The Shero Commerce team attended Shopify's DotDev conference a couple of weeks ago, and the loudest headline coming out of it was, no surprise, AI. Agents, agentic commerce, and AI-native tooling. But the more we compared notes internally, the more a second, quieter story emerged, one that Shopify's own leadership kept returning to across multiple keynotes: getting the fundamentals right.
Tobi Lütke, Shopify’s CEO said it plainly in his fireside chat, framing the year's work as "going back to basics." Vanessa Lee, Shopify’s VP of Product walked through the biggest rewrite to collections and inventory in a decade. The AI headline, it turns out, is really the subplot this year.
Here is what stood out to us, organized by theme, and what each shift actually means for your eCommerce store.
Back to basics: the real headline
Vanessa's session covered the quality-of-life fixes that merchants and developers have been asking for for years. Collections management got what she called "our biggest update to collections in the last decade": variant-level collections (finally, all your red t-shirts can live in one place), visual merchandising, and nesting. On the back office side, inventory is getting a genuine overhaul, starting with a double-sided ledger where, as Vanessa put it, "every inventory movement in Shopify is now accounted for." Bin-level locations are coming next.
Tobi tied it together with a philosophy lesson borrowed from industrial design: Norman Doors, the classic example of a door whose design fights the person trying to use it. His point was that sweating small usability details is the whole job, and that "our software should never, ever make someone feel stupid." It is a fitting thesis for a company spending this much stage time on collections UI and inventory ledgers instead of flashier announcements.
For merchants, none of this requires you to do anything, but it will change how your store feels to run day to day. Variant-level collections mean you can finally group products the way your catalog actually looks, instead of working around single-variant limitations. A more accurate, ledger-based inventory system means fewer oversells and clearer visibility into stock that's damaged, on hold, or otherwise unaccounted for. None of this is a headline feature, but it is the kind of foundational fix that quietly makes a store easier to run.
Agentic commerce is real, but it's not what you think
The AI story at DotDev was less about chatbots replacing your website and more about a structural shift in where shopping actually happens. Shopify's Catalog API is now open to anyone with an API key, available as both a standard API and an MCP server, so AI assistants like ChatGPT, Perplexity, and Copilot can pull in live product data. Notably, Shopify said it is deliberately not building consumer-facing retail experiences on top of this catalog itself. It wants its ecosystem of apps, agencies, and merchants to build those experiences instead. Tobi's own demo was hooking Claude up to a Catalog skill and having it shop for him.
"Dress it up however you like, a search layer that pays out commission on conversion is still an echo of the old Affiliate Feed model, just with much better plumbing behind it."
Mike Warwick , Director of Solutions, Shero Commerce
The most concrete data point we heard: on stores already seeing AI agent traffic, more than half of agent sessions land directly on a product page, compared to roughly 20% for organic search. Agents are increasingly skipping the homepage entirely.
That has a few direct implications for your product pages:
-
Your PDP is now your front door for this channel. Brand storytelling, trust signals, and context that used to live on your homepage need to be present, or at least discoverable, at the product level.
-
Write for agents that can't fully "see." AI agents tend to skip or under-weight images because they are slower to process than text. If a product's key attributes (color, pattern, material) only exist in a photo, an agent may miss them entirely. Descriptive, specific copy matters more than ever.
-
Server-side rendering is worth auditing. Many AI crawlers don't reliably execute JavaScript, so reviews, FAQs, and structured data that only load client-side can be effectively invisible to them. A simple way to check: load your PDP with JavaScript disabled and see what's actually there.
-
"Customers currently trust AI for research and comparison, but not yet for handing over checkout, especially on third-party surfaces. Expect a transition period where product page plus self-checkout remains dominant, with retailer-owned agents closing that trust gap faster than third-party ones."
Daniel Bello , Head of Engineering, Shero Commerce
-
Community conversation increasingly feeds AI answers. One stat cited at the conference put Reddit's share of ChatGPT citations at roughly 13%. Brands can influence this over time by encouraging authentic conversation and leaning into a distinct point of view, rather than trying to manufacture it.
-
"Taste" is becoming its own data category. For low-differentiation products like electronics or supplements, factual specs dominate what agents cite. For taste-driven categories like fashion & beauty agents still need brand voice and emotional signal, not just a spec sheet. A phrase like Rhode's "glazed donut skin" isn't a fact, but it's memorable, gets repeated across community discussion, and feeds back into what agents end up citing. Shopify is extending its merchant knowledge base to capture more of this, alongside a new endpoint that would let brands supply direct answers to agents about their own product
The plumbing nobody notices, but everyone will feel
A few less flashy platform changes are likely to have an outsized practical impact.
Events API is a ground-up rewrite of how Shopify sends data to external systems, replacing classic webhooks rather than simply renaming them. Developers can now define exactly what data comes back in each event payload, meaning far less infrastructure spent just filtering out noise you didn't need.
The stable release lands October 6, 2026. If your store relies on custom integrations (order sync, inventory sync, marketing automation), it's worth confirming with your developer or agency that this migration is on their radar well ahead of that date, not something to discover in September.
Inventory is becoming a true ledger. Shopify is shifting its underlying model from tracking "quantity available to sell" toward tracking "quantity of goods physically held," with sellable quantity calculated from that. Stock now lives in "bins," containers that can be marked fulfillable or not, so damaged or quality-hold inventory becomes its own trackable bin rather than a vague adjustment. The practical upside for merchants is more accurate, auditable stock counts. The catch: integrations (particularly with ERPs like NetSuite or Odoo) that work by periodically overwriting a total quantity, rather than sending incremental adjustments, are more likely to create sync conflicts under this model and may need updating.
