From UX to AX, Designing for a World Where Users Aren’t Always Human
The role of those who design user experiences has always been to make people’s lives simpler and clearer when they interact with digital products or services. For years, it was enough to focus on what users saw, clicked and interpreted, working to build interfaces that could hide complexity and make choices immediately understandable.
This context is now changing with the spread of AI agents, artificial intelligence systems that take on an active role in digital interactions and operate as genuine delegates of the user.
These systems are able to perform real actions, such as making reservations or completing transactions, marking the shift from LLMs to LAMs. In this new landscape, designers must support agents in their work by treating them as an additional type of user. There is one key difference though: they are not people, they do not reason like people, and they work precisely because they do not think. They process information, read structured data and interact with systems through a direct approach that has nothing to do with the visual form of interfaces.
This leads to an uncomfortable realization for designers: what helps a human can hinder an agent. For years, interfaces have been crafted to guide people through reduced information, short sequences of actions and clearly structured steps, all with the goal of lowering cognitive load. This approach works well for the human mind but creates unnecessary fragmentation for an autonomous system. An agent does not need reassurance or guided flows. It performs best when all relevant information is available in one place, when the structure is explicit and when there are no barriers introduced to protect humans from complexity.
This is why User Experience alone is no longer enough. A new perspective is emerging alongside it, often referred to as Agent Experience. This does not replace user-centered design but expands it. The two actors interacting with digital services, humans and agents, perceive the world differently and therefore require environments designed with those differences in mind. A platform that works flawlessly for a person may be inefficient for an agent, while a system optimized for a machine can easily confuse human users. The challenge is to keep coherence, balance and continuity, enabling AI systems to handle complex tasks without losing sight of the human need for control, clarity and understanding.
In practice, especially in e-commerce, designing for agents requires very specific technical decisions. The first concerns content structure. A website must be semantically readable and interpretable. Structured data, such as schema.org standards, become essential to describe products, variants, prices, availability, reviews and shipping policies. Without this semantic layer, an agent cannot interpret the catalog correctly, increasing the risk of providing incomplete or inaccurate information to the end user.
“As systems start making decisions, the designer’s job isn’t to design for them but to design in a way they can understand.”
Another crucial aspect is the availability of clear, reliable and fully documented APIs. An agent must be able to add a product to the cart, check availability or determine delivery costs and times through direct technical interfaces rather than by mimicking human clicks. This requires headless architectures, strong authentication protocols, end-to-end security and the ability to manage tokens or permissions for checkout operations.
More advanced implementations, such as those based on Stripe’s Agentic Commerce Protocol integrated with platforms like Shopify, show what it means to build a truly agent-ready system. In these scenarios, a user can complete a purchase simply by saying something like “Order that black T-shirt in size M for delivery tomorrow.” The agent interprets the intent, queries the catalog, picks the correct variant, calculates costs and availability, manages payment and shows the user only the final confirmation. This workflow is possible only when the website provides clear data, coherent structures and accessible features through secure endpoints.
Designing for agents does not mean abandoning visual interfaces or replacing traditional UX. It means acknowledging an additional, invisible audience that does not see, click or interpret graphical signals but instead reads data and functions. When agents become intermediaries between the site and the user, much of the experience shifts behind the scenes. The quality of communication between systems becomes the deciding factor in how the final user perceives effectiveness.
In the end, the evolution toward digital environments built for both people and agents is no longer theoretical. It has become a necessity. Services that adapt will be more accessible, more functional and more competitive. Those that continue to design experiences only for humans risk becoming difficult for intelligent intermediaries to interpret and, eventually, less relevant for users as well. UX is not being replaced; it is expanding. Designers now face a new and unavoidable responsibility: creating digital spaces that welcome two different forms of intelligence without losing sight of the main goal, which is to make the experiences we navigate every day easier, clearer and genuinely useful.
