Twice a year, we check whether Shopify just made us obsolete

Field notes on platform incentives, from a team that builds on the platform and wants it to win.

Twice a year, Shopify publishes an Editions drop, and twice a year we read it the way you read a weather report before a flight. Somewhere in that list, the theory goes, is the feature that makes a catalog operations company unnecessary.

We have run this ritual for a while now. The feature never lands.

The spreadsheet of overlaps

It is not because Shopify's AI is weak. The Spring '26 Edition alone put Sidekick on every screen of the admin, added Sidekick Pulse to surface recommendations before anyone asks, wired partner tools straight into the conversation through Sidekick App Extensions, and set Campaign Autopilot loose on Facebook, Instagram, Shop, and email. Shopify Magic writes product descriptions, strips image backgrounds, and drafts store themes. Free, on every plan.

Read that list cold and our business should be dead.

For a while we half believed it. Every Editions morning we mapped the announcements against our own workflow list and marked the overlaps (we kept the spreadsheet, and it was not a relaxing spreadsheet).

The flattering theory

Our first explanation was that we are more specialized and more careful than a platform can afford to be. Comforting, and wrong in the way flattering explanations usually are.

Capability was never the question. Shopify employs some of the best commerce engineers working; anything we can build, they can build. The question is who they have to build it for.

Who the feature has to serve

Shopify serves millions of businesses in 175+ countries. That span runs from a store selling its first candle to a store whose catalog no human has read end to end, and every feature that ships has to survive contact with all of it.

When one feature serves that whole span, the median wins. Not by anyone's choice; by arithmetic. A feature included free with every plan is a feature engineered for everyone, and "everyone" has a shape: general by design, safe by default.

That is our read, and we will say it plainly. Horizontal is not a limitation of platform AI. It is the job description.

What the median merchant needs from AI is help in the moment: a description written now, a discount created by voice. Sidekick does this genuinely well, which is exactly what it is for.

A catalog above $250K a month needs a different species of thing. The same enrichment decision applied identically across every variant, verification before anything publishes, rollback after, and a name attached to the error rate.

A platform cannot ship "a name attached to your error rate" to millions of merchants. Accountability does not scale horizontally. That one sentence is most of our business model.

Two layers, both growing

None of this is new. Shopify's own App Store holds 21,000+ apps, twenty years after the platform launched and after every native feature wave in between. The floor keeps rising: whatever everyone needs becomes platform. The ceiling keeps moving too: whatever one account needs, done accountably, stays specialist work. Both layers grow at the same time, which is the part the obsolescence theory always misses.

The same Edition that expanded Sidekick also shipped Shopify Catalog, the pipe that feeds structured product data to AI shopping surfaces, with Shopify reporting that AI searches running on its structured data convert at 2x the rate of scraped data. Our reading: the platform just raised the price of a messy catalog again, for everyone, and left the cleanup to whoever does vertical operations.

Our bet, on the record

Platform AI stays horizontal. Not for lack of ambition, but because the incentives that built a platform clearing $100 billion in GMV in a single quarter point that way, permanently: ship for the median and price it at zero.

Check us on it. If by mid-2028 Sidekick or its successor takes continuous, named ownership of individual large catalogs, error budget included, we were wrong, and we will write that post too.

And the honest line under the thesis: below roughly $250K a month, do not hire us or anyone like us. The native AI is already in your admin at no extra cost, and for most stores it is genuinely enough. The gap we live in starts where "generate a description" ends and "run the catalog" begins.

Where this leaves you

• Platform AI is built for the median merchant because it has to be. Arithmetic, not attitude.
• Watch what a platform gives away free to everyone; that is where the horizontal layer sits.
• The specialist layer above a platform grows with it, not against it. 21,000+ apps say so.
• If the native AI covers your needs, use it. The line is accountability, not intelligence.

Next Editions morning we will open the notes again, same ritual, same spreadsheet. We expect to be impressed. Not replaced.

If you want to know which side of the line your catalog sits on, OKART's Infrastructure Stress-Test maps it, read-only, in a few days.

Sources

1. About Shopify: merchant scale (millions of businesses, 175+ countries), 21,000+ apps in the App Store, platform released in 2006.
2. Shopify Sidekick: official capabilities and "included with your Shopify plan" availability.
3. Shopify Magic: free AI feature suite, text and media generation, available on all plans.
4. The Spring '26 Edition: Sidekick App Extensions, Sidekick Pulse, Campaign Autopilot, Shopify Catalog and the 2x conversion claim.
5. Shopify Q1 2026 results: over $100 billion in quarterly GMV, quarter ended March 31, 2026.