Jakob Nielsen: AI Redefines UX by Centralizing Everything

The Evolution of Interfaces: The Impact of AI on Jakob's Law

The evolution of digital interfaces is undergoing a significant transformation, with artificial intelligence playing a central role in this dynamic. Increasingly, users are completing their tasks through a single interface, often conversational, thanks to virtual assistants like Claude. This trend aligns with Jakob Nielsen's law, which emphasizes the importance of familiar interfaces for users.

The Persistence of Jakob's Law

In 2000, Jakob Nielsen formulated a law that remains relevant despite technological advancements: users spend the majority of their time on other sites, which influences their expectations. This law highlights that users expect new interfaces to function like those they already know. This continuity is crucial for pioneers in user experience (UX), as it ensures that users are not confused by overly radical innovations.

The Importance of Design Systems

To meet these expectations, companies have established dedicated teams for design systems. These teams ensure that each screen adheres to best design practices while innovating within the brand framework. For decades, this meant aligning with web conventions, as that is where users developed their habits.

UX components, for example, show a standardization of the semantic layer for existing design systems; many of these models need to be standardized for agents. The list will grow with A2UI, mentioned later in the article.

The Rise of Conversational Assistants

Today, more and more tasks begin in an assistant rather than on a website or application. Users ask assistants like Claude or ChatGPT to draft emails or retrieve data, centralizing many tasks that previously required multiple interfaces. This centralization simplifies the learning and adoption of new tools, as users no longer need manuals to understand how to interact with these systems.

The Speed of Adoption of New Interfaces

Jakob's law, traditionally seen as a warning against deviations, now explains why the adoption of new interfaces is so rapid. The more interfaces resemble each other, the fewer new skills users need to acquire, encouraging them to try new tools. This homogeneity of interfaces also allows companies to be more creative in inventing new solutions.

The Need for Increased Standardization

Design components and their semantic layer are becoming increasingly crucial. They make explicit the conventions and expectations of users, facilitating navigation between sites and understanding agents. This does not challenge Jakob's law but shifts attention to a different set of surfaces requiring increased standardization.

If most use cases now go through the assistant, the question for anyone building software is not just whether your product works when accessed through this surface — it's whether you have given the agent enough context to use it well by employing standard models.

The Reduction of Surfaces

Historically, each computing task had its own destination, whether an application or a website, each with its own interface. This led to a saturation of home screens and a proliferation of browser tabs. Jakob's law existed to help users navigate this complex landscape.

The Integration of Tasks by Assistants

Assistants, such as Anthropic Claude, Microsoft Copilot, Google Gemini, and ChatGPT, now centralize these tasks in one place. Users no longer need to open a different tool for each task, as assistants integrate these functionalities into a single interface.

These are not just standalone assistants, as the chat tools where people already live have become full-fledged conversational surfaces: Slack now runs AI agents directly in its channels, including third-party ones from Anthropic and others, and Microsoft Teams integrates Copilot and its agents in the same window where colleagues chat.

The Impact of Chat Tools

Chat tools, like Slack and Microsoft Teams, also integrate AI agents into their channels, allowing users to perform tasks without leaving the conversation. Open-source projects, such as OpenClaw, go even further by integrating AI models into various messaging applications.

The conversation that people were already having with each other is now also where they talk to the software, so users do not have to relearn it.

The Simplification of Interactions

OpenAI allows third-party applications to operate within ChatGPT, facilitating tasks like searching for listings or creating presentations without leaving the conversation. Anthropic's model context protocol also standardizes these interactions, connecting an assistant to data and tool systems.

Claude accesses your calendar, documents, and issue tracking system through this single connection.

The Evolution of Assistants

Assistants do not just respond to simple requests. They synthesize information from multiple sources to provide complete deliverables. This reduces the cognitive load for the user, who no longer needs to visit multiple sites to obtain the necessary information.

It shifts towards work and moves between sources on its own. The task of visiting a dozen sites, reading each one, and connecting what they mean — the exact cognitive cost that Jakob's law sought to reduce — is now almost entirely absorbed by the agent. The user states the result; the interpretation happens out of sight.

The Transformation of User Expectations

The number of surfaces each user directly interacts with is decreasing, as the assistant becomes the common surface. The user states the desired outcome, and the assistant does the work in the background.

The Disappearance of Certain Interfaces

Sometimes, a surface does not just reduce to the assistant but disappears entirely. Agents can perform tasks in the background, such as sorting emails or rescheduling meetings, without requiring direct user interaction.

A surface does not always reduce to the assistant — sometimes, it never appears, as the work has become ambient. An agent monitoring your inbox sorts the receipt, flags the renewal, and reschedules the meeting in the background, and the first time you hear about it is a summary afterward, if it even happens.

Actions to Take for Businesses

• Map the user journey: Identify the steps in the user journey that occur in an assistant and those that remain in the product.
• Observe task beginnings: Note the moments when users turn to an assistant before a traditional application to understand user expectations.
• Reevaluate success metrics: Abandon metrics based on time spent in the application, as the assistant can accomplish tasks more efficiently.

When attention focuses on the assistant, the conventions that users carry also concentrate with it.

The Importance of Assistant-Centered Conventions

Jakob's law underscores the importance of reference points. When users spend their time in an assistant, it becomes the reference model. Clear language interactions and contextual responses become the norm, and users expect systems to operate this way.

When time is spent in an assistant, the entire reference set becomes the assistant.

The Adaptation of Businesses to New Expectations

Businesses must learn these new conventions to design systems that meet user expectations. This involves working directly with assistants like Claude or ChatGPT to understand interaction patterns and adjust their features accordingly.

This is the point — assistants are becoming the default in many cases, which aids in adoption.

Do Not Reinvent the Wheel

The clear language request box has become a convention. Companies should avoid reinventing this interface and focus on enhancing the user experience by adhering to these new standards.

I previously referred to the clear language entry as the request box; the broader point is that the request box is no longer one model among others. It is becoming the model — the entry point through which an increasing share of tasks pass.

This is Jakob's law doing exactly what it has always done. It’s just that the place where users spend their time has consolidated, so expectations have consolidated with it.