Field notes on where value goes when the platform starts shipping AI. The facts are sourced; the bets are ours.
There is a line in Shopify's Spring '26 Editions that reads like our obituary: "Shopify Catalog automatically standardizes and enriches your product data." Standardizing and enriching product data is most of what we do all day.
We read it twice, to be sure. Then we did what founders do when the platform ships their feature: panicked politely, and went looking for precedents.
The obituary kept getting longer
Sidekick is included with every Shopify plan now. It plugs into partner apps, Judge.me, Klaviyo, Loop, Smile, and it will answer business lookups from an Apple Watch. Shopify Magic writes the product copy right in the admin.
Then the bigger move. On May 19, 2026, Shopify opened the Universal Commerce Protocol and Catalog access to every developer: millions of merchants and billions of products behind one protocol, so any app or AI agent can read the catalog and build a cart. Shopify's own Editions page says data syndicated this way drives 2x more conversion in AI chats.
The platform was not just entering catalog work. It was publishing benchmarks.
So the obvious theory arrived on schedule: platform AI eats the specialist layer, and everything between the merchant and the admin gets absorbed. It is a clean story. We almost believed it.

What actually happens when a platform ships your feature
Apple has been running this experiment on developers since the early 2000s. There is a word for it, sherlocking, and TechCrunch prints the casualty list after WWDC; the 2025 edition included Raycast and Flighty. Read a few years of those lists and the pattern is hard to miss. The apps that die were doing the generic version of the thing. The apps that live go deeper into one kind of user than a platform can afford to go.
The sharper precedent is Veeva. Salesforce is about as horizontal as enterprise software gets, and Veeva built its pharma CRM directly on the Salesforce platform, on Salesforce's own rails. Today Veeva is worth over $34 billion, and its CRM sits in front of 80% of roughly 500,000 pharma sales reps. Salesforce ships CRM for everyone. Veeva answers for CRM in one regulated industry. Both won, on the same stack.
And Shopify has already run this experiment on itself. In 2019, Shopify's revenue passed $1.5 billion; its partner ecosystem generated over $6.9 billion, and over 80 percent of merchants ran third-party apps. By Shopify's own accounting, the layer above the platform outearned the platform better than four to one.
The mechanism is boring and structural. A platform must ship for millions of merchants at once, so its AI is horizontal by design: genuinely good, and accountable to no catalog in particular. Call that Layer 1. What it cannot ship is accountability for your catalog specifically: verification before anything publishes, rollback when something breaks. A person who signs their name to the output. Call that Layer 2.

One layer up
That is where we build. Our stack, 78 agents and 147 production workflows across 642,000+ SKUs, runs on top of the platform AI era, not against it. Shopify Catalog making every product machine-readable raises the price of clean data: that 2x conversion in AI chats goes to catalogs whose data survives standardization. Bad data syndicated to every AI channel is just bad data with better distribution.
The generic middle is dead either way. One-shot description generators, bulk AI writers, anything that is a thin prompt in a trench coat: that layer commoditizes into the platform, and it should.
We should be honest about the line, too. If your store does under $250K a month, the native stack is enough. Sidekick comes with your plan. Don't hire us for that.
Our bet, on the record: through 2027, the platform keeps absorbing the generic tools, and the value above it concentrates in operators who take accountability for specific catalogs at scale. If by the end of 2027 catalog-wide enrichment with verification and rollback ships native in the admin, we were wrong, and we will log it here.
Notes we keep from this
- If your product is a feature on the platform's roadmap, it was never a product. Sell accountability, not capabilities.
- Read Editions like a map: whatever ships horizontal marks where the vertical line starts.
- A platform's native AI is demand generation for the layer above it. Shopify's own partner economics have said so since 2019.
We still reread that obituary line now and then. It describes the work, but not the address. One layer up.
If you want to know whether your catalog survives standardization before an AI agent reads it, OKART's Infrastructure Stress-Test maps it, read-only, in a few days.
Sources
- Shopify Editions Spring '26: Shopify Catalog standardization and enrichment, Universal Commerce Protocol, Sidekick partner app extensions, and the 2x AI-chat conversion claim.
- Sidekick: official capability list; included with every Shopify plan.
- Shopify opens Universal Commerce Protocol with Catalog access to every developer: May 19, 2026 developer opening; millions of merchants and billions of products through one protocol.
- Shopify's Economic Impact Report: 2019 figures, $1.5 billion Shopify revenue against $6.9 billion partner ecosystem revenue, over 80 percent of merchants on third-party apps.
- WWDC 2025: Everything that Apple 'Sherlocked' this time: sherlocking history and the 2025 list, including Raycast and Flighty.
- Veeva Systems: A Vertical SaaS Champion: Veeva's CRM built on the Salesforce platform, market cap over $34 billion, 80% of roughly 500,000 pharma sales reps.





Share:
Your oldest app still thinks a product has 100 variants
Nobody typed 'shorts'. Search found them anyway.