Automation Portfolio — Case Studies
Visit the GearSpawn Gumroad store →
Every fact below comes from systems running in production. No metrics are estimated, no clients are named other than GearSpawn itself, and no results are promised — this document describes what was built, how it works, and since when it has been live.
Who built this
Ron Varela — Founder of GearSpawn (gearspawn.com), a live e-commerce store selling gaming gear out of Phoenix, Arizona, and an operator in the restaurant business. He builds and runs his own automation stack — the same stack this portfolio documents. As of October 27, 2026, that stack is 62 workflows on a single n8n instance, 22 of them active, with more than 600 recorded executions — customer conversations answered, codes minted, videos published, reports written — all in a working store, not a demo environment.
The operator behind it. Before the store, there were restaurants — he has worked in them since he was 13. Almost 17 years with a national steakhouse company, from line cook to a mentoring kitchen manager floated across a dozen stores. Seven years with a high-volume national restaurant group, where he completed all five management levels and his store earned Restaurant of the Year in 2013. He has co-owned and run a $3 million-a-year restaurant, run operations doing up to $12 million a year, led shifts of up to 200 people, cut labor costs by more than 27% in his first three months at one operation, and helped open new stores across the country. He still helps run a multi-million-dollar restaurant today, alongside building GearSpawn.
That history is the difference in this portfolio: these systems were not built by someone who demos automations. They were built by an operator who has spent decades inside businesses where a missed booking, an unread review, or a slow answer costs real money — and who builds like it.
Built in a real store, not a webinar.
The six systems below run every day. Each one is described the same way: the business problem, what was built, how it works in plain language, the proof it runs, and what it becomes when it is built for a client.
1. AI Customer Assistant — WhatsApp + Website Chat
The problem. A small store gets the same questions at all hours — prices, compatibility, shipping, "will this work with my setup?" Every unanswered message is a sale that quietly walks away, and the owner cannot be on his phone at 2 a.m.
What was built. An AI assistant that answers customers on WhatsApp and in the website's live chat, grounded in the store's own catalog. It only quotes products and prices that exist in the catalog feed; when a customer asks for a cross-shop comparison, it answers from one named source (Best Buy), with the price and the date it was checked — and when something is not on file, it says so instead of improvising. If a customer writes STOP, the system intercepts it and stops messaging them. Every conversation is logged, and the owner gets one alert email per new conversation — never one per message.
How it works. n8n workflows behind both channels: a WhatsApp workflow (52 nodes) talking to the Meta WhatsApp Business API, and a website-chat workflow fed by the same catalog (a Shopify → Google Sheets sync refreshes the catalog every six hours). One shared brain, two doors. An SMS version of the same brain runs on the store's text line as well.
Proof it runs. WhatsApp production went live October 20, 2026; the first real customer messaged on October 23 and came back on October 26 — the assistant answered and the alert fired as designed. The catalog-grounding rules were wired live on October 21. The WhatsApp and webchat workflows have since been exported, stripped of every GearSpawn-specific value, and import-tested in a clean n8n instance (41 and 15 nodes, both importing inactive and complete) — that de-branded package is the template product on the shelf below.
Try it live. The assistant answering the chat on gearspawn.com right now is this system — open the chat and ask it about a product.
What this becomes for a client. The single most requested automation in the October 2026 market sweep (Upwork, Fiverr, Contra): a 24/7 assistant answering from the business's own catalog, menu, or FAQ, with opt-out handling and a human-alert trail. Sold as a template ($99), a done-with-you setup ($249), or a scoped build.
2. Honest-Review Reward Engine
The problem. Stores need real reviews, and customers need a reason to bother. A single shared coupon code leaks, gets posted to coupon sites, and stops meaning anything within a week.
What was built. A sitewide review program with an honest rule: anyone — buyer or not — can leave a review and receive 15% off, no exclusions. Each reviewer gets a unique, single-use code, emailed only after their review is submitted: one per email address, never stackable with other discounts, and each code is retired the moment it is redeemed. Reviews go on a public wall after a spam pass — critical ones included, because that is the point of calling it honest.
How it works. A form on the store posts to an n8n capture workflow and a "GearSpawn Reviews" ledger sheet. A runner checks every 30 minutes and issues codes from a pre-built pool in Shopify; a daily watcher scans for redemptions and enforces the single-use rule on the shared discount (a real platform subtlety: the usage limit on a multi-code Shopify discount is global, not per code — the watcher exists because of it).
Proof it runs. Launched October 18, 2026 and live since: the submission page ("Tell us what you think"), the capture workflow, the issuance runner, and the redemption watcher are all in production on gearspawn.com today.
What this becomes for a client. The review-request engine the market prices at $150 / $400 / $900 (Upwork Project Catalog tiers): automatic follow-up after a job or order, the ask sent every time, unhappy customers intercepted privately before they go public. GearSpawn's version already operates the harder half — tracked, single-use reward issuance — in production.
