What is Experience Architecture and Why You Can’t AI Without It

What is Experience Architecture and Why You Can't AI Without It

How we think about user interfaces has shifted from static designs to creating 'invisible' systems that generate experiences on the fly.

Share article

A common misconception about designing for AI is that user interfaces are mostly going away. In reality, they're evolving - moving from static screens into more dynamic, conversational, and adaptive experiences. The rules of engagement are changing how people and AI connect, communicate, and collaborate.

UX designers have an opportunity to shape these rules, to build systems that enable people to grow beyond "playing AI Mad Libs with a prompt" toward richer, more intuitive interactions. To fully engage with how these dynamic interactions work requires a structure to ensure experiences are coherent, consistent and delightful. This is where experience architecture comes into play.

Here's what we'll cover:

What is experience architecture?
A good runtime experience relies on good architecture
Three examples of how experience architecture shows up
Get to know the building blocks of success
Architect your AI design practice

What is experience architecture?

Experience architecture is the structural layer of product design. It provides an "invisible" scaffolding that complex, generative AI systems use in real-time responses to create experiences that surface the right things at the right time and in the right places. Good UX design for these runtime experiences enables performant, human-centered interactions where people understand what the AI is doing, can trust its outputs, and feel confident engaging with it. Experience architecture makes that possible.

To create systems that work well with human communication patterns, experience architects partner closely with conversation designers. Conversation design defines the flow, tone, and recovery patterns that make interactions feel natural. Experience architecture provides the structural models and signals that keep those interactions coherent across contexts, modalities, and agents.

Experience architecture defines the rules of engagement for dynamically generated interactions, answering new types of experience questions:

• How does the system represent status and progress to the people using it?
• How does the system attribute the sources and evidence used for its outputs?
• How does the system explain the reasoning it used to get to an output?
• When humans need to be engaged, how does the system keep humans in the loop?
• How does the system represent what it knows about the person using it?
• How does the system surface questions and outcomes from background work?

The answers to these questions live in the underlying experience models and frameworks. By shaping how dynamic AI behaviors unfold, experience architecture ensures interactions are not only functional but also seamless and delightful for users.

Return to the top

A good runtime experience relies on good architecture

An important distinction to recognize when building AI systems is the difference between design time and runtime experiences. Design time experiences shape the system's capabilities, while runtime experiences shape how someone perceives and experiences those capabilities. Both are fundamental, but runtime design ultimately defines whether an interaction feels coherent, human-centered, and reliable.

Design time defines the models, workflows, ontologies, guardrails, and prompts that determine how the AI should operate. It focuses on prompt editors, training dashboards, orchestration tools, and such that prioritize usability for builders and admins who configure these systems.

Runtime is the moment of use, or when the system is actively working with real data, users, and contexts. The runtime experience centers on what end users encounter when they engage with the system: signals like status indicators, citations, agent hand-offs, memory, and guardrails that make the system's behavior understandable, trustworthy, and smooth in real time.

Return to the top

Three examples of how experience architecture shows up

To make this more tangible, let's look at a few examples of experience architecture that we use to make experiences on the Agentforce platform better for the people who use it.

1. Making system status visible

Status indicators are ways that a system communicates to its users what's happening and what it has done. One of the hallmark heuristics - or cues - of good interaction design, these take on new forms for AI systems to keep people informed and give appropriate feedback.

When do they show up: While the system is working.

What do they do: Shows progress ("retrieving…", "generating…", "agent A is planning…").

How do they help meet people's goals: Reduces uncertainty, provides reassurance that the system is active, and gives clear expectations for wait time.

There are four common types of status indicator in current AI systems:

1. System typing indicator or "thinking" indicators.
2. Progress feedback on multi-stage processes and workflows.
3. Streaming responses, when the words that are part of an answer appear to show progress.
4. Reasoning indicators that convey how the system is reasoning to provide an answer.

2. Attributing sources and evidence

Citations in traditional contexts are references that attribute information to external sources. They're used in research, journalism, legal writing, and beyond to signal important contributions to an author's work. Similarly, citations for generative AI systems provide people with an affordance, or clues, to understand what contributed to the system's output.

When do they show up: After the system produces an output.

What do they do: They show evidence of where information came from, which subsystems or agents contributed, what documents or APIs were used.

How do they help meet people's goals: Builds trust in the system and provides an opportunity for verification. Have a way to check the provenance of an answer or follow links for more context.

3. Handing off to humans

Agent handoff: This occurs when the AI system reaches a point where a person needs to step in, such as when a customer service agent is unable to fulfill the user's request. The agent needs to gracefully transfer that interaction to a human agent who can help the person accomplish their goal or clearly explain why the goal can't be met.

When does this show up: The AI system reaches a point where a person needs to step in, such as when the task requires judgment, empathy, or domain knowledge the system can't provide.

What does it do: Shifts control from the AI to a human, carrying forward context like conversation history, user intent, and relevant data so the human doesn't start from scratch. Agent handoff can also happen from human to agent, when the agent takes over from the human.

How does this help meet people's goals: Provides a smooth, transparent transition that preserves trust, minimizes user frustration, and empowers the human to continue seamlessly.

Return to the top

Get to know the building blocks of success

While many aspects of designing for AI experiences are just good UX design, there are a few key pillars that make an experience architecture practice more successful.

Cross functional collaboration

To design the experience architecture for AI-driven interactions, work with product managers, engineers, and software architects to understand existing capabilities and the systems that drive them.

Understanding your AI

It's critical to understand the software architecture and how AI systems are generating interactions: what are the inputs, what happens with those inputs, and what the system outputs in response. These are the building blocks for your experience architecture.

For example, on the Agentforce platform, the reasoning engine is the heart and soul of generating dynamic interactions. Understanding how that works helps designers create interaction patterns that are aligned with how the system actually performs and accurately reflect it in runtime experiences.

Experience modeling

Mapping the experience architectures, frameworks, experience models, and principals is essential. Make a map of where dynamic experiences are happening in the UI, what's happening, and when it's happening. Define how and where that happens across a user's journey and assess how well the current capabilities are meeting users' goals.

Market awareness

Look for examples of how this is done elsewhere and draw on analogous patterns for ideas and inspiration.

Understanding the outcomes that matter most to the people

When using your system, what do people want to achieve and how should people feel when interacting with your product or service?

Sketch the possibilities

From hand-drawn sketches to AI-generated interactive prototypes, make your ideas tangible. Do it in ways that let others fluidly collaborate with you. Although a common tendency is to start UX with designing for design time experiences, it's far more effective to begin with working out the desired runtime experience. When you're clear on what you want the system to generate, it's much faster to craft a usable and efficient design time experience.

Return to the top

Architect your AI design practice

Experience architecture is critical to your AI design practice. It's the structural layer of product design that keeps dynamically generated experiences coherent, consistent, and aligned with intended outcomes. It serves as the foundation that allows AI systems to deliver clarity and reliability at runtime.

Beyond being part of the product itself, it defines the outcomes that matter most for people and the holistic journeys that help them accomplish those outcomes across every touchpoint.

Special thanks to Adam Babin, Claire Bain, Cassy Funk, Patrick Hermiller, and Vineet Mishra.

Return to the top

Share article

As a design leader on the Agentforce platform, I lead teams who design intentional interactions to help people understand what the AI is doing, trust in its outputs, and feel confident engaging with it. I'm fortunate to be collaborating with a talented group of people focused on agent...Read More observability, multi-agent interactions, and runtime experiences. What I enjoy most about designing AI driven experiences is unraveling how to work with this new design "material" and how it can be shaped to anticipate and meet users' needs in meaningful ways.

More by JoEllen