How Ecommerce Leaders Are Actually Using AI in 2026
How Ecommerce Leaders Are Actually Using AI (and What's Still Just Hype)
AI is in every ecommerce keynote, every vendor pitch, and every leadership offsite. But behind the noise, retail leaders are doing something more interesting than the discourse suggests: they're quietly running experiments, killing the ones that don't work, and building real workflows around the ones that do.
This piece draws on 100+ conversations with ecommerce executives recorded on Noibu's podcast, The Ecommerce Toolbox: Expert Perspectives, between 2024 and 2026 — covering brands including Hanna Andersson, Pendleton, Ribble Cycles, Bazaarvoice, Salesforce, Knix, Jordan Craig, Camper, Omoda, Baerskin, and many others. The patterns below are what surfaced across them.
The gap between AI keynote and AI reality
If you only read press releases, you'd think AI has already transformed ecommerce. Every platform vendor has launched an AI agent. Every analytics tool has added a chat interface. Every keynote leads with the same thesis: AI is changing everything.
The reality on the ground is messier — and more interesting.
What we heard, consistently: AI is quietly powering backend operations customers never see. Marketing analytics. Supply chain visibility. Developer productivity. Support automation. Internal data wrangling. This is where AI is delivering measurable ROI today, and it's where the most experienced leaders are placing their bets.
It's loudest in the customer-facing applications that get the press coverage — conversational search, AI stylists, agentic shopping — most of which are still pilot-stage at most retailers we spoke to. And it's almost entirely absent from the deepest layer of the stack: site operations and monitoring, where it could arguably have the most leverage.
Three patterns emerged about where AI sits in retail operations today. They're the spine of this piece.
What's actually shipping right now
The clearest AI wins in ecommerce in 2026 aren't the flashy ones. They're concentrated in three operational use cases.
Developer productivity and lean-team leverage
This is where the wins are loudest, most measurable, and least talked about in public.
Gus Fune at Baerskin Tactical rebuilt his company's entire ecommerce stack with a four-person team in three months — a project that would have required 15+ engineers two years ago. The acceleration came from AI-assisted development across architecture decisions, code generation, and debugging.
Rob Varon at Jordan Craig runs a nine-person digital team that operates at the scale of a 50-person operation. The team uses Relay.app as an agent orchestration layer connecting workflows across multiple AI models — Anthropic, ChatGPT, Gemini, Perplexity — to handle daily reporting, sale launch monitoring, influencer identification, and wholesale prospecting. They've automated more than 50 distinct workflows.
The pattern across these cases: AI isn't replacing engineers. It's letting small teams ship like large ones. The brands compounding this advantage in 2026 are mostly the ones that started in 2023 and 2024.
Personalization at scale — when done with restraint
The category that gets the most press coverage is also the one with the clearest wins for the brands that have done it well.
Ewoud Frielink at Omoda built an AI-powered stylist on top of large language models that gives customers natural, contextual style advice. Roughly 80% of users rate the stylist's suggestions as helpful — a number that compares favorably to the click-through rates of traditional recommendation engines. More importantly, the AI stylist surfaces information about customer intent that traditional analytics tools never see, because customers volunteer context (occasion, preference, mood) they would never enter into a search bar.
Frielink calls the model "lean-back ecommerce" — proactive curation that anticipates intent based on purchase history, wishlist signals, and browsing behaviour. It isn't reactive personalization. It's brand-led curation at scale.
The pattern across personalization wins: leaders treating AI as an extension of brand experience, not a replacement for human judgment. The brands that have failed — and there are many — typically rolled out generic recommendation engines without thinking carefully about what their brand voice required. Or, worse, scaled personalization on top of fragmented customer data and ended up recommending the wrong things to the wrong people, just faster.
Backend analytics and decision support
The least visible AI use case is arguably the most economically valuable.
Sienne Veit at Invisible Stuff described using AI to accelerate marketing analytics and supply chain visibility — work that would have required dedicated analyst hires in the previous era. The output isn't customer-facing; it's better, faster, more confident internal decisions.
