AG3NTIC.AI // VOLUME 1

Shopify is quietly building a new AI stack

Most people missed this. Shopify just picked up Vantage Discovery, an AI search startup from former Pinterest engineers. That’s their sixth AI-related acquisition in the past year.

They’re not bolting on AI. They’re rebuilding the foundation for how commerce will work. Search, merchandising, operations, and marketing are all being rethought from the ground up.

This is infrastructure, not features.

Read more

Amazon launches an agent that actually takes action

Nova Act is Amazon’s new AGI agent. You give it a goal. It opens a browser. It searches, filters, clicks, books, buys.

It’s not public yet but it’s real and working.

This is a serious look at how AI will move from content generation to task execution across the web.

Read more

Alibaba is testing demand before they make anything

They’re generating photorealistic product images with AI, putting them online, and only manufacturing what gets engagement or sells.

No prototyping. No dead stock. No guessing.

And it’s outperforming their traditionally designed products.

This will be standard for brands who understand speed and margin.

Research paper

Etsy and eBay are rebuilding the front end like it’s TikTok

These marketplaces are moving away from static search and filters. They’re leaning into personalized discovery feeds powered by AI.

It feels less like shopping and more like scrolling a social feed. The more you engage, the more it adapts.

This isn’t about improving UX. It’s about increasing session time and conversion by learning the customer faster than they learn what they want.

WSJ article

Old Navy has a smarter ops stack than most DTC brands

They rolled out RADAR across 1,000 stores. It uses RFID, AI and computer vision to track inventory in real time, restock products automatically and sync online orders with physical locations.

It’s not innovation theater. It’s real operational efficiency.

If you’re still doing weekly audits and reconciling across systems by hand, you’re already behind.

More here

^{What’s on my mind this week…}

MCPs are the system layer that breaks the current SaaS model

Every SaaS tool today is built around its own UI, its own database, and its own workflow. Each one needs to be logged into, configured, checked. Most of the intelligence still comes from the operator. The tools just help execute.

Agents shift some of that, but on their own, they don’t solve the core problem. You still end up with isolated processes. You’re just triggering them with prompts instead of buttons.

Model Context Protocols change this completely.

MCPs give agents a shared understanding of your business. They let agents and other processes operate with access to real-time goals, state, memory, decisions, and priorities. Not just data. Context.

When an agent sees that conversion is down, it doesn’t just report it. It knows what products are affected. It knows which offers were running. It knows what changed in traffic. It knows who else is working on the same problem. And it can do something about it because everything is speaking the same internal language.

This removes the need for the human to sit in the middle of a bunch of tools trying to connect dots. The coordination happens inside the system.

What replaces SaaS isn’t one app. It’s a network of agents working off the same context layer. That’s how pricing, lifecycle, attribution, support and forecasting all stay in sync without teams of people managing them manually.

MCPs don’t improve software. They change what software is.

They’ll feel invisible. But they’ll be the most valuable layer in the stack. Whoever owns that wins.

If you’re building something real in AI and commerce, send it in

This newsletter is for people actually doing the work. If you’re building, testing or shipping something with real impact, drop me a note. If it’s good, we’ll feature it.