E-commerce
Tail Friends: a pet store that AI built in a week and now runs every day
A real Shopify brand, live in 18 countries. AI researched the products, wrote every listing, set the prices, made the videos and the music, runs the social accounts, watches the ads twice a day and pulls the plug on losers, finds new products every Sunday and hands me a KPI review every Monday. I approve. It does the rest.
I had built two Shopify stores before AI. Each took weeks, a lot of evenings and a lot of money on apps. Tail Friends went from idea to live store in about a week, with a couple of hours a day from me. The full blueprint is here. This page is about what happened after launch, when the store had to run.
A week in the life of the store
- Every morning a digest tells me what happened overnight: orders, ad spend, anything stuck, anything that needs a yes.
- Twice a day an ads monitor reads Meta and TikTok, applies written kill rules (stop an ad that has spent past a limit with nothing to show for it) and recommends, never makes, any increase in budget.
- Every day an engagement run works through Instagram and TikTok: likes, short comments on other people’s pet posts that react to something actually in the video, and a small number of messages asking creators for permission to repost their clips, all logged so nobody gets messaged twice.
- Twice a day new posts go out to Instagram, TikTok and Facebook, but only the ones I marked Approved in the review sheet.
- Every Sunday a research job scans supplier catalogues for new products, prices them against the store’s margin rule and shortlists the best five.
- Every Tuesday product pages without a demo video get one, checked for warped paws and floating products before upload.
- Every Monday the KPI review ends in decisions, not charts.
The content machine
Product videos start from real product photos and are turned into short clips with Higgsfield. Every Reel gets an original instrumental from ElevenLabs. Captions follow one format, and nothing is published until I tick a box. The AI label stays on where the platforms ask for it.
The rules that make it safe
One hub document holds the rules every agent reads before touching the store: price at about two and a half times landed cost, never under a 40% margin, one welcome code only, no fake reviews, no fake discounts, no “best” or “number one”, no medical claims, nothing published and nothing spent without my approval. The same rules live in the shared brain that every job reads. When a rule changes, it changes once.
What broke
Plenty, and it is the useful part:
- Paid ads were humbling. Meta delivered clicks at prices that made the maths hard for a new store with no reviews. TikTok sent traffic that left almost immediately. The kill rules caught the worst of it, and one week a payment error meant the ads delivered nothing at all without anyone noticing for days. Every rule on spend now has a written cap, and nothing restarts without one.
- Browsers are moody. Instagram served a phone-sized layout to a desktop window one morning and the engagement run could only like posts, not comment. The run now drops to likes only instead of failing when that happens.
- Supplier logins expire on a timer. Exactly 72 hours, whatever you do. The daily run now checks the clock and warns me before it lapses.
What I would tell you
Write the operating rules before the first product listing, and put them in one place every agent has to read. The store is only as good as the rules it follows when you are asleep.