UX design for AI: a practical guide to avoid common mistakes

As long as Artificial Intelligence doesn’t dominate the world, there will be a need for someone to design the products that use it. For some years now, that someone has often been me. I’ve gone through all the stages of the hype cycle: climbing from the hill of skepticism to the peak of inflated expectations, tumbling into the valley of disillusionment, and finally ascending toward enlightenment. Now, from this vantage point, I reflect on the experience gained while enjoying the view from the plateau of productivity.

Jokes aside, artificial intelligence, in its many forms and applications, has an exceptional ability to amplify the value of a digital product. However, the work needed to turn this potential into reality is far from straightforward.

It’s not just about technical issues like APIs or features—there’s no shortage of options, and we’ve already discussed them. The real challenge lies in product design, an often underestimated aspect that I’ve grappled with many times over the years. It’s a job that has taught me a lot.

Here, I’ll try to distill some ideas into 10 tips I’d give to anyone designing digital products in the AI era.

1. Let the magic shine

Even in the world of digital platforms, first impressions are everything. If my product integrates AI, I need to make it noticeable right from the start, in the first steps, in a simple and direct way. Making it noticeable doesn’t mean loudly proclaiming that I’ve released an extraordinary AI feature.

To convince my users, I need to immediately show them the tangible value that AI offers. They should experience an unexpected benefit, a moment of sudden satisfaction, at a speed that exceeds their expectations. Let them feel that acceleration.

My tip
An example? A user activates a subscription, expecting to go through a long series of steps. With AI-powered onboarding, the process becomes streamlined, and in just a few moments, the setup is complete and ready to use. Wow. That feeling of positive surprise, which we’ve all experienced at least once with the best AI products, is worth more than any marketing campaign.

2. But let them take the wheel  

Now that our users have grasped the potential of the AI™-powered platform, they must take the lead. Losing a sense of control can be intimidating or, more commonly, make users feel like they’re wasting time with a tool that imposes its own logic instead of adapting to their needs.

For this reason, after sparking that initial excitement, it’s crucial to reassure users that they are not being dragged along but are firmly in control. AI is just a copilot, an assistant that enhances the journey without ever taking control away.

My tip
Provide tools to customize the AI’s behavior and options to undo or adjust automated decisions. For anyone at the wheel, it’s essential to feel that the steering responds to their commands.

3. Help them choose the right path

How do you start a conversation or give instructions to an AI? After the initial experiments, users often find themselves stuck in front of language models (LLMs), unsure of what to ask. It’s the classic paradox of choice: if I can ask for anything, where do I begin?

Smart design eliminates this problem by avoiding leaving the user staring at a blank page. The interface welcomes them with well-thought-out examples that inspire and guide them toward a practical and effective use of the AI’s capabilities.

My tip
Provide your users with suggestions and templates to make writing prompts easier and reduce effort.

4. Make yourself understood 

Artificial intelligence is no longer just for experts. Today, the real challenge is making it accessible to everyone. If a product can be used by a wide audience, why complicate it with technical jargon or options that are incomprehensible to the average user?

The key lies in striking the right balance: the necessary technical complexity doesn’t have to disappear but should be hidden behind a clear and well-organized interface. Build a hierarchy of options that guides users step by step, focusing on the most common actions while simplifying access to advanced features for those who need them.

My tip
Use simple, user-friendly language. When necessary, employ design solutions like tooltips or collapsible menus (accordions) to guide users step by step while keeping the interface clean and organized.

5. Don’t hide the AI

It’s essential for users to immediately understand where and how AI operates within a digital product. This information should stand out at a glance, thanks to well-designed visual cues. This approach not only enhances the user experience by making AI-powered features more comprehensible but also strengthens trust and transparency.

Users know exactly where AI is making a difference, which reduces uncertainty and highlights the role of technology within the product. A well-designed interface leaves no room for doubt: AI is noticeable yet seamlessly integrated into the overall experience.

My tip
Incorporate distinctive elements to represent AI’s involvement: a subtle glow, a magic wand, a robot icon, or any symbol that harmoniously blends with the product’s style.

