Designing AI Experiences
Over the past few years, I’ve had the opportunity to work with teams leading ambitious AI initiatives focused on solving real user problems. I’ve seen firsthand how thoughtful AI experiences can create meaningful impact for both users and businesses. As a designer, I find this space incredibly rewarding and exciting.
Back in March, I attended the "Designing AI Experiences" training course by Nielsen Norman Group, where I learned, among many other things, that successful AI experiences are not the ones that feel the most futuristic, they are the ones that help users feel informed, supported, and in control.
AI products are everywhere now, but many still feel confusing, unreliable, or hard to trust. The issue is often not the AI itself, but the experience around it.
Designing good AI experiences is less about adding chat interfaces and more about helping users understand:
What the AI can do
What it cannot do
How much they should trust it
How they stay in control
One of the strongest ideas from the course was how AI changes traditional UX rules. Unlike normal software, AI is probabilistic. The same prompt can produce different results, which means designers must design for uncertainty, supervision, and recovery from failure.
Because of this, good AI products focus on trust, transparency, and usability.
TRUST
Establishing user trust in AI products is essential for engagement and adoption. To help ensure the AI experiences we design actually match users expectations, we could use the CREED framework:
Competence: Can it do the job?
Reliability: Is it consistent?
Explainability: Do users understand it?
Ethics: Does it act responsibly?
Design: Is it built with users in mind?
The framework shifts the conversation away from “how smart is the AI?” toward “how trustworthy is the experience?”
TRANSPARENCY
The course also emphasised the PROVE Framework for AI Tool Evaluation:
Problem Alignment: Does it solve a real problem?
Risk Assessment: Is it safe for your data?
Output Quality: Is the work good enough?
Velocity: Does it save time?
Experience: How much friction does it add?
P and R are gates. If a tool doesn't solve a real, high-frequency problem or if the tool is not safe to use with your data, stop there. The tool is not worth adopting.
O, V, and E are trade-offs, that should be scored and weighted based on your context.
USABILITY
Several practical principles stood out around User Agency and Control:
Allow users to opt-in and opt-out of AI processes
Expose system controls
Offer multiple output options
Allow users to act on existing outputs
Allow users to stop a process, especially if it costs money.
Key Takeaways for Effective AI Design:
AI usability matters more than novelty
AI strategy must align with business objectives
Traditional UX research methods still work for most AI systems
Technical constraints directly impact what features are feasible
Trust requires transparency and explainability
UI of AI: use common AI components appropriately for familiarity
The biggest lesson is simple: successful AI experiences are not the ones that feel the most futuristic. They are the ones that help users feel informed, supported, and in control.
AI Product Design
If you are looking to implement effective user interfaces for your AI products or features and are not sure where to start, Balbuena Design can help. We typically support customers across Branding -> Website -> AI Product -> Design System.
If you’re building AI experiences and want to create products that users can actually trust and use confidently, feel free to get in touch.