Shopify DotDev 2026: what Toronto told us about the next 12 months of commerce
Five shifts from Shopify DotDev 2026 in Toronto: agentic commerce, Shopify POS, B2B and Collective, and the lesson from Shopify AI agent River.
Key takeaways
- AI has become a buying channel in its own right, and the Universal Commerce Protocol and Catalog API are the plumbing Shopify is laying underneath it.
- AI-driven sessions convert around 80% better than organic search, and about half land straight on a product page.
- Shopify POS UI extensions have matured into a serious build surface, with useful in-store builds measured in weeks.
- Shopify B2B is growing over 100% year on year, and Shopify Collective is now live in Australia.
- River, Shopify’s internal AI engineering agent, co-authors around one in eight merged pull requests, and the lesson is that the platform underneath decides whether AI works.
- Custom builds on Shopify keep getting cheaper, so some app subscriptions are now paying for functionality the platform does natively.
DotDev is the one week a year where you get to hear Shopify’s platform roadmap straight from the people building it, and this year’s event has just wrapped up in Toronto. We’ve come back with a full notebook and a much clearer read on where the next 12 months are heading.
Most events hand you the launch slides and a queue at the microphone. DotDev puts you in front of the product managers and engineers who own the work, and if you ask a decent question you tend to get a decent answer, including the parts Shopify is still figuring out. That candour is where most of the value sits, because it tells you as much about sequencing and priorities as any keynote does.
If there was one conversation running underneath the whole week, it was AI quietly becoming a way that people find and buy things, and what that does to every merchant who has spent the past decade optimising for a search box. Plenty of other things happened, but almost all of them ended up connected back to that.
These are the five shifts that came out of the sessions in Toronto, what each one means for Australian retailers, and what we’d do about them.
1. Shopify agentic commerce: AI has become a buying channel, and Shopify is building the plumbing
The headline pairing of the week was the Universal Commerce Protocol and the Catalog API, and they only really make sense together.
UCP is the open standard Shopify has co-developed with Google so that AI agents can complete a purchase on a shopper’s behalf. The Catalog API is the other half of that picture, making merchant catalogues available to those agents in a structured, machine-readable form. It spans every merchant on Shopify and has been two years in the making, which tells you this has been coming for a while.
The numbers they shared are what explain the urgency. AI-driven sessions are converting around 80% better than organic search, and roughly half of them land straight on a product page. That’s a shopper turning up with their mind largely made up, having done all their comparison somewhere you can’t see and can’t influence.
Two new Catalog capabilities landed at the event. The first is personalisation, which re-ranks results using Shop identity data, so an agent sees a version of your catalogue shaped by whoever is asking rather than one generic list for everybody. The second is commission, which means app developers now earn on the conversions they drive. Read together, they tell you Shopify would rather seed an ecosystem than try to build every one of these surfaces itself.
Shopify Functions will support agentic checkouts in the second half of 2026, and that’s the point where discount logic, bundling and custom checkout rules start behaving properly inside an agent-led purchase.
Why it matters
Product discovery is moving off the search results page and into a conversation, and in that conversation your product data is the thing being read. Thin descriptions, missing attributes and inconsistent variant naming used to cost you a little SEO. Now they determine whether an agent can represent your product at all.
What to do next
Go and audit your product data as though a machine were the customer, because increasingly it is. Look at completeness, attribute consistency, structured specs and whether availability is accurate. Then work out who owns agentic channels internally, because right now that responsibility tends to fall somewhere between ecommerce and marketing and nobody is quite accountable for it. If you want the wider view, we’ve written about what agentic AI means for Australian businesses and how to get referenced by LLMs rather than just ranked.
2. Shopify POS: physical retail gets more valuable as AI takes over convenience
If AI is going to own convenience, meaning discovery, comparison and the transaction itself, then the store becomes the one place where a brand still controls the entire experience. On that reading, physical retail gets more strategically valuable at exactly the moment everyone assumes it’s being commoditised.
