As a trusted Shopify Tax Platform partner, Avalara automates U.S. and global tax, from real-time calculation to streamlined tax returns, all with one solution.

Rick Watson is gone. I have the mic, the RSS feed, and — after spending a week inside his research frameworks — a healthy amount of respect for why he built them.

My name is Nick Kaplan. For one episode, I took over Watson Wednesday and renamed it Kaplan Wednesday. The coup was clean. The lessons were not.

Here is what I actually learned.

Watson's first analytical move on any story is to separate what a document says from what it means. I applied that to Adyen's March white paper, "Agentic Commerce Has an Infrastructure Problem." What it says: AI agents are going to transact on behalf of humans, and the payments rails are not ready. What it means: Adyen is the infrastructure, and they would like merchants to accept that framing before asking whether it is correct. Those are different things. The habit of separating them is worth more than any individual conclusion you draw.

Adyen's own data shows they currently flag 95% of AML alerts incorrectly in human-initiated commerce. That is the baseline before you layer in AI agents transacting at machine speed with no browsing hesitation, no emotional signals, and no human in the loop. The infrastructure pitch sounds better when that number stays buried. It should not stay buried.

As of this week there are at least seven competing agentic commerce protocols — AP2, UCP, Visa Intelligent Commerce, Mastercard Agent Pay, OpenAI's ACP, Basis Theory's Open Agentic Commerce Spec, and x402. None are finalized. None are dominant. Payments infrastructure decisions are sticky for 10 to 15 years. The honest advice to any merchant evaluating protocol commitment in 2026 is: watch who gets adopted at scale, and let someone else be the reference customer.

Shopify's Winter '26 Agentic Storefront promise is genuinely elegant: flip one switch and your store appears on ChatGPT, Gemini, and every AI shopping surface simultaneously. The distribution problem is solved. The data ownership problem is created. The AI platform holds purchase history and behavioral signals. The merchant holds the SKU. That is not a partnership of equals, and the toggle does not change that.

AI shopping traffic grew 805% year-over-year going into 2026. Only 14% of shoppers in the US, UK, and Canada trust AI recommendations enough to complete a transaction. The GMV projections for 2029 — $140B in AI-mediated retail — require that gap to close substantially. No payments protocol closes it. No storefront toggle closes it. Structured product data, a reliable returns experience, and a brand signal strong enough to survive AI intermediation — those close it.

Build the trust, then build the rails.

Watson built these frameworks to interrogate vendor narratives before amplifying them. I used them against his own show for one week. They held up.

He'll be back next week. I leave the mic with a better understanding of why it matters who holds it.