Split fulfillments are arriving before the 2026 holiday season. A single order can now be fulfilled in multiple ways at once, for example partly shipped and partly picked up in-store. If you use third-party logistics providers or offer more than one fulfillment method, it's worth confirming with your provider that they correctly evaluate each fulfillment on an order individually. A 3PL that only looks at an order as a whole risks over fulfilling it.
Our CEO, Beth Shero hosting the panel 'Shero Commerce on migrating from Adobe Magento to Shopify' along with Colin Sinclair and Hayward PeirceB2B merchants: real progress, with an important caveat
There was genuine, concrete movement on B2B pain points. Shopify now supports tying net payment terms to a specific calendar due date (rather than only "30 days from invoice") by vaulting a customer's payment method and using an automation to charge it on the due date, a workflow that works for both cards and ACH. There are also two viable approaches now for passing credit card processing fees on to B2B buyers as a surcharge.
The important caveat: Shopify's current AI and agentic shopping infrastructure does not yet support B2B. Wholesale pricing, gated catalogs, and customer-specific pricing are explicitly out of scope for agent-facing surfaces for now. If your business runs on tiered or negotiated pricing, "AI-ready" claims from any vendor should be evaluated with that gap in mind.
"UCP has no B2B support today. Anything wholesale, price-list driven, or with customer-specific pricing is out of scope for agentic surfaces for now, and that's worth stating plainly before you sell agent-readiness to any B2B merchant running tiered pricing."
Daniel Bello , Head of Engineering, Shero Commerce
Checkout extensions and agentic surfaces: a real gap to plan for
This might be the most immediately actionable finding from the whole conference. Shopify Functions (the newer, more flexible way to customize checkout logic) work fine within Shopify's new agentic checkout infrastructure. Checkout UI extensions, the older visual customizations many merchants rely on for things like age verification or terms-of-service acceptance, currently have no compatibility with it at all. If an AI agent tries to complete a purchase on a store where one of those extensions is flagged as business-critical, the transaction gets escalated back to the merchant's own checkout, pulling the customer out of the agent's flow entirely.
Shopify's stated direction is to turn that hard escalation into something lighter (essentially, "just fill in an extra field" rather than "start over on our site"), beginning with a release planned for August 2026 that also adds delegated identity, attribution tools, and additional custom fields, so merchants can keep visibility into sales that happen through agents. Either way, merchants keep merchant-of-record status, their order data, and their marketing rights. What changes is that the moment of interaction may no longer happen on the merchant's own interface.
If your store depends on checkout UI extensions for anything business-critical, it is worth auditing now which ones truly need to block checkout versus which are cosmetic, and revisiting that plan again once the August release lands.
Security is still mostly about hygiene, not sophistication
"Everything else in app security is a rounding error next to this. Threat actors targeting these apps are opportunistic, not sophisticated, and they monetize mainly through checkout redirects and ransom."
Daniel Bello , Head of Engineering, Shero Commerce
On the app security front, the takeaway was refreshingly unglamorous: the overwhelming majority of incidents trace back to basic credential hygiene rather than sophisticated attacks. Threat actors targeting eCommerce apps tend to be opportunistic, not sophisticated, and they monetize mainly through checkout redirects and ransom. Two specific, easy-to-miss risks came up: apps that keep calling back to a domain a partner let lapse (which an attacker can cheaply re-register and use to intercept requests), and apps requesting broader data permissions than they actually use, which makes them a more attractive target for no real benefit.
If you work with agencies or developers on custom apps, it's a fair question to ask whether they scan for accidentally committed credentials, rotate secrets on a regular schedule, and request only the data scopes an app actually needs.
Liquid is getting an AI-native rewrite
Shopify is reworking its Liquid templating language to be more legible to both people and AI systems. A new, lightweight base theme (currently called "skeleton") ships with 93% fewer lines of code than the current default theme. New template tags make sections of a theme more modular and give AI tools a more explicit, checkable contract to work against when making edits.
This isn't a hypothetical concern. Shopify shared that its Sidekick AI assistant has already been used for 25 million theme edits this year, and roughly one in five merchants are already using AI to edit their own theme. Whether your team edits themes directly or leans on an AI copilot to do it, this next release cycle should make that process noticeably more predictable.
The Shero Team at DotDev 2026What this means for your store
A line from one of the keynotes stuck with our team: code will likely hold less value going forward, and what endures is data, taste, and distribution. Data means the information unique to your business that helps you sell. Taste means your specific point of view within your category. Distribution means the relationships and trust you've built with your customer base.
That framing matters because it lines up with everything else we heard this year. AI agents are getting better at reading a product catalog, but they still need brand voice and community signal to make taste-driven recommendations. The platform is investing in ledger-accurate inventory and a rebuilt events system, but that's plumbing, not differentiation. The actual advantage available to any merchant right now is the same one that's always been available: a distinct brand, a well-run operation, and genuine trust with your customers. AI agents are simply becoming one more audience reading those signals.
Shero will be tracking how these changes land over the next few release cycles, from the Events API deadline in October to the checkout changes expected in August. If you want a read on how any of this applies to your eCommerce store, get in touch with us!
Written by Mike Warwick, Daniel Bello, Kevin Pina and Katie Evans from the Shero Team. -