a checklist here, a routine prompt there, a tracking setup somewhere else. Nobody reading one post sees the whole system. This is the whole system: every agent, routine, and pipeline I actually run in production, ranked by what it's actually worth, not by how interesting it was to build.

I'm ranking by three things: hours it saves me per week, measurable traffic or revenue impact, and how much would break if it stopped running tomorrow. Interesting-but-low-impact stuff sits at the bottom on purpose.

The ranked list

#AutomationImpactWhat it does
1AI Price RadarHighest, revenue + productA full product, not a script, built and run with AI as the engineering partner
2Pre-merge SEO gateHigh, quality controlClaude routine reviews every blog PR against my checklist before merge
3AI traffic trackingHigh, visibilityGA4 + Search Console setup that separates AI referral traffic from classic organic
4AI Overview citation systemHigh, compoundingStructural rules, answer-first, schema, tables, that earn citations across every post
5n8n content pipelineMedium-high, timeFans a topic into draft, schema, and image generation without me touching each step
6IndexNow pushMedium, speedInstant Bing crawl notification on every publish, now fired by a deploy-triggered routine
7Pre-launch checklistMedium, consistencyThe standard every page gets held to, now the literal prompt the SEO gate routine runs
8Weekly content auditMedium, risk reductionCatches broken links, stale dates, and orphan pages before a reader or a bot does
9Multi-country content systemMedium, scaleLocalization logic that keeps country pages distinct instead of templated duplicates
10Agent framework evaluationLower, exploratoryOngoing testing of LangGraph, CrewAI, and AutoGen for client-facing automation work
๐Ÿ“Š
Why AI Price Radar tops the list

Everything else on this list supports content or search visibility. AI Price Radar is the one where AI assistance went all the way to a shipped product with its own users, its own data layer, and its own revenue path. Scripts save time. A product changes what's possible.

What's actually running right now

Three pieces moved from "something I do manually" to "something that runs without me" in the last month alone, I wrote the exact setup in I automated my SEO checklist with Claude routines:

  • A GitHub-triggered routine that reviews every blog PR against my pre-launch checklist and leaves inline comments instead of silently merging.
  • A scheduled weekly routine that audits all three sites for broken links, stale freshness dates, and orphan pages, and opens a GitHub issue I triage Monday morning.
  • An API-triggered routine fired from my deploy script that reads the commit diff and submits only the changed URLs to IndexNow, no more manually copying URLs into a curl command.

I broke down when each of these makes sense versus a plain n8n workflow in Claude Routines vs n8n: when to use which, the short version: anything needing judgment goes to a routine, anything with fixed steps stays in n8n.

The systems underneath the content

A few pieces aren't single automations, they're standards that every other automation on this list enforces. Worth naming separately because they're the reason the rest of the list actually compounds instead of degrading.

  • The pre-launch SEO/AIO/GEO/AEO/SXO/LLMO checklist. Originally a manual document, now literally the prompt text the pre-merge routine runs against every PR.
  • The AI Overview citation rules. Answer-first structure, tables over prose, schema on everything, these aren't automated, they're the judgment the routines are checking for.
  • The multi-country content framework. Keeps 80+ localized pages distinct enough to avoid duplicate-content flags, enforced manually during writing, checked automatically during the weekly audit.
๐Ÿงฐ
The actual stack, end to end

Astro + Cloudflare Pages for the sites themselves. n8n for deterministic pipeline work. Claude Code and Claude Routines for anything needing judgment, PR review, audits, deploy-triggered pushes. GA4 and Search Console for measurement. IndexNow and Rich Results Test for crawl and schema validation. Supabase where a data layer is actually needed, like AI Price Radar. Nothing exotic, the leverage is in how the pieces are wired together, not in any single tool.

What I'd rebuild first if I lost it all

  1. The AI traffic tracking setup. Without it I'm optimizing blind, this is the measurement layer everything else depends on.
  2. The pre-merge SEO gate. Quality control compounds, losing it means slow content-quality drift I wouldn't notice for months.
  3. AI Price Radar's data pipeline. It's a product with users, not a nice-to-have.
  4. Everything else. Genuinely rebuildable in a weekend each, now that the patterns exist and are written down.

Key takeaways

  • Ten automations are currently live across my sites, ranked here by impact, not by how recently I built them.
  • The highest-impact item isn't a script, it's AI Price Radar, a shipped product, proof that AI-assisted building scales past content tasks.
  • Three Claude routines now run unattended: PR review, weekly audits, and deploy-triggered IndexNow submission.
  • The underlying checklists and citation rules matter more than any single automation, they're what the automations are actually enforcing.
  • The stack is ordinary, Astro, n8n, Claude Code, GA4, IndexNow, the leverage comes from how it's connected, not from exotic tooling.

What I'll do next

Adding the agentic-commerce readiness checklist from the Black Friday prep post as an eleventh item once it's running as a standing audit instead of a one-time pass. I'll update this ranking whenever something new earns a spot or something old stops pulling its weight.

FAQ

What's the highest-impact AI automation you run? +
AI Price Radar, a full product built and operated with AI as the engineering partner, not a script or content pipeline. It has its own users and revenue path, which puts it ahead of anything that only saves time on content work.
Do you use Claude Routines or n8n for most of your automation?
Both, split by task type. n8n handles deterministic, high-volume pipeline work like content fan-out. Claude Routines handle anything needing judgment, PR review, audits, deploy-triggered actions, where the right steps depend on what's actually found during the run.
How do you measure whether an automation is actually worth keeping?
Three criteria: hours saved per week, measurable traffic or revenue impact, and how much breaks if it stops running. Anything that scores low on all three gets cut rather than kept out of sunk-cost attachment.
What's the simplest piece of this stack to copy?
The deploy-triggered IndexNow push. It's a single API call wired to fire after deploy, covered with the exact script in IndexNow vs Google Search Console. No agent judgment required, just a webhook and a working IndexNow key.
Share this

Want this built for your team?

I design AI agents and growth automation that run without babysitting. If that sounds useful, let's talk.

Get in touch โ†’

Keep reading