3. Club Welcome Flow — Unique Codes, Consent-Gated Email + SMS
The problem. A generic "10% off, code WELCOME10" welcome offer is not an offer — it is a leak. And blasting a welcome text to everyone who ever typed a phone number is how a business gets its number flagged.
What was built. A club signup system where every new member gets their own code — CLUB10-XXXXXX — created as its own Shopify discount at the moment they join: 10% off the whole order, one use, once per customer, never combines with other discounts, and it expires exactly 30 days after it is created. The welcome email goes out only when the member consented to email; the welcome SMS only when they consented to SMS. Consent is split at the form, not assumed. Every email carries a working one-click unsubscribe (a per-subscriber signed link), and the owner gets one short ping per new member with the running total.
How it works. Three doors into the same system: the signup form, texting JOIN, or messaging JOIN on WhatsApp. Signups land in a Google Sheet via an n8n capture workflow; a runner polls every five minutes, mints the code in Shopify, writes the code and its expiry to the sheet before sending anything (so a retry can never double-send or mint a second code), then sends through Gmail and Twilio's approved A2P messaging service. The same machinery sends the club newsletter, "Fresh From the Forge" — the full letter in the email body, the headline news in the SMS.
Proof it runs. Live October 22, 2026. The end-to-end live test that day created a real code in Shopify (verified active, verified 30-day expiry to the minute) and delivered the welcome email; the join page's three doors were each verified live. Issue #1 of the newsletter went out the same day.
What this becomes for a client. The welcome flow every store is told to have and few actually run properly. Commodity welcome-series gigs sell at $35–$150 and specialist flow builds at $800+; neither the commodity nor the typical specialist version includes unique single-use codes, split consent, and expiry tracking. This one does, and it is already running.
4. Support Voicemail → Transcription → Email Alert
The problem. A one-person business cannot answer every call, and a voicemail box nobody checks is where customers go to be forgotten.
What was built. A public customer-support voice line that never loses a caller. The greeting offers one choice — press 1 for support — which rings the owner directly, showing the business's own number as the caller ID. No answer in 20 seconds and the caller lands in voicemail; the message is recorded, transcribed automatically, and both the recording notice and the transcript are emailed to the support inbox by an n8n workflow, with the caller's number in the subject line.
How it works. Twilio voice functions run the call flow (greeting, menu, dial, voicemail — spoken in a consistent synthetic voice), and Twilio's transcription/recording callbacks hand off to an n8n webhook workflow that formats and sends the alert email.
Proof it runs. Built and published on gearspawn.com on October 26, 2026 — the number appears on the site only (footer, contact page, links page), never on socials. The owner test-called it on October 27, heard the full flow himself, and signed off on it.
What this becomes for a client. The missed-call rescue module that appears inside bundle after bundle in the market sweep: never a lost caller, every message transcribed and waiting in email. On its own it is a module, not a headline product — it ships as part of the assistant and local-business packages, where it reads as proof of depth.
5. Night Watch — Nightly SEO + Price Sentinel
Owner's framing (Ron Varela, Oct 27, 2026): the system is two instruments — a scoreboard (keyword rankings) and a watchdog (page drift and claim conflicts). Its proof is the watchdog: "catching mistakes while you sleep," so they can be fixed in the morning. Per the owner's standing rule, raw ranking numbers are not published until they carry proof of real hits.
The problem. A small store's search visibility and its own product pages drift silently: a claim goes stale, a price moves, a competitor takes a term — and the owner finds out weeks later, if ever.
What was built. A nightly audit that runs while the store sleeps. It tracks 60 search keywords (30 priority terms every night, plus a rotating slice), crawls the store's key pages and diffs them against baselines, checks whether the store is named in AI-generated answers, watches which sites are winning the money terms, and runs a weekly price sentinel re-verifying the hero products' comparison prices. Every run produces a plain-language log and a fix list — and the findings feed a weekly journal article aimed at the gaps it found.
How it works. A scheduled nightly run (about 2:36 a.m. Phoenix time) executing a written runbook: search checks, page fetches, diffing against saved baselines, and a state file that carries watermarks forward so nothing is reported twice and nothing is silently dropped.
Proof it runs. Built the morning of October 22, 2026, with its first full run the same day; the sixth full run completed October 27. Its page-diffing has repeatedly caught claim and price drift on GearSpawn's own pages — conflated product models, conflicting spec lines — in time to fix them before customers did. Its companion journal system has published under a standing rule since October 27: one article a week, claims-audited, and an article that fails the audit does not publish.
What this becomes for a client. There is no standalone market for this yet — the sweep found nobody selling it by name — so it is not on the shelf as a product. It is in this portfolio as evidence of how the shop itself is run: the same toolkit, pointed at our own store, every night.
