Every ecommerce platform has an AI story now. Most of them are fiction.
Open any vendor’s homepage and you’ll see it: “AI-powered search.” “AI-driven recommendations.” “Intelligent automation.” The language is everywhere. The substance, in most cases, is not.
Here’s what’s actually happening: the majority of ecommerce platforms are bolting on third-party AI tools, wrapping them in their own branding, and calling it innovation. The merchant gets a chatbot that hallucinates product specs. A “smart” search bar that doesn’t understand the difference between a 3/8” brass fitting and a 3/4” steel one. Product descriptions generated by a model that has never seen their catalog and doesn’t know their customer.
And for the merchants running complex operations—B2B distributors managing tens of thousands of SKUs, manufacturers with configurable products, specialty retailers with nuanced compliance requirements—this kind of sloppy AI implementation isn’t just unhelpful. It’s dangerous.
The Real Problem: AI Without Context Is Just Noise
The fundamental issue with most AI in ecommerce today is context—or more precisely, the complete absence of it.
Large language models are extraordinary general-purpose tools. But general-purpose is exactly the wrong thing when a buyer needs to find the right replacement gasket for a specific compressor model, or when a merchant needs to understand which product categories are eroding margin this quarter versus last.
Generic AI doesn’t know your catalog taxonomy. It doesn’t understand that your business sells to both contractors and homeowners at different price tiers with different fulfillment logic. It hasn’t internalized your inventory constraints, your vendor relationships, or the seasonal patterns that drive your purchasing decisions.
Without that context, AI outputs aren’t insights. They’re guesses dressed up in confidence.
The industry has a term for this now: hallucination. And most platforms are treating it as an acceptable tradeoff. I don’t think merchants should accept that tradeoff at all.
The Data Privacy Question Nobody Wants to Answer
There’s another dimension to this that doesn’t get enough scrutiny: where does your data go?
When a platform integrates a third-party AI tool, your product data, your customer behavior, your pricing—all of it potentially flows through external systems. Some of these integrations are well-architected with proper data boundaries. Many are not. And most merchants have no visibility into the difference.
This matters enormously for B2B operators. If you’re a distributor with negotiated pricing, custom catalogs, and proprietary product data, you need to know—with certainty—that your competitive intelligence isn’t being fed into a model that also serves your competitors.
At Miva, we made a deliberate architectural decision: AI operates inside the merchant’s own environment, using their store-specific data, with clear boundaries around what’s processed and how. Not because it was the easiest path—it wasn’t—but because it’s the only approach that respects the trust merchants place in their platform.
Intelligence Should Amplify Operators, Not Replace Them
There’s a narrative in tech right now that AI is going to automate away entire job functions. In ecommerce, that narrative is mostly wrong—and pursuing it actively harms merchants.
The best merchants I know have deep intuition about their business. They understand their customers, their margins, their competitive dynamics. What they lack isn’t intelligence—it’s time and visibility. They’re buried in spreadsheets, toggling between dashboards, manually pulling reports to answer questions they already know how to ask.
That’s where AI creates real value. Not by replacing the operator’s judgment, but by compressing the time between question and answer.
When we built AI Insights into the Miva Admin, we designed it around a simple question: what would an experienced merchant want to know right now, and how fast can we surface it? The result is embedded intelligence that works inside the workflow—not a separate tool merchants have to context-switch into, not a chatbot they have to prompt-engineer, but contextual analysis that meets them where they already work.
Instant performance snapshots. Trend identification across product categories. Margin analysis that connects revenue data to cost data in real time. The kind of operational clarity that used to require a dedicated analyst or a BI tool that nobody had time to configure.
Connected Commerce Requires Connected Intelligence
This is part of a larger thesis we’ve been building toward at Miva, and it goes beyond AI.
The future of commerce—especially for mid-market and enterprise merchants—is connected. Not connected in the way that term usually gets used (another app store integration, another API call to another point solution), but genuinely connected: a single system where product data, order management, customer intelligence, ERP integration, and now AI all share the same context.
Most platforms are architecturally incapable of this. They were built as storefronts first, and everything else—B2B logic, ERP connectivity, margin visibility, multi-channel management—was bolted on after the fact. AI is just the latest thing getting bolted on.
When your platform is a loosely connected collection of apps and plugins, AI can only be as good as the data it can access, which usually means fragmented, inconsistent, and incomplete. You end up with an AI layer that sits on top of a broken data model and produces confident-sounding garbage.
Connected intelligence requires a connected platform. It requires data flowing natively between your storefront, your operations, and your analytics—not through middleware and webhooks and third-party data pipes. That’s the architecture we’ve built at Miva, and it’s why our AI capabilities can deliver genuine operational insight rather than surface-level summaries.
What Practical AI Actually Looks Like
I’ll be blunt: the ecommerce industry has an incentive problem when it comes to AI. Platforms are rewarded by investors and analysts for having an “AI story,” not for having AI that actually works. The result is a lot of announcements, a lot of demos, and very little measurable merchant impact.
Practical AI is boring by comparison. It doesn’t make for a great keynote. But it does make for better businesses.
Practical AI means a merchant can ask “what happened to my margins in Q4?” and get an answer in seconds, not days or weeks. It means product discovery that actually understands technical specifications and attribute relationships, not just keyword matching with a neural network bolted on top. It means operational intelligence that’s aware of your ERP data, your inventory levels, and your customer segments simultaneously—because it lives inside the same system.
That’s the standard we’re holding ourselves to. Not “how do we add AI to the feature list,” but “how do we make merchants measurably better at running their businesses.”
The Path Forward
AI in ecommerce is going to mature rapidly. The merchants who benefit most won’t be the ones who adopted it first—they’ll be the ones who adopted it intentionally.
That means choosing platforms where AI is built into the architecture, not layered on top. Where your data stays yours. Where intelligence amplifies your expertise rather than substituting a generic model for your hard-won knowledge.
The hype cycle will burn through. The tools that remain will be the ones that actually made merchants more profitable, more efficient, and more competitive.
That’s what we’re building toward. Not the most impressive demo—the most useful platform.
Rick Wilson