The Retail Hive’s Head of Content, Ed Lawson, interviews Peyman Naeini, Field CTO at Shopify
Customers are moving from using AI assistants are moving from as discovery tools to active shopping companions, helping them research, compare and (increasingly) transact. Although the trend is moving quicker in the Far East and the USA, it’s beginning to snowball in Europe as well. Shopify’s Field CTO Peyman Naeini explains why the next wave of retail won’t be won through clever marketing alone, but through clean product data, trusted information and frictionless customer experiences, and how you can lay the foundations for success right now.
Key takeaways:
Agentic commerce is shifting discovery from search to conversation
Customers are no longer just searching for products, they’re describing needs, preferences, budgets and use cases in natural language.
As AI assistants become more involved in discovery and decision-making, you need to think beyond keywords and optimise for intent-rich conversations.
Product data is becoming a competitive advantage
Throughout our interview, one message comes through clearly: AI can only recommend what it can understand. Structured product data, accurate inventory, reliable sizing information and clear product attributes are becoming foundational to visibility. If you invest in these basics now, you are laying the groundwork for future discoverability.
Trust will determine
adoption
While AI agents are likely to play an increasing role in routine purchases and product recommendations, customers still want transparency, control and confidence. To win, you need to combine AI-ready infrastructure with trustworthy information, strong operational performance and clear brand guardrails.
Q: Agentic commerce is being positioned as a major shift from search to action. In practical terms, what does that actually mean for brands and retailers today?
Peyman Naeini, Shopify: At its simplest, agentic commerce is about turning intent into action. Rather than a customer manually searching, filtering, comparing and checking out across multiple tabs, they can increasingly describe what they want in natural language and have an AI assistant do a lot of the heavy lifting for them.
So instead of searching for “black dress”, someone might say: “I need a black midi dress for a wedding in May, size 10, under £150, ideally in natural fibres.” An AI agent can then shortlist products, compare options, check stock availability, surface sizing information, apply discounts and even move the customer through to checkout.
What’s interesting is that this goes well beyond discovery. We’re also seeing the early stages of agents helping with post-purchase experiences like tracking, returns, care instructions and even reordering essentials.
The impact won’t happen evenly across every category or customer journey. Over the coming months and years, I think we’ll see much stronger conversational discovery and assisted checkout experiences. Further ahead, more routine purchases could become increasingly delegated to AI assistants, while more expressive or emotional purchases will still remain very human-led. People will still want to make the final call on things tied to identity, taste or aspiration. The role of the agent is to remove friction, not replace personal judgement.
Q: Do you think shoppers will genuinely trust AI agents to make purchasing decisions for them? Including handling payments?
PN: Trust grows by task and by doing.
I think many customers will become comfortable quite quickly with AI helping to shortlist products, monitor prices, reorder repeat purchases or narrow down options. We already rely on algorithms in many parts of our lives. Commerce is really the next stage of that evolution.
Where things become more nuanced is around higher-consideration or emotionally driven purchases. In those cases, people still want control. The AI becomes more of an advisor than a decision-maker.
The biggest drivers of trust are transparency and reliability. Customers need to understand why something is being recommended. They need confidence that product information is accurate. And they need clear controls around budgets, preferred brands, sustainability preferences and purchase approvals.
On payments specifically, trust tends to follow familiarity. If customers are using secure, well-known payment rails and trusted wallets like Shop Pay, they’re far more comfortable authorising purchases or setting spending limits for repeat actions.
Q: We’ve spent years optimising around keyword search. What changes when shopping becomes more conversational?
PN: Discovery starts to become much more contextual. Customers won’t just search for products, they’ll describe situations, preferences, budgets, lifestyles and constraints.
That changes the way brands need to think about visibility. Historically, the goal was ranking well on a search engine or marketplace. Increasingly, discovery will happen through AI assistants, messaging platforms, voice interfaces and embedded commerce experiences.
In that world, you win placement by being easiest for machines to understand and trust. That means clean product data, accurate availability, reliable fulfilment information and structured content become incredibly important.
Keep three promises: Be findable through clean structured data. Be factual with accurate and trustworthy information. And be fast when it comes to checkout and fulfilment.”
Q: Shopify has been talking about agentic storefronts and enabling products to be discovered and purchased within AI tools. Can you explain what that looks like in practice?
PN: The idea is to make products and storefronts legible to AI systems in a structured and permissioned way.
For merchants, that means making catalogues, pricing, inventory and product metadata accessible so AI assistants can accurately surface products and help customers transact. We’re also making it easier for customers to move seamlessly from an AI-powered recommendation into a product detail page or directly into an accelerated checkout experience like Shop Pay.
What’s important is that merchants still maintain control. Brands decide what information is shared, where it’s surfaced and under what conditions. Attribution matters too. Retailers need visibility into where demand is coming from and which assistant experiences are actually driving value.
