AI Redefines Digital Interaction: Designers Under Pressure

A Revolution in Digital Interaction

For decades, our way of interacting with software has relied on well-established principles. The traditional user interface, inspired by the desktop metaphor, has long dominated our digital experience. This model, which mimics a physical workspace, is familiar to us: opening applications, moving files into folders, navigating through menus. Even with the advent of mobile and web technologies, this structure has continued to guide our interactions with software.

However, the rise of artificial intelligence (AI) systems challenges this established grammar. Conversational interactions, although present for some time with assistants like Alexa or Siri, were until now limited to scripted responses. New AI models, on the other hand, are capable of understanding user intent and acting accordingly, thus modifying interfaces in real time.

We are witnessing a transition from task-centered interfaces, where each step is controlled by the user, to intent-driven interfaces, where the user expresses a need and the system begins to work immediately.

The Emergence of New Interaction Models

As part of this transformation, the startup Thinking Machines Lab, founded by Mira Murati, former CTO of OpenAI, recently published a research report on interaction models. This company proposes a new approach to AI, with models designed to enable real-time multimodal exchange, integrating audio, video, and text.

In the field of design, an interaction model is the framework that connects the various functions of a product. The models developed by Thinking Machines Lab aim to transform this interaction into a seamless and integrated experience.

A Significant Coincidence

Although the term "interaction models" is used at two different levels, there is a connection between them. The first layer concerns the internal workings of the system: how it processes data, maintains context, and responds to users. The second layer, that of design, pertains to the user experience: what the user sees, what they click on, and how they perceive the system.

Design plays a crucial role in translating the model's capabilities into a user-friendly interface. When the model acquires new skills, such as the ability to hold a real-time conversation, the design must adapt to make these capabilities accessible.

In demonstrations from Thinking Machines Lab, interaction models simultaneously take into account audio, video, and text, responding even while the user is speaking. They can intervene in case of an error, react to what they see via the camera, and translate languages instantly.

In one example, two users ask the model to create a graph of Uber's revenues and costs. The graph, generated in real time, is a generative interface created to meet this specific request. While the model works in the background, the users ask a question about Uber's performance this year, and the model continues the conversation while finalizing the graph.

What fundamentally changes is the pace of interaction. Instead of waiting for the machine to respond, the exchange unfolds seamlessly, making the interaction more natural and fluid.

Challenges for Designers

With the evolution towards continuous interaction, the role of designers is changing significantly. Here are some key considerations:

• Reconstructing the mental model. The desktop metaphor provided a tangible image for navigating software. With AI, this image becomes abstract. Users must understand a system that interprets their intentions and acts accordingly. Designers need to make this process visible and comprehensible.

• Changes in input and navigation. A user may start an interaction in a chat, move to a generated interface, and then to a permanent dashboard. Designers must establish new conventions to guide the user through these transitions.

• Deliberate choices of intervention points. Older interfaces gave control at every click. The new, faster interfaces require designers to decide where to insert options for revision, cancellation, and confirmation.

• Increased importance of human judgment. AI models have limitations. For example, a study by TCS Research showed that models overestimate task durations and cannot accurately gauge elapsed time. Designers must anticipate points where human intervention is necessary for contextual decisions.

Some predict that traditional interfaces will disappear, replaced by AI agents capable of managing end-to-end tasks. However, where human responsibility is crucial, the user interface remains essential for understanding and intervening in the process.

In enterprise software, where decisions often need to be audited, new interaction models must make the system's actions visible and traceable, allowing users to intervene and revisit past actions.

The Challenge of Building a New Relationship

Designer Frank Chimero stated that people ignore design that ignores them. This is a test that new interaction models must pass. The desktop metaphor worked for forty years by providing a clear mental model. The new models, whether conversational, agentic, or multimodal, still need to achieve that clarity.

We are entering a new era of relationships between users and systems, a relationship we still need to learn to define and master.