6. Humanize less, humanize smarter

It’s hard to resist the temptation to humanize AI, especially in conversational interfaces. But blurring the line between human and machine doesn’t add value to the tool; in fact, it introduces a layer of opacity that creates an unpleasant user experience. I’m not talking about dystopian scenarios like Her. It’s unlikely that your users will spend their nights chatting with an AI model until they fall in love with it. Instead, think of a more common and practical scenario, like an app for booking a dentist appointment. Imagine a user’s reaction upon discovering, after a brief exchange, that the friendly voice on the other end is actually artificial.

Now, compare this to a more transparent approach: at the start of the conversation, the user hears, “Hi, I’m your AI-powered assistant. I’m here to help you book an appointment. Let me know how I can assist.” This small detail changes everything. Clarity builds trust and allows users to engage with the system with realistic expectations.

There’s no need to disguise the machine as a human; what matters is designing an interaction that is effective, transparent, and focused on quickly solving the user’s problem.

My tip
Make it clear to your users that they are interacting with artificial intelligence.

7. Conversational? Sure, but don’t overdo it.

Why does AI always have to involve chatting or talking? The conversational interface was one of the key factors behind ChatGPT’s massive success. But you’re not building ChatGPT. Think twice—maybe even three times: is a chat really the best solution? Are you giving users an advantage, or are you forcing them into an unnatural commitment?

For many of us, providing direct input feels simpler and more natural than engaging in a conversation. It’s not just about cognitive effort; in many contexts, a chat can feel out of place. Instead of helping the user, it risks draining energy and slowing down an action that could have been handled more efficiently with a well-designed control panel or dedicated commands.

My tip
Conversational interfaces can be powerful, but they’re not a cure-all. So, choose wisely. Not everything has to be a chat. Let’s talk about it (kidding, of course).

8. Fail gracefully

No amount of training can stop AI from making mistakes. But what matters isn’t avoiding errors—it’s handling them gracefully. A well-managed failure won’t lead to negative judgment; it can even enhance the positive perception of the user experience.

This is a crucial aspect of design. When AI fails to meet a request, it should still provide a helpful response or suggest alternatives.

My tip
For instance, a chatbot that doesn’t understand a question could ask for clarification or offer other options, rather than returning a cold and frustrating error message.

9. Handle awkward scenarios

A well-trained model should intelligently handle unexpected inputs, recognizing its own limitations. If I ask a beauty website bot for a political opinion, the system should be able to steer the conversation back to its area of expertise without creating awkward situations. This type of handling requires careful planning for edge cases right from the early stages of development.

Anticipating such scenarios allows the model to respond naturally and professionally, avoiding unnecessary detours and keeping the conversation relevant to the user. When a bot knows how to manage out-of-context requests, it not only maintains the coherence of its role but also strengthens the perception of reliability and thoughtful product design.

My tip
During the design phase, always define edge cases to develop effective solutions that address potential issues.

10. Fancy sharing some feedback?

Unlike the advice given to a gentleman, a product should frequently ask if you enjoyed the experience. The system needs to be trained to address this proactively but not too bluntly, to avoid catching the user off guard—and potentially missing out on valuable feedback.

Questions like “Would you like to improve this conversation?” or “Was my response helpful?” help gather valuable data. The more responses your system collects, the better the algorithm can improve over time.

My tip
Think of feedback mechanisms like a thumbs-up, a heart, or other micro-interactions to make providing feedback a simple and intuitive experience.

To recap

Designing experiences for AI-based products means navigating a constantly evolving landscape, where technology and design must work together to unlock the potential of AI without ever losing sight of real user needs. The key to success lies not just in the power of algorithms but in the ability to translate them into interactions that are simple, clear, and meaningful.

From the initial excitement to the inevitable moment when something doesn’t go as planned, every aspect of the user experience demands attention, care, and foresight. AI isn’t an end in itself but a means to improve people’s lives and solve real problems. As designers, it’s our responsibility to build a bridge between what technology can do and what users truly want.

There are no universal solutions on this journey: every product is unique and requires a different balance between complexity and simplicity, between innovation and familiarity. But one thing remains constant: the value of a design that always puts the user first. With these principles, we can create products that not only work but inspire trust, satisfaction, and a genuine connection between people and technology.