Welcome to a New Era of Agent Experience Design

As Chief Experience Officer at Salesforce, I work with business leaders to bring AI agents into their customer touchpoints and employee experiences. Most leaders are at the very beginning of a greater vision for how to integrate AI and they ask a fundamentally human-centered question: How do I make sure agents will work well for people? 

For this reason, experience matters. More than ever.

Experience design is innately human-centered, meaning it focuses on creating positive interactions, feelings, and outcomes for people. How do we ensure AI agents are human-centered as they’re integrated into businesses? We are building a new expertise in agent experience design, where AI agents are treated as a new kind of user. Great customer experiences will rely on this rising category of design.

As we partner with our new digital coworkers, AI agents will have unique strengths as users, people will have unique strengths as users, and they will need simple ways to communicate with each other. With this in mind, here’s what you need to know about agent experience design. 

What we’ll cover:

What is agent experience design?
AI agents become powerful work delegates
People become agent orchestrators
Human-to-agent interfaces become flexible and personalized
A new foundation for the era of experience

What is agent experience design?

When I first wrote about designing trusted generative AI, I envisioned AI agents as a new category of users. They can increasingly interact with the world, learn from those interactions, change behavior, and remember what they learned. While an AI agent certainly doesn’t have human-like experiences, it is, in its own way, a user with unique goals, needs, and limitations. 

Agent experience (AX) design is the development and “optimization of digital environments so that AI agents can efficiently and effectively operate within them,” and orchestrate successful human-centered outcomes. I define AX as encompassing both design for agents and design of agents, because both are necessary to make sure AI agents prioritize people’s goals and needs. 

Another way to look at it is AX should result in great usability for agents and usability of agents. Thus, great AX leads to great UX.

If the AI agent encounters a broken process, protocol, or even inconsistent wording across systems, it’s unable to perform that task. That’s a bad experience for the agent that, in turn, leads to a bad customer or employee experience. 

Let’s consider a simple example:

An agent designed to help with customer order changes gets a call from a customer who wants to change the at-home delivery time for their new refrigerator. Next, this agent needs to know who the customer is, package-tracking info, product details, order history, and how to communicate with delivery service coordinators. Finally, to complete this request, an AI agent needs to connect the dots across multiple kinds of data, systems, companies, and even other AI agents.

Building effective solutions for AI agents will drive new and different design requirements. As we’re learning through practice, agent experience design is a critical part of success with AI agents. A new era of experiences designed to work concurrently for both agents and humans – and interactions between the two – is an exciting prospect. 

Back to the top

AI agents become powerful work delegates

AI agents are poised to reshape every customer touchpoint across all sectors. One strength that an AI agent brings to the workplace is its ability to learn across vast amounts of information, and find effective ways to solve problems. Consider the Era of Experience, which explores how an AI learns from its interactions and experiences. 

Agents are purpose-built AI for specialized kinds of actions. They have specific expertise, designed to help them successfully complete certain jobs. As people delegate work to agents, agents need to reason their way to the correct outcomes, rather than following explicit pre-determined steps. If they encounter challenges, such as incomplete information or an unexpected result, a well-designed agent experience will help AI agents successfully navigate those issues.

Building agent-friendly systems will require new skills and standards. My team’s mission is to make Salesforce fast, compelling, and easy to use. We can extend these priorities to agent experiences. AI agents favor executing tasks directly through systems and services, bypassing traditional user interfaces. Let’s consider some criteria for designing an agent-friendly system:

• High-quality APIs and fast response times.
• Well-organized knowledge management.
• Effective ways to learn from structured and unstructured data.
• Agent-to-agent protocols that standardize agent actions across clouds. 

Designers are developing new skills to make sure agents are optimized to produce human-centered outcomes:

Back to the top

People become agent orchestrators

As agents become ubiquitous, every person may be managing – or orchestrating – multiple agents on a daily basis. How do you bring it all together to know that agents are making the right progress? Agent experience design needs to give people efficient ways to steer multiple agents toward the right goals with the right level of control over those outcomes.

This level of orchestration certainly gets easier if an interface flexes to a person’s individual preferences. There are many different ways that people prefer to communicate with one another, and the same will be true for communicating with AI agents. 

This is far more than just a text box and conversational interface. Good communication includes helping people understand an AI agent’s capabilities. It means knowing how to organize and optimize data so that an agent user has access, knows its goals, and knows how to navigate all that complexity on behalf of a human user. There needs to be intuitive ways to know whether progress is going in the right direction. These are just a few new orchestration design challenges for agent experiences. 

Also, AI agents need structured ways to learn when and how to interact with humans in situations where they need help. AI agents will need to know when to give control back to humans, or to ask humans to verify an AI agent’s reasoning process. These kinds of interactions need to be delivered by the agent to a person at the appropriate time and place, through effective user experiences.

Human-to-agent interfaces become flexible and personalized

This is the topic that tends to get the most attention from leaders who value design. For years we’ve relied on interactive demos and prototypes to specify how an experience should work. Now, AI agents are disrupting decades of investment in graphic user interfaces. And yet, most design teams are rushing to define guidelines that reduce clicks and help people and AI work together on shared information.

These interfaces can be as simple as one single screen where a user can perform many different tasks. These work well for early generations of AI agents. And it’s exciting when voice-based technologies are integrated into these interactions to make them feel more natural and engaging to people.

Deep personalization also improves ease of use for people as they learn how to interact with agents. Experiences that flex and reflow to a person’s preferences will be essential as AI becomes ubiquitous. Enabling more UI flexibility sets the foundation for agentic experiences. With the release of SLDS 2 (beta), we’ve brought a new CSS-based architecture that gives customers the ability to deeply customize the look and feel to achieve a consumer grade user experience.  

As we’ve explored different kinds of dynamic and personalized user interfaces, we’ve learned what matters most is staying flexible as people and agents find their new stride together. They will search, strategize, create, converse, analyze, monitor, and collaborate their way into this new future. We no longer need to squeeze every person into the same user journeys or click paths. We need to stay flexible and partner closely with our customers as they discover what makes AI agents feel like an embedded and invaluable part of how they get work done. 

A new foundation for the era of experience

Revolutions in the way we work are never driven by sheer brilliance of technology. What drives it is the willingness of people to try something new and learn through their experiences. Every interaction with marketing, sales, retail, websites, products, and customer service contributes to a person’s trust in a business. This is a core tenet of experience design. 

As AI agents become new users of your systems, they will make decisions on the best way to complete a task on behalf of the people and businesses they serve. What will they find when they arrive at your doorstep? Prepare for their arrival with Data Cloud to harmonize your data and metadata. This will make sure AI agents can understand how your business works and get a comprehensive view of your customer relationships. In parallel, agent experience design will ensure that AI agents can turn that data into powerful and compelling experiences for your customers and employees.

Back to the top