Ron Ijack at Knix put the operating principle most clearly: start with outcomes, not tech. The framing matters because it changes the question from "what can AI do?" (which leads to demos and pilots) to "what specific decision could I make faster with better data?" (which leads to shipped workflows).
"Start with outcomes, not tech."
— Ron Ijack, Knix (The Ecommerce Toolbox podcast)
What's still in pilot: the customer service question
If there's one AI category with genuinely divergent outcomes, it's customer service.
Jenna Flateman Posner, Founder of Chief Digital Agency, made a strong case for conversational AI in customer service — particularly for replicating the personalized service of in-store shopping in a digital environment. Her thesis: AI can do the parts of customer service that scale, freeing human agents to focus on the parts that require empathy.
Jared Kugel at Tire Agent argued the opposite case in an episode pointedly titled "Why human connection beats AI." His thesis: in categories where customers are spending real money on complex decisions, the AI hand-off is the moment trust breaks. Tire Agent has explicitly chosen not to replace agents with bots — and the company's customer satisfaction numbers support the bet.
Both leaders are right — for their categories. The honest read across the conversations: customer service AI works when the questions are simple, the brand voice is informational, and the stakes are low. It fails when any of those conditions break down.
This is the use case with the widest gap between "what every vendor is selling" and "what actually ships value." Worth approaching with a clear-eyed view of which side of the line your category sits on.
The unspoken blocker isn't tooling. It's data.
The single most consistent finding across every conversation: every leader who described a paused or killed AI initiative came back to the same root cause.
Mena Wouters at Bazaarvoice put it most directly in her episode on the podcast: AI can be fantastic for scaling content, but it will scale things ineffectively if the underlying system isn't in a good spot. Her phrase for it — "scaling chaos" — captures something important. AI doesn't fix broken systems. It accelerates them.
"AI is going to scale things ineffectively if your system is not in a good spot. Your foundations need to be clean. Is your inventory in a good spot? Is your availability in a good spot? If not, you're going to scale in the wrong way."
— Mena Wouters, Bazaarvoice (The Ecommerce Toolbox podcast)
If product data is inconsistent, AI surfaces that inconsistency at higher volumes. If the inventory feed is unreliable, AI-driven recommendations promote unavailable products. If customer data is fragmented, personalization sends the wrong content to the wrong people — just faster.
Matt Ezyk at Hanna Andersson echoed this from the engineering side. Clean, connected customer data is the non-negotiable foundation for any meaningful AI adoption. Heritage brands modernizing their stacks have a particular version of this problem — decades of legacy systems, all encoding business logic in ways AI can't safely interpret without significant cleanup work first.
Rick Ragusa at Pendleton framed it as the precondition for agentic commerce: without discipline around the data, you can't drive the agentive experience. The leaders who got AI working invested first in data discipline. The leaders who skipped that step are now paying for it.
The agentic commerce question
Of all the topics in AI today, agentic commerce has the widest gap between discussion and adoption. Every leader we spoke to is talking about it. Almost none have done anything about it yet. The small group who have started are moving with more caution than the discourse suggests is warranted.
Rick Ragusa at Pendleton framed it as a platform shift, not a channel shift: instead of optimizing for screens, brands now have to optimize for intent, personalization, and conversational interaction. He believes the shift will be fast — comparable to how quickly mobile reset customer expectations a decade ago — and that it will bring power back to brands willing to embed AI directly into their own sites rather than rely solely on third-party intermediation.
Nitin Mangtani at Salesforce, who runs Commerce Cloud and Retail Cloud, was more measured. Agentic AI is going to change how consumers shop, how employees work, and how commerce systems operate — but it isn't replacing the fundamentals of retail. The leaders winning will be the ones who treat agentic commerce as a layer on top of solid retail discipline, not a substitute for it.
Ram Venkataraman at Kibo Commerce echoed the framing: the brands that win agentic commerce won't be the ones with the most AI features. They'll be the ones with the cleanest data and the most disciplined architecture underneath.
