This article is based on a conversation between Stephany Cardet and Matthew Darrow from The Unexpected Lever podcast.

‍

The rise of agentic AI—systems designed to act autonomously and interact seamlessly with users—is reshaping how we think about design in technology. Unlike traditional SaaS applications, which rely on graphical user interfaces (GUIs) with buttons, dropdowns, and menus, agentic AI emphasizes dynamic, conversational interactions.

The user can design and define their own experience. For teams designing AI Agents, crafting intuitive, trustworthy, and empowering user experiences is critical.

This article explores the core differences between designing for SaaS and Agentic AI, what UI means in this context, and strategies for creating user experiences that prioritize personalization and trust without overwhelming users.

‍

From SaaS to Agentic AI: A Paradigm Shift

Agentic AI represents a massive technological evolution but does not change the core principles of UX. The fundamental design challenges of usability, trust, and user understanding are merely addressed differently in Agentic AI than in SaaS.

‍

SaaS: Predictability and Structure

The design of SaaS applications centers around GUI-based experiences that are intuitive and predictable. Features such as buttons and dropdowns guide the user, offering clear indications of what can be done within the software. This structure creates a controlled, predictable user journey.

These classic interfaces prioritize error prevention by quickly communicating to users what they can do in the product.

User understanding is therefore easy to achieve, but what they can do in the product is limited. Often, the human workflow must be designed around the software, and users have to request customizations as feature gaps.

‍

AI Agents: Flexibility and Autonomy

In contrast, Agentic AI focuses on creating an interaction model that feels like collaborating with a human teammate. Gone are many of the filters, folders, and buttons that are hallmarks of SaaS applications.

Agentic experiences are fluid and conversational. These systems predict user intent, offer proactive suggestions, and dynamically adjust based on user input. They empower users to shape their own experiences around their workflow.

Rather than rigid error prevention, the agentic experience offers error correction. Moreover, these systems infer what users want, rather than requiring explicit instructions. They respond to user feedback and also learn how best to communicate with individuals.

As Stephany explains, “It’s about helping users understand AI outputs, providing opportunities to fix errors, and ensuring users feel in control despite the system’s autonomy."

‍

What Does UI Look Like in Agentic AI?

In the world of agentic AI, the UI serves as a "familiar bridge" for interaction but is no longer the centerpiece. Consider it as a medium for communication—akin to how Slack or iMessage facilitates a conversation with another person but isn't the conversation itself.

As AI continues to evolve, the interface itself may become less critical. The intelligence behind the agent becomes the core feature, with the UI acting merely as a conduit for interaction.

“You can think of the UI as the body and the agent as the brain,” Cardet explains. “The interface is just a medium to facilitate communication. The intelligence driving the experience is what matters.”

‍

Characteristics of AI Agent Interfaces:

1. Adaptive and Minimalistic: The UI adapts to the user’s needs, offering just enough scaffolding to support interaction.
2. Multiple Modalities: Interactions may occur through text, voice, or integrated environments like email or CRM systems.
3. Human-Like Engagement: The UI mimics natural human conversation, mirroring user communication styles and preferences. Think of how a person's tone and body language guide your interaction with them during a conversation.

This approach reduces cognitive load, enabling users to focus on their goals rather than the mechanics of the interface.

Stephany shares: “We designed Ava, our agentic AI, to feel like collaborating with your best work friend. She learns your communication style, intentions, and how you prefer to tackle work. This allows her to not only handle tasks but also mirror your unique style back to you, creating a natural and fluid exchange.”

‍

Overcoming Adoption Barriers

However, this flexibility is a double-edged sword. Because the experience is so different from working in traditional SaaS applications, many users don’t yet intuitively know how to use AI tools.

In fact, users may be hesitant or actively frustrated while trying to engage these tools without familiar features like a save button, filter options, or explicit instructions.

A conversational interface can feel intimidating because it doesn't tell you where or how to start. The most successful users are the most creative, but that is not always practical for people just trying to do their jobs.

‍

Design recommendations:

• Build trust through transparency. The agent can clearly communicate what it knows or doesn't know and invites collaboration. Users are shown how the AI fits into their workflow and the outcomes they can achieve via that collaboration.
• Encourage curiosity in users by designing affordances and efficiencies to guide them in using the system (e.g., quick actions and conversational prompts to get them started).
• Focus on proactive experiences where the system actually anticipates user needs and offers solutions based on past behavior rather than waiting for explicit instructions.
• Create feedback mechanisms that allow users to validate or critique the agent's outputs via a thumbs up/down or by providing comments. This builds trust and rapport and also helps the agent learn to collaborate better with individual users.

‍

Conclusion: Crafting Invisible Excellence

Agentic AI represents a leap forward in how we interact with technology. By focusing on human-centered principles—personalization, trust, and simplicity—designers can create systems that feel like collaborative teammates, not just tools.

As Stephany says, “The best design is invisible. It’s technology so powerful that you barely notice it. It just works and feels completely natural to you.”

‍