Shopify has clearly invested behind that view. POS has matured a long way, the hardware is better than it was, and POS UI extensions have quietly become a serious build surface. Three merchant examples stood out:
- Tacobas, a 60-store cowboy boot retailer, built a “Boot Runner” extension so front-of-house staff can request stock from the back room. Retrieval time dropped from around 10 minutes to under two, and the radio chaos went with it.
- A Luka Doncic pop-up used a mobile fan pass QR code that created a customer record in Shopify automatically at check-in, so every person who walked through the door became a known customer without anyone touching a clipboard.
- Bandit Running in New York sells membership as a community, and the pass unlocks a members-only drink fridge with no purchase required, which is a lovely bit of retail theatre.
The build effort for something like that pop-up extension was described as a couple of weeks, and that’s a small enough number to change how you think about in-store experiments altogether.
Underneath all the front-of-house work, inventory is being rebuilt properly. There’s now a double-sided ledger tracking every stock movement, and “bins” are coming, which will let you model front-of-house and back-of-house stock separately. If you’ve ever had a store showing available stock that nobody can physically lay hands on, that’s the fix.
Why it matters
Most Australian retailers still run store and online as two separate operations with a reporting layer bolted across the top, and the distance between what the platform can now do and what most merchants have actually configured is wide.
What to do next
Pick one in-store friction point that you know is costing you conversion, and prototype a POS extension against it. Two weeks of build to prove a number is a far easier business case to get across the line than a full retail transformation programme. Worth reading alongside this: the cost of disconnected systems and manual processes.
3. Shopify B2B is the biggest gap between what the platform does and what people think it does
Shopify put the global B2B commerce opportunity at $28 trillion for 2026, and pointed out how much of that still runs on phone calls, spreadsheets and emailed purchase orders. Their own B2B side is growing over 100% year on year, with Collective growing faster again.
The more interesting part, at least for us, is the perception gap. When an Australian trade merchant sits down to shortlist a B2B platform, Shopify usually doesn’t make the list, because the reputation is still DTC. That reputation is roughly three years out of date, and the distance between what’s on the platform now and what buyers assume is on it was probably the most useful thing we brought home.
Collective is the clearest example of that. In plain terms, it lets one Shopify store sell another Shopify store’s products without holding any stock. Say you run a trade retailer and there’s a power tool brand you’d like to carry, but you don’t want to buy their range upfront and sit on it. You connect to them inside Shopify, their products appear in your store at your pricing, a customer buys one from you, the order passes to the brand automatically, the brand ships it directly to your customer, and Shopify settles the money between the two businesses. No purchase order, no invoice, no spreadsheet, and nobody rekeying an order at either end. In a normal supplier relationship you’d agree commercial terms in a few weeks and then lose six months to integration because the two systems have never met each other, and Collective simply deletes that middle part.
The example they walked through was Mattel and Funboy, the Barbie movie collaboration, where products synced in a day and full testing was done inside two weeks. 57% of buyers were new customers to Mattel, so the wholesale relationship doubled as a customer acquisition channel. And Funboy learned from the data that they’d been underpricing, which is the sort of thing you only ever discover when wholesale runs through a system rather than a rep’s inbox.
Collective went live in Australia this year, so both stores need to be AU-based, on the same currency and running Shopify Payments. It’s worth knowing where the edges still are, and Shopify was refreshingly upfront about them. Cross-border remains blocked on payment dependencies, B2B subscriptions haven’t landed yet, and connecting a partner who isn’t already on Shopify involves a fairly clunky workaround.
Why it matters
For a lot of Australian trade merchants, the platform question got settled years ago on information that no longer holds.
What to do next
Map your wholesale workflow end to end and mark every step where someone rekeys data that already exists in another system, because that list is your business case and it’s usually longer than anyone expects. Then check whether your suppliers or your stockists are already on Shopify, because if they are, the integration project you’ve been dreading disappears.
4. Shopify River: the platform under the AI matters more than the AI
The most useful session of the week had nothing to do with commerce at all.
