Agentic shopping is already here. While Shopify rolls out things like the global catalogue, UCP and Agentic Channels, some platforms are already a couple of years into that journey. Michael Tutek is one of them. I remember loading up the Arc'teryx site one day and there it was - an AI shopping assistant, live and working, while most of the industry was still writing think-pieces about whether this stuff was even real.

Michael is the founder and CEO of Preezie, the guided-selling and agentic commerce platform working with brands like Puma, Lorna Jane and Arc'teryx. He started on the sales floor at The Good Guys, clocked that 30% of in-store visitors bought something against barely 1% online, and has spent the best part of a decade trying to close that gap.

We get into why Michael thinks a chat box bolted onto your store isn't the future, and why agentic commerce is. Plus how you stop an AI assistant confidently making things up, why feeding it your old human chat transcripts can make it worse not better, and the out-of-stock feature that's worth 20-30% of the whole tool's value on its own. We also get into why Michael flatly refuses to build products that try to change shopper behaviour.

This episode is sponsored by The Support Heroes - dedicated Shopify support agents who actually know what they're doing. Head to thesupportheroes.com.

Also brought to you by Product Pelican (https://apps.shopify.com/product-pelican), the native PIM for Shopify. Michael's whole point is that good AI starts with good data - Product Pelican keeps your catalogue clean and consistent at scale, so whatever you bolt on top of it, agentic or not, has something solid to work with. Install it from the Shopify App Store.

In this episode:

From throwing-water-on-your-face mornings to The Good Guys sales floor: the 30% vs 1% insight that started everything
Why Preezie pivoted from e-commerce quizzes to a full agentic shopping platform after one day with Microsoft
The case against conversational commerce: why agentic beats a chatbot box
Keeping AI honest: onboarding it like a new hire, and why your old human chat logs can make it worse
The out-of-stock play worth 20-30% of the tool's value on its own
Why it still matters to let shoppers scroll through 10 pages of products
Turning real shopper questions into user-generated FAQs for SEO and GEO
Why Michael refuses to build products that try to change shopper behaviour
The roadmap: an "e-commerce brain" built natively for the AI era

YouTube chapters:

00:00 — Cold open: agentic shopping is already here
01:51 — From bricklayer's son to The Good Guys sales floor at 14
03:19 — The 30% vs 1% conversion gap that started the obsession
04:44 — Quizzes first, then Microsoft's tap on the shoulder about AI
07:04 — Spotting Preezie live on Arc'teryx before anyone else had it
07:48 — Sit and wait, or do something: why agentic beats conversational
10:30 — The out-of-stock feature worth 20-30% of the tool alone
13:40 — Training AI like a new hire, and performance reviewing it
16:56 — Killing support tickets with a mini AI on the contact form
18:12 — How much instruction is too much (it depends on the brand)
20:37 — Turning shopper questions into content: transcripts, FAQs, SEO and GEO
23:59 — Why the paradox of choice means discovery isn't going anywhere
28:17 — Don't try to change shopper behaviour: the electric car test
31:19 — The roadmap: building an e-commerce brain for the AI era
32:01 — Wrapping up