Case Study
Tour World Crafts: AI at work in a one-person business
Tour World Crafts sells folk art and ethnographic objects from the Craft World Tours Collection, assembled over four decades through a travel organization that operated from 1969 to 2011. I curate the collection and run every part of the business myself: listing, photography, restoration and packing. AI is what lets one person do that carefully.
The challenge
Every object needs a careful identification, a provenance story that is honest about what is and is not known, a price I can defend, and a complete listing that matches the fields of my store and my eBay integration. Doing that for a large collection, one piece at a time, by hand, does not scale.
The approach
- A written SOP. A Master Product Listing SOP & Template defines a strict, 12-section listing format whose field order mirrors the WP Lister dashboard. See the listing workflow.
- An independent second opinion. I identify each object with Google Lens first, and the AI makes its own identification before it sees mine, so the two can genuinely disagree.
- Provenance cross-checks. Every listing is checked against the public Craft World Tours Itinerary Archive.
- Honest language by design. Listings use probabilistic wording such as “likely,” “possibly” and “suggests,” and never assert provenance that cannot be confirmed. See prompting and guardrails.
- Enforced syntax. A finished SKU format (CWT-[Region]-[Year]-[Number]) is applied consistently, and every edit returns the complete listing, never a partial update.
- Pricing with reasons. The AI researches and proposes a price with its rationale; I make the call.
What stays human
AI does not photograph, restore, handle or pack anything, and it does not get the last word on what an object is. Condition, authenticity judgment, photography and customer service stay with me. The AI drafts and checks; I decide.
The stack around it
The same approach extends to the rest of the operation: WordPress and WooCommerce for the store, eBay for a second sales channel, MailerLite for the newsletter, Jetpack for forms, and Google Analytics 4 with consistent campaign tagging for attribution. The details are in The Automation Stack Behind the Store.
Takeaways
- Write the process down first. A good SOP makes AI output repeatable.
- Give the AI context, strict output rules and a defined role, then review what it returns.
- Build honesty into the prompt: uncertainty should show up in the wording, not be hidden.
- Keep a human decision at every point that carries risk.
This is the same thinking I bring to client work. See my Web Integration services and Portfolio, or get in touch.
More in this section
- AI Overview
- The AI Product Listing Workflow
- Prompting, Guardrails & Human Oversight
- The Automation Stack Behind the Store
- AI Tools & Resources
Looking for the rest of my work? See the Portfolio, Web Integration services, Projects, Video Learning, or learn more about Marc.