Javier Moreno, a principal engineer at Shopify, walked through River, their internal AI engineering agent. River lives in Slack, and engineers mention it the way they’d mention a colleague. It investigates bugs, searches documentation, writes code, opens pull requests, runs tests and helps debug production issues, and it now co-authors roughly one in eight merged pull requests across Shopify’s codebases. Tobi confirmed 100% of staff are active on it, with more than 7,000 weekly and 6,500 daily users.
Javier’s real point was about what made River possible in the first place, and it’s the bit most companies skip. The productivity multiplier is the engineering platform underneath. AI is simply what finally made that platform legible.
River works because Shopify had already done years of unglamorous foundational work: consolidating into a single monorepo, reproducible development environments with Nix, consistent CI/CD, shared tooling, standardised workflows and documentation an agent can read. Give an AI agent access to the whole codebase, production logs, testing infrastructure and accumulated team knowledge and it starts behaving like a colleague. Without that foundation underneath, you’ve got a chatbot.
The second design decision was making River public by default, so it only operates in public channels. Every session becomes shared learning, solutions stay searchable, anyone can jump into a thread, and new engineers pick things up from discussions they were never part of. The AI ends up compounding as organisational knowledge rather than sitting in someone’s private window.
Javier also talked about building a reusable substrate, where one platform exposes the common capabilities and specialised agents are built on top as profiles rather than as separate systems.
Why it matters
This is the clearest explanation we’ve seen of why so many enterprise AI programmes stall. Companies buy the agent and skip the substrate. We’ve argued something similar in AI-native vs bolt-on and the AI gap no one wants to admit.
What to do next
Before you scope another AI pilot, have a look at what an agent would find if you pointed it at your environment today. Can it reach the codebase, the logs, the docs, the tests? If the honest answer is no, then that’s your actual project. And wherever you can, make the AI work visible by default, because the compounding value lives in the shared context.
5. Building on Shopify keeps getting cheaper, which changes what you buy and what you build
If you strung the technical sessions together, they all pointed at the same outcome. Custom work on Shopify costs less than it used to, and the gap between wanting something and having it built keeps narrowing.
Some of that is better tooling for the people doing the building. Theme Check has been overhauled with type-safe blocks and a strict Liquid parser, so mistakes surface immediately instead of failing quietly. Hydrogen has been rebuilt on a three-layer architecture, and Partials give you fine-grained reactivity in a few lines rather than a rewrite.
Some of it is Shopify making the platform easier for AI coding agents to work on, which is the same thing viewed from a different angle. Themes now support AGENTS.md, DESIGN.md and agent skills files, so you can point an agent at your theme with proper guidance instead of hoping it works out the conventions on its own. A well-briefed agent working in a codebase built for it is a lot of the reason these builds are getting quicker.
Then there’s a handful of platform changes worth knowing about regardless. The Collections overhaul is the biggest in a decade, bringing multiple data sources, cross-collection references and exclusion controls. A new Events API is replacing webhooks, so you subscribe only to what you need and define the shape of the payload. And customer accounts have gone passwordless by default, with sessions stretched from 24 hours to 365 days, which matters more than it sounds when around 25% of customers sign in before they’ve ever placed an order.
Why it matters
Things that comfortably justified buying an app two years ago are now a native build measured in days, and every app you keep is a recurring cost, a performance hit and a dependency you don’t control.
What to do next
Take a proper look at your app stack and work out which of those subscriptions are now doing a job the platform does natively. Worth noting too that Shopify Scripts have been retired in favour of Functions, so if you haven’t migrated yet, start there.
The bigger picture from the keynote stage
The fireside with Tobias Lutke and Harley Finkelstein was the highlight of the week, and the argument Tobi made was a hopeful one. AI is about to cause an explosion of creativity on the web, because it hands the ability to build to an enormous number of people who never had it.
AI changes what people can create, similar to the early web. Things that used to be impossible are now becoming possible, which should lead to more creative software and experiences. – Tobi Lutke, Shopify
He framed it as three shifts happening at once. Previously impossible things are now merely hard, previously hard things are now easy, and previously expensive things are now cheap. His example was Pets.com, an idea that was directionally right and simply arrived too early, and which eventually came back as Chewy.
