"An e-commerce store that runs itself" is a phrase I usually don't trust - so I want to be honest about which parts of it are real

If you've been around IH for more than a week, "launch a store that runs itself" probably sets off the same alarm it sets off in me. It's the exact phrasing of every dropshipping course and passive-income funnel that's ever separated a founder from $997. So I want to talk about the version of that idea I'm actually building - and be specific about where it holds up and where it doesn't yet.

Here's the premise I keep coming back to. If you sell products online, the product is maybe 10% of the job. The other 90% is machinery: a storefront that converts, ads that bring traffic, descriptions written for every SKU, support answered for every "where's my order," and someone quietly processing orders so nothing falls through. Most sellers I know spend nearly all their time on that machinery and almost none on the thing that actually differentiates them - choosing good products.

So the question I got interested in wasn't "can AI do e-commerce." It was narrower and more useful: which parts of running a store are genuinely delegable to agents right now, and which parts still need a human no matter what the landing page says?

What I think actually works today, based on running it:

Store setup. Standing up a storefront that's ready to sell, instead of wrestling a template for two weeks. This one's solidly real - it's structured work with clear inputs.

Product content at volume. Descriptions, copy, and SEO written per item. This scales well because the marginal cost of the 200th description is near zero, and that's exactly where humans burn out and cut corners.

First-line support. 24/7 answers to the repetitive 80% of tickets. Real, with a caveat below. Order ops. Processing and keeping the backend moving so things don't slip through cracks.

Where I won't oversell it:

Ad campaigns can be launched and managed by an agent, but the judgment calls, is this creative on-brand, is this a market worth spending into, still benefit from a human in the loop. Fully hands-off ad spend is where I'd tell you to be skeptical of anyone's claims, including mine.

Support handles the repetitive volume well, but the angry-customer edge cases and anything touching refunds or trust, you still want a person seeing.
Product selection, the actual taste, the "will this sell" instinct, is the one thing I have zero interest in automating. That's the part that's yours. The pitch isn't "AI picks your winners." It's "AI runs the store so you have time to."

That's the honest shape of it: the operations become largely autonomous, and the founder's job collapses down to the one high-leverage thing only they can do: find and choose products worth selling.

What I'd love the IH crowd's read on:

If you sell products, what % of your week is machinery vs. actually sourcing/choosing products? Does the 90/10 split match your reality or am I overstating it?
Which of these would you actually hand to an agent tomorrow, and which would you never let out of your own hands and why?
For those who've tried automating any of this: where did it quietly break?

Full disclosure so it's not a stealth pitch: this is the problem I'm building on (Worknet - the e-commerce agents handle store, ads, content, support, and ops). So I'm biased. But I'm more interested right now in pressure-testing which parts are actually delegable than in selling anyone anything - and if the answer from people doing this daily is "the automatable parts aren't the hard parts," that's the most useful correction I could get.

Where's the line for you?