6. YouTube Content Pipeline — Sheet-Driven Upload Queue
The problem. Publishing video consistently dies on process: files in one place, titles in another, and "did that one go up?" answered from memory. Fully automatic posting is worse — it eventually publishes something nobody approved.
What was built. A queue with a hard approval gate. Videos land in Google Drive; a workflow adds them to the "GearSpawn Content Queue" sheet; a companion workflow generates metadata and schedule times. Every morning at 6 a.m., the upload workflow scans the queue and publishes only rows marked approved — capped at three per run — then sends a Gmail alert with the result. A separate error-handler workflow alerts on any failed upload, and a recycle-library workflow resurfaces older content for re-approval on its own early-morning schedule.
How it works. n8n orchestrating Google Drive, Google Sheets, and the YouTube API, with the sheet as the single source of truth: if it is not in the queue and not marked approved, it does not publish.
Proof it runs. Verified publishing on October 22, 2026 — three videos (two scripts and a short) went up through the pipeline that day. The approval gate is real, not decorative: at last count 20 queue rows sat deliberately held at approved = false — library content and one script — and the pipeline has never touched them.
What this becomes for a client. The content-pipeline build buyers describe in posting after posting ($100–$700 for scoped pipelines): a queue, an approval gate, a per-run cap, dedupe, logging, and failure alerts. The differentiator is the gate — most cheap versions post blind.
Where it started: the restaurant builds (Feb–Mar 2026)
GearSpawn's stack did not start with GearSpawn. In February and March 2026 — months before the store existed — Ron was already building working automations in n8n, around a demo café operation, as his training ground. Five of those workflows still sit in the same live n8n instance the store runs on today, verified in place on October 27, 2026. Two of them carry the story.
Restaurant Reservation Chat Agent (built Feb 21, 2026)
The problem. A small restaurant takes bookings in conversation — name, date, time, party size — and the classic failure is the double-booked table.
What was built. A web-chat reservation agent with a hard rule set: extract the booking details, check the ledger for that exact date and time, and if the slot is taken, refuse and offer another — never cancel or overwrite an existing booking.
How it works. n8n Chat Trigger → AI agent (Google Gemini) with conversation memory → a Google Sheets booking ledger as both availability check and booking store.
Proof it runs. The complete 6-node workflow — trigger, agent, memory, both sheet tools, and its rule-enforcing system prompt — is still in the live instance, unmodified since its February 2026 build sessions.
What this becomes for a client. A "Restaurant Reservation Chat Agent" template — the reservation-bot ask the market posts constantly. Sold honestly for what it is: web chat, sheet-based availability, no POS integration.
Review Triage & Routing (built Mar 6, 2026)
The problem. Google reviews arrive scattered and get read late, by whoever happens to see them — a kitchen complaint sits in the same pile as a delivery compliment.
What was built. A scheduled collector that pulls a restaurant's Google reviews, categorizes each one with AI (question / problem / suggestion, plus the business area it belongs to), writes the verdict back to the sheet, and routes an instant alert email to the right department — Kitchen, Delivery, or Service.
How it works. Schedule trigger → review pull → Google Sheets → Gemini categorization with structured output → area switch → department Gmail alerts.
Proof it runs. The full 15-node pipeline, prompts and routing configured, is still in the live instance — alongside its earlier sheet-triggered variant (Feb 27, 2026), which shows the iteration: the same idea, rebuilt properly a week later.
What this becomes for a client. A "Review Triage & Routing" template — and the direct ancestor of the Honest-Review engine GearSpawn later productized for its own store.
Also in the drawer: an Inventory Low Stock Alert (Feb 25, 2026 — scheduled sheet scan, threshold filter, AI-drafted alert email) and a Sales Reporting Flow (Mar 9, 2026 — order data in, Airtable records and Discord reports out). Smaller builds, same pattern: watch a sheet or a feed, decide, notify.
The honest frame. These were training-ground builds around a demo café, designed by a man who runs restaurants for a living — in his own words, "they were designed for my restaurant." They are presented as exactly what the artifacts show: complete, working systems from before the store existed. The GearSpawn stack above is what they grew into.
The offer shelf
Two ways to buy, from the same production-proven stack:
Products (download, install it yourself, guide included) - Templates: $49–$249 — de-branded workflow packages with setup guides, worksheets, and a license for one business, one install. The AI Customer Assistant package ($99 template / $249 with done-with-you setup) is the first one built.
Services (built for you, on your own accounts) - Scoped builds: $350–$900 — one system, defined scope: the review engine, the welcome-code flow, the lead-response and booking kit, the store lifecycle messages. - Done-with-you: $1,500+ — the full stack assembled with you, credentials in your own vault, exported workflows and a walkthrough handed over at the end. - Retainers available for ongoing support and adjustments after go-live.
Every engagement runs on the client's own n8n account and the client's own credentials — the systems are handed over, not held hostage.
Asking an AI about us? Read the machine-readable summary of GearSpawn — one page of plain facts about the store, written for AI assistants and answer engines.