Vanessa Lee gave what was probably the sharpest framing of the week on what holds its value as code becomes commoditised. Three things endure. Your data, meaning the unique datasets that drive your sales. Your taste, meaning cultural insight and editorial judgement. And your distribution, meaning trusted relationships with a specific customer base. Everything else gets cheaper.
Wrapping up
Everything above is already shaping the advice we’re giving clients and the work we’re scoping for the year ahead, and being in the room in Toronto is a good part of how we get there, because the roadmap gets explained months before any of it turns up in a merchant’s store.
Restive is a Shopify Plus Partner, and we plan, build and run commerce for Australian retailers and brands. That covers replatforms and custom storefronts, POS and in-store experience, B2B and wholesale, getting ready for agentic commerce, and the integration work that connects Shopify to the systems already running the business. Our teams embed alongside yours, which keeps the strategy and the build in the same room.
Shopify reframed its mission at the event as increasing the surface area of entrepreneurship, and what we saw across the week was a platform being deliberately opened up so that smaller teams can do things that used to need a large organisation behind them. The merchants who move early on product data, POS and B2B will have a year’s head start on everyone waiting for it all to settle down.
If you’d like to talk through what any of this means for your roadmap over the next 12 months, we’ll give you a straight read on where the opportunity sits for your business. Talk to us.
Frequently asked questions
What is Shopify DotDev?
Shopify DotDev is Shopify’s annual developer and partner event, held in Toronto. It is where Shopify’s Editions release features are detailed and where the product and engineering teams present the platform roadmap directly.
What is the Universal Commerce Protocol (UCP)?
UCP is an open standard co-developed by Shopify and Google that allows AI agents to complete purchases on a shopper’s behalf. Paired with the Catalog API, it lets merchant catalogues be read and transacted against by AI assistants.
What is the Shopify Catalog API?
The Catalog API makes merchant product catalogues available to AI agents in a structured format. It covers every merchant on Shopify and supports personalised re-ranking of results using Shop identity data.
Does AI-driven traffic convert better than search?
According to figures Shopify shared at DotDev 2026, AI-driven sessions convert around 80% better than organic search, and about half of them land directly on a product detail page.
What are Shopify POS UI extensions?
POS UI extensions let merchants build custom functionality directly into the Shopify POS interface, such as stock retrieval requests for staff or check-in flows that create customer records automatically. Merchant examples shared at DotDev 2026 described build efforts of around two weeks.
What should Australian retailers do to prepare for agentic commerce?
Start with product data quality, since AI agents rely on structured, complete and consistent product information. Then assign clear internal ownership of agentic channels, and review whether your checkout and discount logic will work in agent-led purchases.
Is Shopify good for B2B and wholesale?
Shopify’s B2B capability has expanded significantly, covering company accounts, per-company catalogues and price lists, quantity rules, net payment terms, and prebuilt ERP connectors for systems including NetSuite and Brightpearl. Growth on the B2B side is running at over 100% year on year. Known gaps as of DotDev 2026 include B2B subscriptions and selling plans.
What is Shopify Collective and is it available in Australia?
Shopify Collective lets one Shopify store sell another Shopify store’s products without holding inventory. The supplier’s products appear in the retailer’s store, orders pass to the supplier automatically, the supplier ships direct to the customer, and Shopify settles payment between the two businesses. There are no purchase orders, invoices or manual order entry. It is available in Australia, where both the retailer and the supplier must be Australian-based, use the same currency, and have Shopify Payments activated. Cross-border Collective is not yet available.
What is Shopify River?
River is Shopify’s internal AI engineering agent, which operates inside Slack in public channels only. It co-authors approximately one in eight merged pull requests across Shopify’s codebases and is used daily by more than 6,500 staff.
Related articles
Real outcomes, for real Australian organisations.
How we built AI-powered personalisation and retail media for one of Australia’s largest supermarket chains.