In partnership with

Hey there! This week, we're diving into something that's reshaping how we think about AI interactions: the end-to-end experience of working with AI agents.

Why Agent Interfaces Matter Now

The chat interface we've all become familiar with? It's just the beginning. As AI agents become more sophisticated, we're moving beyond simple Q&A to complex, multi-step interactions that need thoughtful design consideration. In our recent survey of designers working on AI agents, a key insight emerged: the empty chat box is becoming obsolete. As one practitioner notes, "most people don't know what they want or how to start when it involves a series of steps!"

GenAI apps like ChatGPT, Gemini or Claude are all presented as chat interfaces. These first forms of Agents leave a lot to be desired from a human-interaction perspective. If we are to treat these agents as collaborators they need to be able to receive feedback in-context and share understanding, capabilities and an approach to the request.

Designers and creative technologists are just showing up to the party of the AI researchers who have built this amazing new technology. It is time to bring GenAI out of the lab and rapidly iterate on how we expect it to work.

This is the subject of fast principles, each week we will continue to dig into different approaches teams are taking to build these AI agent technologies into production ready experiences.

The "Magical Deliverable"

Every user comes to an AI agent with a goal in mind – what we call the "magical deliverable." It's that perfect outcome they're hoping for, whether it's:

• A perfectly summarized research paper

• Code that works exactly as intended

• A presentation that captures their vision

• An illustration that sparks the imagination

But here's the thing: getting there isn't just about the final output. It's about the creative journey.

Understanding the User's Journey

Before we can design better AI experiences, we need to understand how users currently tackle their tasks. This means looking at:

🔍 Current Process

• What apps are they jumping between?

• Where do they get stuck?

• What frustrates them most?

❌ Pain Points

• User’s struggling to explain what they want

• Getting off-track responses from agents

• Unclear next steps for tasks

• Agents inability to articulate needed user input

Building Better AI Experiences

From our latest research with designers in the field, here's what we're learning about creating more effective AI interactions:

1. Set Clear Boundaries

• Be upfront about what the AI can (and can't) do

• Guide users toward requests that will succeed

• Prevent off-topic wandering

2. Design for the Whole Journey

Think beyond the chat box:

• How does the user start?

• What happens after they get a response?

• Where might they need human help?

• When can they add feedback and edit the work?

3. Create Smart Defaults

Take inspiration from tools like Google Image FX:

• Offer pre-written prompts

• Include dropdown options

• Make experimentation safe and easy

Real-World Success Stories

Perplexity AI nails this with their research boards:

• Step-by-step guidance

• Progressive feature revelation

• Built-in documentation tools

• The deliverable is a research room the user may ask questions to with high quality sources that can be refined into a research paper, white paper, FAQ or other document.

Loom shows how AI can support entire workflows:

• Record → Transcribe → Edit → Task Creation

• Each step flows naturally into the next

• The deliverable is a fast, informative video edited to be easily accessible to the viewer with focus areas, chapters and a searchable transcript making it a high-quality deliverable that is fast to craft

Your Turn

As you design AI experiences, ask yourself:

1. What's the user's "magical deliverable"?

2. How can you guide them there step-by-step?

3. Where might they need help along the way?

Key Challenges From the Field

Our research with AI designers reveals several critical challenges:

1. Human-AI Collaboration: "When an AI agent spits out a massive deliverable, a human still has to check the work. How many of us like to be handed someone else's work and then are asked to take over?" This insight from a research participant highlights the importance of thoughtful, progressive hand-offs between AI and humans.

2. Control and Refinement: key to designer’s work is noticing the details, "small details matter in design, it is crucial to adjust bits and pieces of visual outputs." The ability to directly fine-tune AI-generated content is becoming increasingly important. It is frustrating to the user to view output as view-only and not be able to go in and directly edit.

3. Dynamic UI Evolution: There's growing excitement about "dynamic/generative UI" as a "highly underrated area of LLMs." The potential for an AI agent to learn a brand’s design language and then create personalized pages for an individual human user opens up new possibilities for adaptive interfaces. Imagine tailoring content to people’s learning styles and interests, these dynamic UI aren’t far off.

🔮 Looking Ahead

The future of AI interaction design is being written right now – by people like you. We're moving beyond simple chatbots to sophisticated AI agents that understand context, guide users through complex tasks, tailor interfaces to individuals and most importantly, work seamlessly with human collaborators.

There’s a reason 400,000 professionals read this daily.

Join The AI Report, trusted by 400,000+ professionals at Google, Microsoft, and OpenAI. Get daily insights, tools, and strategies to master practical AI skills that drive results.

Sign up now for free and work smarter, not harder.

What We’re Reading This Week

Lex Fridman podcast interview with Dario Amodei is the CEO of Anthropic, the company that created Claude. Amanda Askell is an AI researcher working on Claude's personality, and Chris Olah an AI researcher working on mechanistic interpretability. It is a long interview, but shows an intimate picture of how LLMs are made, the inner workings of Anthropic and the frontier of where autonomous, personalized agents are heading at Anthropic.

Suggestion Box

Forward it to a friend and have them signup here.

Until next time, keep innovating and stay curious!

Jason and Anastasia