I’d encourage retailers to think about AI assistants almost as a new high-intent acquisition channel. Similar to marketplaces in some ways, but potentially with much deeper personalisation.
The retailers who operationalise those basics well are going to be in a very strong position.
Q: What are the first practical steps retailers should be taking now if they want to become “agentic AI-ready”? And how important is discoverability in all of this?
PN: Retailers should start by getting the fundamentals right. That means investing in clean, accurate product data, strong variant-level imagery, clear sizing and fit information, reliable stock visibility and structured product feeds that stay up to date. Operational basics matter too. Accurate pricing, realistic delivery promises and a frictionless checkout experience all become increasingly important when AI systems are surfacing products and helping customers make decisions.
Technical foundations matter too. Structured schema, clean taxonomy and fast checkout experiences become increasingly important in a world where AI systems are interpreting and surfacing products on behalf of customers.
Discoverability is absolutely critical. AI assistants will prioritise sources they can easily parse and trust. In many cases, clean structured data and operational reliability will matter more than clever marketing copy.
That’s a shift some brands will find uncomfortable because historically a lot of differentiation sat in storytelling and aesthetics. Those things still matter enormously, but the underlying data layer is now becoming commercially strategic.
Q: Structured product data keeps coming up in these conversations. Where are brands getting it wrong today?
PN: AI systems reason through attributes and structured information rather than vague descriptions. If data is incomplete or messy, recommendations will become weaker very quickly.
One of the biggest issues is inconsistency. We still see product titles packed with marketing language rather than useful detail, missing attributes around things like fabric, fit or silhouette, poorly maintained size guides, limited variant imagery, inaccurate availability information and unclear returns policies. Individually those things might seem minor, but together they create friction for both customers and AI systems trying to interpret products accurately.
These gaps create friction for both customers and AI systems.
Retailers who invest in high-quality product data now are effectively laying the foundations for future discoverability.
Q: Is this the death of SEO as we know it?
PN: I don’t think SEO disappears. It evolves.
The emphasis shifts away from simply targeting keywords towards building authoritative, structured and genuinely useful content.
That means:
- Structured data and feeds become even more important
- Content needs to answer real customer intent
- Product and editorial content should be machine-readable
- Site architecture and taxonomy matter
- Fast, reliable experiences still matter enormously
In reality, brands will probably optimise simultaneously for both traditional search and AI-assisted discovery for quite some time. There’s a lot of overlap between the two.
Q: If transactions increasingly happen within AI interfaces, what happens to brand storytelling and brand.com experiences?
PN: I actually think brand.com becomes more important, not less.
The transactional layer may become more distributed, but the brand site increasingly becomes the canonical source of truth. It’s where retailers can fully express their identity, build community, showcase editorial content and deepen customer relationships.
The experience itself may evolve though. Instead of endless product grid browsing, we’ll likely see more guided and conversational experiences:
- “Help me build a holiday wardrobe”
- “Find products that work with this jacket”
- “Reorder my skincare routine”
- “Create a capsule wardrobe for work travel”
The underlying merchandising model becomes more dynamic and intent-driven.
At the same time, brands still need to maintain their distinct voice and point of view. AI can help scale discovery, but customers still connect emotionally with taste, creativity and human curation.
That’s why things like styling rules, creator content, lookbooks, memberships and post-purchase experiences become even more valuable.
Q: What risks should retailers be paying attention to as this evolves?
PN: There are a few big ones.
Bad data is probably the most immediate risk. If inventory, pricing or product attributes are inaccurate, AI systems will surface poor recommendations and trust erodes quickly.
Retailers also need clear rules around substitutions, positioning and how products are represented across external assistant ecosystems.
Hallucinations remain an industry challenge too. The best approach is grounding AI systems in structured, up-to-date product information rather than relying purely on generative responses.
Privacy and compliance obviously remain critical as well. Brands need clear governance around consent, data sharing and regional regulations.
And finally, there’s a strategic risk around dependency. Retailers shouldn’t build their entire future around a single assistant or platform. Diversification and strong first-party relationships will still matter enormously.
Q: Finally, what should retail leaders prioritise over the next 12 months if they want to stay competitive?
PN: I’d focus on getting the foundations right first.
That means:
- Improving product data quality
- Fixing size and fit information
- Strengthening variant imagery
- Stabilising live inventory accuracy
- Streamlining checkout
- Reducing fulfilment friction
- Improving returns experiences
At the same time, retailers should start experimenting responsibly. Pilot conversational discovery tools. Test a handful of assistant channels. Build internal understanding around attribution and governance.
Importantly, this also requires organisational readiness. Teams need clear policies around human oversight, escalation paths and brand guardrails.
Q: Anything else you’d leave retailers with?
PN: Keep three promises:
Be findable through clean structured data.
Be factual with accurate and trustworthy information.
And be fast when it comes to checkout and fulfilment.
Do that, and agents will bring you customers. Fail at any one, and they won’t.