What leaders are actually doing about it falls into three patterns:
- Cleaning the data foundation — product feeds, inventory accuracy, customer data unification, schema markup
- Auditing site infrastructure — making sure pages render reliably, events fire predictably, and the technical underpinnings could support automated traffic
- Running small experiments — usually one or two test cases with AI shopping assistants on their own sites, to see what actually works
What they're not doing: making big bets on specific agentic commerce platforms. The reason, consistently, is that nobody knows yet which platforms will matter most in 18 months — ChatGPT Shopping, Perplexity, Amazon Rufus, Google's agentic search, brand-controlled assistants, or something else entirely.
The most useful reframe came from one Bazaarvoice executive: agentic commerce isn't a thing you build a strategy for. It's a state of operational readiness you build toward. If your data, your inventory, your site, and your team are healthy enough to support automated traffic, you'll be ready for whatever agentic commerce becomes. If they're not, no strategy will save you.
The five habits of AI-mature ecommerce teams
The most consequential pattern across the conversations is that AI adoption has bifurcated into two groups — and the gap between them is widening every quarter.
The AI-mature teams share a recognizable profile. So do the ones falling behind.
The teams falling behind share the opposite profile: they start with technology purchases and look for use cases. They have fragmented data they keep meaning to fix. They have one AI champion, often siloed in marketing or product. They overstate their wins and understate their failures.
The gap between these two profiles will compound over the next 18 months. The leading group will get further ahead. The lagging group will spend more on tools while shipping less value.
What this means for the next 12 months
If the past 12 months were about experimenting with AI, the next 12 will be about operationalizing it. Three patterns to track:
The data layer becomes the moat. As AI tools commoditize, the differentiator becomes the quality of the underlying data. Brands with clean, unified, real-time data will be able to apply any AI capability that arrives next. Brands without it will be locked out.
Agentic commerce arrives faster than expected — but more narrowly. The conventional wisdom is that agentic commerce will be massive and slow. The pattern from these conversations suggests the opposite: it will arrive in specific categories first (high-consideration purchases, B2B, complex configurations) and won't generalize for several years. Brands in those leading categories should be preparing now.
The cost dynamics will reshape the conversation. As AI platforms become transaction layers — taking referral fees, conversion fees, or both — the margin math will get hard. Retailers who don't model the per-transaction economics today will get surprised in 2027.
The practical takeaway from every leader who's shipping real AI wins: audit the data foundation. Start with the unsexy use cases. Make AI literacy a team expectation, not a specialist role. Run multiple pilots in parallel and kill the ones that don't ship value in 90 days. And invest in data discipline every quarter — because the gap between AI-mature teams and the rest is already compounding.
Start with the foundation, not the feature
The most consistent message from the leaders we spoke to: AI is a multiplier, not a magic wand. It accelerates whatever's underneath it — clean data, good systems, disciplined teams — or it accelerates the chaos.
If you can't tell whether your data foundation is ready for AI to run on top of it, a free Noibu website audit is a fast way to find out. It surfaces the hidden friction, technical issues, and data gaps that quietly erode every AI initiative running above them — the data your team would need clean before agentic commerce is even a conversation.
Related topics
- The ecommerce site problem nobody talks about (because your tools don't show it)
- Ecommerce Product Analytics: How to Build Roadmaps That Drive Conversion
- 5 ecommerce monitoring metrics every Head of Ecommerce should track
- Listen: The Ecommerce Toolbox podcast
About Noibu
Noibu is the leading ecommerce analytics & monitoring platform, purpose-built to help retailers protect and grow online revenue. By unifying site monitoring, experience analytics, and conversion growth opportunities in a single pane of glass, Noibu captures the most important end-to-end shopping data, without the complexity of traditional analytics tools.
Noibu surfaces critical site errors, performance issues, and customer journey friction that block conversions, then ties every insight directly to business impact, session replays, and full technical context. This makes it easy for ecommerce teams to understand why things are happening and what to prioritize, without dedicated analytics headcount.
The result: faster decisions, better collaboration across teams, optimized customer experiences, and revenue growth.
