The internet as we know it is dying. What comes next?
The future of online shopping envisioned by Andi Flütsch
Most of us have experimented with AI tools by now — drafting emails faster, summarizing long documents, researching things that used to take hours. These tools are genuinely useful, and they represent the question most people are currently asking: how can AI help me do what I already do, but faster?
It’s the right question for today. But it’s the wrong question for the future.
What interests me is the next move: the one that fundamentally changes the way we work and interact with technology. The question is: “Will we still need to do this at all?”
This article summarizes my vision for online shopping. I invite you to think similarly about your own field.
The web was always a workaround
The visual, navigable web we built over thirty years was never the ideal solution. It was a workaround.
We manually curated product metadata and designed product pages, driven by UX flows to guide human visitors visually and emotionally towards a purchase decision. Every button, every product photo carousel, every “customers also bought” recommendation was purposely built to compensate for one simple fact: our systems just didn’t understand what you needed.
With AI, that constraint is disappearing.
The personal assistant layer
Imagine this, five years from now. You want a new pair of running shoes — a specific model, your size, in pink. You tell your personal AI assistant. Within seconds, it has checked every available supplier, compared prices, verified your size is in stock, confirmed delivery times, and cross-referenced your payment preferences. It gives you three options ranked by your stated priorities. You say yes. Done.
You never visited the online shop. You never saw the meticulously designed product page. You never entered your credit card details.
Or the even more futuristic version, ten years from now: your AI-powered home assistant notices that you’re running low on milk. It doesn’t ask you. It just checks your preferred grocery stores, finds the best price and then places the order.
In both cases, the transaction happened entirely without a human interacting with the online shop. And this isn’t as far off as it sounds — a version of it already exists: When you search for flights on Google or a comparison site, you’re not visiting each airline’s website one by one. A system is querying dozens of airline databases simultaneously, pulling structured data in real time, and surfacing the best results. That’s the machine layer, already working and trusted by millions. What’s coming is simply that logic applied to everything else we buy.
What happens to the online shop
The honest answer: the transactional online shop as we know it today will disappear.
Not immediately, and not completely. But the elaborate architecture we build to guide humans through a purchase journey becomes redundant when an AI handles that journey on their behalf.
What replaces it is something far less glamorous but more powerful: structured data. A manufacturer provides all the available data of a product in a machine-readable format so that any AI, anywhere, can understand exactly what is being sold.
And the barrier to entry is lower than you might think. A single, well-structured and standardized text file from a manufacturer is enough. A small artisan shoe brand with an accurate, complete product file competes on equal terms with companies with big marketing budgets, because the machine layer rewards completeness.
The economic incentives are clear. If your product data is incomplete, the AI won’t show your product — and will surface a competitor’s product instead.
This produces something paradoxical: the AI-mediated web will have better, more reliable product information than the human-browsed web ever did. Not because companies became more ethical — but because invisibility became the punishment for non-compliance.
An unexpected bonus: the end of wasteful returns
Here’s where the story gets genuinely optimistic.
Customers return around 30% of everything bought online — three times the rate of physical stores. In fashion, that figure reaches 40%. In 2022 alone, 9.5 billion pounds of returned products were sent to landfill in the United States, because it’s often cheaper for retailers to destroy a return than to restock it.
(Source: https://www.cleanhub.com/blog/ecommerce-returns-environmental-impact)
Most returns happen for the same reasons: wrong size, wrong fit. These are information failures. Customers didn’t have accurate enough data, so they ordered two sizes and sent one back.
Take the running shoe scenario: foot-scanning technology (already available in apps from major shoe brands today) gives your AI assistant a precise record of your measurements. The product metadata includes exact dimensional data from the manufacturer. This allows the AI assistant to match and order the one pair that will fit. No guessing. No “I’ll try a 42 and a 43.”
Fewer shipments. Fewer returns. Fewer products destroyed. This gives manufacturers an additional economic incentive to share complete and accurate product metadata — not just to be visible to AI agents, but to cut the reverse logistics costs that are bleeding their margins.
The death of the upsell
There’s another part of the current online shopping that won’t survive this transition: the upsell.
The entire architecture of e-commerce is designed to influence you at the moment of purchase. “You might also like.” “Customers also bought.” Bundle offers at checkout. We’ve grown so used to these that we barely notice them — but they are not services to the customer. They are commercial nudges designed to increase basket size.
In the AI assistant model, they become impossible because you as the human will not be on the checkout page. And that is a feature, not a bug.
Your personal AI works for you. The moment it starts recommending products because a retailer paid for placement, it has betrayed its core value proposition. An assistant that can be commercially influenced is not an assistant — it’s a salesperson in disguise. Users will be acutely sensitive to this, and any AI caught playing both sides will lose trust and market share fast.
The “classic” transactional upsell dies. But something more interesting replaces it.
A brand’s AI can observe purchases over time, build a genuine understanding of what you like, and reach out afterwards — not with “you bought running shoes, here are more running shoes,” but with something more considered: a notification six months later that the model you love comes in a new color, or that something you’d genuinely want is on sale.
For brands willing to invest in that genuine relationship, the long-term value is arguably higher than any upsell at the checkout carousel ever generated.
Two internets, not one
What emerges is a web split into two distinct layers.
The first is the machine layer — the infrastructure of APIs, structured data feeds, and ordering systems where transactions happen. It’s largely invisible to humans, and it rewards accuracy and completeness over beauty.
The second is the human layer — brand storytelling and desire creation. This is where people go not to buy things but to want them. And this layer is migrating: the brand experience that once lived on a product page is moving into video, into social media, into experiences designed to inspire rather than to process a transaction.
When you encounter a product in the human layer and feel ready to buy, you won’t be taken to a checkout. You will be directed to your AI assistant, which already knows your size, your preferences, your delivery address.
Of course, that level of personalisation only works if users have genuine control over their own data. That trust still needs to be earned — by the technology and by the companies building it.
What this means for you
The internet is not dying, but its purpose is being redefined. The front-end of e-commerce is moving away from web pages and towards AI assistants. This shift demands a new way of thinking.
If you are a developer, your focus will shift from building beautiful user interfaces to creating robust, well-documented APIs and data feeds that AI agents can easily consume.
If you are a marketer, your job is no longer about optimizing checkout flows. It’s about creating the brand story and the desire for the product on the human layer — social media, video, and other experience-rich platforms. How will you make someone want a product enough to ask their AI assistant to order it?
If you are a product manufacturer, you must provide the most complete metadata set possible for your products to online shop platforms, to ensure the AI assistant will surface your product.
If you are a brand owner, the question is how to prepare for a world where your website is not your primary sales channel. Are you building a brand that transcends the product page?
The one question worth sitting with, whatever your role:
What part of this process exists only because a human needed to be in the loop?
Answer that honestly, and you’ll know exactly where to focus next.
Written by Andi Flütsch
If this resonated, I’d love to talk.