There's an objection you hit within the first two minutes of any conversation about agent commerce: won't Shopify (or Stripe, or Google) just do this for you? It's a good question, and the honest answer has two halves. You need both — because if you only believe the first half you build the wrong company, and if you only believe the second you sound like you're in denial.
The first half: yes, the platforms are building the rail. Shopify has wired over a million merchants into ChatGPT's Instant Checkout. Google launched its commerce protocol with Target, Walmart and Etsy among the founding partners. For a catalogue product sitting on Shopify, the buy button is genuinely getting solved — and pretending otherwise is how you end up competing with a giant on its home turf. So we don't build the buy button. We never will.
A rail is not a meter
Here's the distinction the objection skips. A platform builds the plumbing. It will never be the thing that tells you the plumbing is failing you. Shopify will happily show you a green “✅ connected” badge. It will not tell you that an agent tried to buy your hero product and gave up because the size variants were ambiguous, or the price only existed inside an image. You don't ask the restaurant to write its own reviews — and you don't ask the rail-builder whether the rail is letting you down.
We've seen this film before. Google built Search — and rather than killing the optimisation industry, it created it: Ahrefs, SEMrush, a multi-billion-pound layer that exists precisely because Google won't grade your pages for you. Stripe built payment rails, and a whole analytics-and-optimisation stack grew on top of them. The pattern is dependable: the rail's existence is what creates demand for an independent meter sitting above it. Agent commerce is the next instance of the pattern, not an exception to it.
Four things the platform structurally won't do
It won't grade its own homework. A neutral “did the purchase actually complete, and where did it drop out” is something the rail-builder has no incentive to measure against itself. It won't guarantee your data is complete. The platform hands your catalogue to the agent; whether every line carries the price, stock, shipping and variant detail an agent needs to transact is your problem — and the data is blunt that completeness is the single biggest driver of outcomes. It won't cover anything off the catalogue. Your B2B side, your services, your brand-level terms, the parts of you that aren't a tidy product — none of that is in the feed. It won't feed anyone but its own partners. Shopify feeds OpenAI; if you also want to be ready for Perplexity, Gemini or whatever ships next, you need a layer that isn't betting the shop on one roadmap.
And most of the world isn't on that platform anyway
Shopify is roughly a quarter of identifiable stores. WooCommerce — about 4.5 million self-hosted stores — has no one centrally making it agent-ready. Neither does the custom, headless and B2B world, which is where the largest pool of commerce by value actually sits. For most sellers, “the platform will handle it” isn't a competing offer — it's a thing that was never on the table.
So the position isn't against the rails. It's on top of them. Use Shopify's checkout. Use Stripe's. Use Google's. Then use Actum to prove it actually works, fix the data that doesn't, and cover everything the rail never touches. We don't sell the buy button; we sell the proof that buying works — and the reach beyond the catalogue.