AI that does the work
e-commerce founders shouldn't.
Bespoke AI tooling for the workflows draining your team — returns, post-purchase, creative pre-pro, finance reconciliation. Built inside your stack, owned by your team, paid back in months.
Two modes: Teach (we set it up together) or Do (I build it end-to-end). Six-week minimum for Do mode.
Teach mode. Or do mode.
Pick the path that fits your team. Either way, you own the build at the end.
Teach mode
We work inside your stack — I show your operator exactly how to use Claude / GPT / Shopify's AI toolkit for the workflow draining the most hours. You leave with the skill, not a dependency on me.
- ✦1–2 half-day workshops · live
- ✦2–3 weeks of follow-up support
- ✦Your stack, your prompts, your owner
Do mode
I build the full system end-to-end inside your stack, pilot it with real cases, then hand it over with a plain-English run-book. Six-week build plus 30-day stabilisation.
- ✦Full build by me · 6 weeks
- ✦30 days stabilisation included
- ✦Plain-English run-book + training
Six places AI actually earns its keep.
These six workflows pay back inside two months on average. The other 80% of "AI for e-commerce" use cases don't — and I'll tell you straight.
Content generation
Bulk product descriptions, SEO copy at scale, newsletter drafts in your brand voice.
- ●Bulk product descriptions
- ●SEO copy at scale
- ●Newsletter & email drafts
- ●Blog & social from catalogue
Marketing automation
Hook variants for paid social. Subject-line tests at volume. Creative briefs from your top-performing ads.
- ●Paid social hook variants
- ●Subject-line test sets
- ●Creative brief drafts
- ●Storyboard & shot lists
Operations & file work
The boring work that drains your team. Image resizing, thumbnail generation, bulk SKU clean-ups.
- ●Image resize & thumbs
- ●Bulk product imports
- ●SKU & catalogue clean-up
- ●File org & naming
Inventory & pricing
AI watching stock levels. Dynamic pricing rules. Re-order alerts. Margin alerts when a SKU starts losing money.
- ●Stock-level monitoring
- ●Dynamic pricing rules
- ●Re-order & margin alerts
- ●Shopify AI toolkit setup
Customer service & insights
Tone-matched draft replies. Tickets summarised. Reviews and feedback clustered into themes.
- ●Tone-matched ticket drafts
- ●Conversation summaries
- ●Review & feedback clustering
- ●Repeat-question playbooks
Business strategy & finance
Sales-trend analysis. Best-seller and dead-stock surfacing. Reconciliation between Shopify, Meta, Klaviyo and your accounting.
- ●Sales trend analysis
- ●Best-seller / dead-stock
- ●QuickBooks · PayPal · HubSpot
- ●Daily margin reconciliation
The path from idea to compounding savings.
Twelve months. Five phases. From naming the workflow to owning the savings.
The workflow is named
We've shadowed your team, identified the workflow draining the most hours, and signed a one-page scope.
A working build in your stack
Inside Klaviyo, Shopify, Notion — wherever it belongs. Your operator has demoed it twice. The first hours are already being saved.
Stabilised + owned
Defect count is below one a week. The named owner on your team can fix it at 9pm. Payback timer started.
Part of the operating system
The team can't remember how they ran the workflow before. The hours saved are reinvested somewhere more strategic.
Compounding savings
Builds in production are quietly compounding. Total client-side savings of $40k–$200k+ logged across the engagement.
Recent builds. Real workflows.
Six workflows out of the last 27. Pick a similar size and category to yours.
Product description generator
Newsletter campaign drafter
AI returns desk
Sales-trend analyst
Customer-service co-pilot
Inventory & re-order alerts
Two modes. Two price ranges.
How the price compares.
| Option | Cost | Notes |
|---|---|---|
| Teach mode (1–2 sprints) | $2,400 – $4,800 | We set it up together · your team owns it |
| Do mode (full build) | $14,800 – $24,000 | Six-week build · 30-day stabilisation |
| Hire a senior AI engineer | $15–22k/mo | + benefits · 12-mo ramp · long search |
| Pay for a vendor SaaS | $2–8k/mo forever | Locked into someone else's roadmap |
Who this is — and isn't —
I turn down about half of all AI build applications. Here's the line.
- ✦DTC brand, $1M–$30M ARR
- ✦Stack is Shopify + Klaviyo + Slack / Notion
- ✦A workflow is draining real hours
- ✦You can name one operator to own the build
- ✦You'd rather own the tool than lease a SaaS
- ·You want a SaaS product (use them — they exist)
- ·You can't name an operator to own it
- ·Pre-revenue — wait until you have orders to automate
- ·Your team won't show up for the build sprints
- ·You want AI doing work no human reviews
"Half the value came from the AI systems she built into our returns desk. The first month paid for the year — and the team actually uses it on Mondays."
The questions founders ask first.
Six honest answers to the six questions every applicant asks.
What's the difference between Teach and Do mode? ＋
Teach mode = we set the tool up together inside your stack — I show you exactly how to use Claude / GPT for the workflow draining your team. You leave with the skill. Do mode = I build the full system end-to-end and hand it over with a run-book.
What stack do you build inside? ＋
Whatever you already pay for. Most builds live inside Shopify, Klaviyo, Notion, Slack, Zapier or n8n, with thin custom code where existing tools fall short. I won't sell you a new SaaS subscription as a side-effect.
Do I own everything? ＋
Yes. Every prompt, every connection, every line of code. The run-book, the model choices, the API keys — all yours, in your accounts. No hidden middleware on my server, no licence fees.
What's the smallest project you take on? ＋
$2,400 in Teach mode — usually one workflow, one team member, one half-day session plus follow-up. Smaller than that, you're better off in the Community.
What if the AI gets it wrong? ＋
Every build I ship has a human approval step somewhere in the loop — even when the AI does 95% of the work. Fully autonomous workflows have a defect rate e-commerce can't absorb at scale.
Do you work with Claude, GPT, or both? ＋
Both, and Gemini where it fits. The model choice is whichever earns its keep for your workflow — usually Claude for long-form content and customer service, GPT for quick edits and code.
Ready to give your team
the workflows they want?
AI builds · by application. I read every application personally and reply within two business days.