What These Case Studies Are
Every workflow below is a real shipped piece of work – either built on Auburn AI’s own properties or delivered through this consulting practice. Client system names are real where permission was given, anonymized where it was not. No demos, no mockups, no theoretical builds.
Case Study 1: Bulk SEO Meta Rewrite – 119 Pages Corrected for $0.14
The Problem
A content site had accumulated 120 pages with broken meta descriptions – a mix of truncated titles stuffed into the description field, missing descriptions entirely, and a handful that blew past the 160-character limit. Fixing them by hand in WordPress would have taken the better part of a day.
The Build
An n8n workflow pulled every page’s existing meta data through the Rank Math REST API, sent each record to Claude with a prompt constraining the output to 140-155 characters and the page’s actual topic, then pushed the rewritten description back through the same API. Total nodes: eleven. Total build time: about two hours, including testing on a staging environment first.
The Outcome
119 of 120 pages were corrected in just over two minutes. One page had a malformed Rank Math record that needed a manual fix – caught by an error branch that logged failures to a Google Sheet rather than silently skipping them. Total Claude API spend: $0.14. One version of this engagement was later productized into a Chrome extension that runs the same rewrite loop on demand from the browser toolbar.
Case Study 2: Multi-Site n8n Recovery – Three Silent Failures Found and Fixed
The Problem
A 17-workflow n8n stack spread across two sites had three flows that had stopped producing output without throwing visible errors. The client assumed they were running. They were not.
The Diagnosis
The failure pattern turned out to be stale credential propagation – API keys had been rotated in the credentials manager but the affected workflows were still holding a cached reference to the old credential node ID. n8n did not surface this as an error at the workflow level; it just returned empty payloads downstream. A separate bug in the Market Signal workflow was a template-rendering issue where a variable reference was evaluating before the HTTP response was available in scope, causing a silent null pass-through.
The Outcome
All three workflows were diagnosed, patched, and confirmed producing correct output in a single session. Zero downtime – the fixes were applied to inactive executions and verified in test mode before re-enabling production schedules. The credential propagation pattern is now documented as a standard check in Auburn AI’s n8n audit checklist.
Case Study 3: FTC Disclosure Cleanup – 82 Posts in 90 Minutes
The Problem
A content site with affiliate revenue had 82 published posts that either lacked an FTC disclosure entirely or carried one that did not meet current placement requirements. With enforcement activity increasing in the affiliate space, this was an active compliance risk, not a theoretical one.
The Build and Outcome
An n8n workflow fetched each post via the WordPress REST API, used Claude to detect whether a compliant disclosure was present and correctly positioned, then patched non-compliant posts by injecting a standardized disclosure block at the top of the content field. All 82 posts were updated in 90 minutes. Five workflow templates that auto-generate new affiliate posts were also updated to include the disclosure block in their output by default, so the problem cannot accumulate again.
Case Study 4: Daily Email Triage Pipeline
The Problem
The Auburn AI consulting inbox was getting a mix of inbound leads, client follow-ups, vendor pitches, and noise. Reading and triaging manually first thing every morning was eating 20-30 minutes that could go toward billable work.
The Build
An n8n workflow fetches unread messages via IMAP each morning, passes the subject and body to Claude Sonnet for categorization into three buckets – LEAD, CLIENT, or VENDOR – and drafts a suggested reply for anything in the LEAD or CLIENT category. Results land in a lightweight CRM-style Google Sheet with the category, a one-line summary, and the draft response ready to review and send. Nothing is sent automatically; every reply goes through a human eye before it leaves the inbox.
Productized Offering
This workflow is available as a productized build for consulting clients who want the same setup adapted to their inbox and their categories. The base configuration takes about three hours to install and tune.
Case Study 5: AI-Narrated Podcast Pipeline
The Problem
Publishing a daily or near-daily podcast alongside a content site is genuinely time-consuming when you are recording, editing, and uploading manually. The goal was to convert existing blog posts into listenable episodes without adding daily manual work.
The Build
An n8n workflow pulls newly published posts, sends the cleaned text to ElevenLabs for text-to-speech narration, receives the audio file, and publishes it directly to Buzzsprout via the Buzzsprout API – complete with title, description pulled from the post excerpt, and episode number. The workflow runs on a schedule. After the initial setup and voice selection, the pipeline produces five episodes a week with zero manual touchpoints.
What These Cases Have in Common
Looking across all five, a few things hold:
- Fast turnaround. None of these took longer than one working session to build and ship. Most took two to four hours including testing.
- Fixed cost. API costs were predictable before the build started. The SEO rewrite case is an extreme example at $0.14, but even the larger workflows ran within a defined budget.
- Real error handling. Every workflow has a failure branch. Silent failures are the worst kind – they look like success until something downstream breaks.
- Repeatable as productized workflows. Each of these can be adapted and delivered to another business with similar infrastructure in a fraction of the original build time.
Have a Workflow You Would Rather Not Run by Hand?
If something on this list looks close to a problem you are sitting on, or if you have a repetitive process that belongs in a workflow and not on your daily task list, email alexander@auburnai.ca. A short description of what you are running and what the manual version currently costs you in time is enough to start a conversation.
Have a process you would rather not run by hand anymore? Email Alexander.