Ready for AI shopping? Most retailers aren’t thinking far enough ahead
I’ve been deep in writing an ebook on AI in shopping. But this week’s update changed the pace.
OpenAI’s announcement of instant shopping inside ChatGPT triggered a wave of questions from retailers. Not hype. Not futurecasting. Real questions. Right now.
So instead of waiting for the full playbook, here’s what I’ve been telling people privately — typed up for anyone who wants to get ahead.
AI fundamentally can be thought about as just another channel. At the same time, long-term, it’s also a massive behaviour shift. That’s a tricky balance between the short and long-term.
And the best way to plan for that?
Ask yourself:
“If this was the only channel our customers used to shop — how would we still win?”
What would we double down on?
What would need fixing, fast?
What would make us proud to be chosen, even without ever meeting the customer?
That’s how to think clearly.
Because while your team debates 1% gains on an existing channel, this new channel is already open and you could be early into the mix with a breakthrough your competitors weren’t thinking about.
What you still control (and should start auditing now)
The funnel hasn’t disappeared. It’s just sAI breaks the funnel.
But it doesn’t take away your leverage.
Customers might skip your homepage, your filters, your entire experience. But they’ll still rely on your data. What shows up when you’re not in the room will matter more than what’s on your PDP.
You still control:
- Product data. Structured, complete, machine-readable.
- Imagery. Clear, varied, and relevant to use cases.
- Merchandising logic. How bundles, variants, and cross-sells are set up.
- Copy. Short, specific, and benefit-led.
- Delivery and returns. Especially if an AI assistant is comparing fulfilment logic on the customer’s behalf.
- Stock signals. Real-time, reliable, and visible.
Ask yourself:
Would we still convert if our experience was invisible, and all we had was data?
If not, fix that before anything else.
What to ask your payment gateway if you want to be AI-ready
To make this real, consider the variety oIf AI agents are going to start checking out on customers’ behalf, your gateway needs to be prepared. Most aren’t.
Here’s how to test if they’ve even started thinking about it:
- Can you support agent-led transactions without redirects or front-end input? If not, the agent can’t complete the checkout without human involvement.
- Do you offer tokenised payment methods that work across sessions and channels? Stripe’s Shared Payment Token API is a good signal of where this is going.
- Are you exploring programme-level approval rules or agent authentication? Delegated payments will need agent whitelisting or permissions — not just card storage.
- Can we speak to someone about your roadmap for delegated or AI-powered checkout? Even if they don’t have a feature yet, the conversation matters.
If they can’t answer any of these, you don’t need to switch providers. But you do need to start planning around that limitation.
Equally, if your agency can’t ask this on your behalf, have a word.
Marketplaces already behave like AI agents
Want a live preview of AI shopping? Look at how your products sell on Amazon. (Other marketplaces are available)
You don’t control the funnel. You don’t own the narrative. And yet people still buy. Or don’t.
That’s what AI assistants will replicate — data-in, decision-out.
Here’s what still matters:
- Product titles and categorisation.
- Image quality and order.
- Reviews, Q&A, and average rating.
- Variant logic and fulfilment configuration.
Do a quick test.
Search for one of your products on a marketplace.
Does it represent you well? Would you buy it if you knew nothing about your brand?
Now look at your D2C version. If the marketplace performs better, your problem isn’t AI. It’s consistency.
AI Searches Differently
Most teams are treating AI like an enhanced search bar. It’s not. It’s a buyer.
Your next customer won’t type “running shoes”.
They’ll ask:
“Find me 3 pairs under £100 that are waterproof, wide-fit, and reviewed by runners.”And they’ll get a shortlist.
No homepage. No filters. No scrolling.
You won’t even know they were looking unless your product is selected.
This shift means:
- Your attributes need to be visible and structured.
- Your imagery must carry context and relevance.
- Your content has to explain benefits, not just features.
We’re helping brands map these interactions — not to predict the future, but to spot where they’re already invisible. AI isn’t waiting for permission to replace your funnel. It’s already doing it.
Workshop: map your AI shopping path
This exercise helps your team understand how AI will interact with your catalogue — and where your current setup breaks down.
Step 1. Choose a realistic AI prompt
Start with how your customer might shop through an assistant. Try:
- “Find me a birthday gift under £50 for a 5-year-old that arrives by Friday.”
- “Show me a waterproof hiking jacket that fits in a carry-on and isn’t black.”
- “What’s the safest car seat under £300 that doesn’t need isofix?”
Step 2. Play the assistant
Without using your website, try to:
- Find eligible products
- Validate price, shipping, and attributes
- Shortlist the best options
- Explain the selection
(Better yet, use ChatGPT’s “Agent Mode” and ask it to perform the AI prompts in step one exclusively on your website to know how you perform outside of your competitors)
Step 3. Spot what’s missing
Look for:
- Unstructured or inconsistent data (e.g. missing size tags, vague materials)
- Delivery filters that don’t match real fulfilment logic
- Products that sound great on a PDP but vanish when reduced to raw data
Step 4. Turn it into action
- What could merchandising fix this week?
- What needs a PIM or platform update?
- Where do you need to rethink categories, tags, or variant logic?
This is how AI will decide. Not based on branding or conversion flows — but data fidelity.
Don’t replatform. This is a readiness problem.
AI shopping isn’t something to replatform for.
Every major ecommerce platform — Adobe, Shopify, the rest — will adapt and already have.
They’ll expose APIs. They’ll integrate new workflows. You don’t need to lead on that.
What you need is to be findable, selectable, and buyable.
That means:
- Data that’s accurate, readable, and relevant.
- Teams who know how to structure product catalogues for discovery, not just UX.
- Internal accountability for how you’re preparing.
You don’t need a new stack.
You need new habits.
While others stall, you can act
It’s tempting to wait.
Luxury brands will say, “We rely on in-store experience.”
Furniture brands will say, “Our customers want to see the product.”
High-AOV brands will say, “This won’t work for us yet.”
But AI is closing those gaps fast. Image-to-video, AI sizing guides, multi-modal comparisons — all in motion already.
The smart brands aren’t betting on feature timelines.
They’re asking:
- What are we already exposed to?
- What can we test today?
- Where are we invisible?
And then they’re acting.
That’s what readiness looks like.
JH – The Breakthrough Agency.
We work with Adobe Commerce and Magento brands who know growth is being stalled by friction. Whether it’s disjointed UX, misaligned strategy, or data that doesn’t serve the customer, we help teams move from output to outcome. Topics like AI, product discovery, and payment readiness aren’t abstract — they’re part of the weekly conversations we have with clients.