Let's Talk

Meaningful AI experiences: 4 foundational tips

The world of AI and end-user experiences is an emerging one. At Topp, we’ve been designing & prototyping a wide variety of AI-driven experiences for several years now. The following are some foundational perspectives when getting started with designing for AI:

1 ) To spot opportunities, understand your tech constraints 

2 ) Design flexible, modular structures

3 ) Build for continued service & product evolution

4 ) Set the expectation bar low, know your goal

These tips are meant to unlock some common issues and bottlenecks when getting started with AIs beyond the merely conceptual level. We’ll expand on these in just a moment.

A bit of background

When we read about AI in the media, it’s easy to think about super intelligences taking over the world, and whether it’s productivity, political conflict, or the environment - in the end, humans are simply optimised out of the equation.

While there are ethical topics that are important to consider when approaching the design of AIs (or systems supported by AIs) there are also great opportunities for creating better experiences and solutions for people in ways we haven’t seen before. These leaps will require getting our hands dirty.

For clarification, when we talk about AI we’re not only talking about the systems that mimic our own intelligence, but instead an encompassing collection of technologies that can go by many names: machine learning (ML), natural language processing (NLP), deep networks, and more. Essentially, AI provides us with the opportunity to create systems, that are build for customization and growth over time.

There are many applications of AI, and today the application of this broad area of technology is most mature in areas like logistics or medical diagnosis. There are certainly aspects of the end user experience in almost all of these, but let’s focus in on the more nascent area of user-facing digital services and products, typically found in the consumer space.

As users, we are most familiar with things like chatbots, recommendation services, or digital assistants.

AI can be considered an added ability to build new services and products around or add new value to existing ones. We believe that the scale and impact of these abilities could be along the scale of how the Internet as a whole impacted opportunities in the 1990s.  

We’ve been exploring the emergent area of AI and UX at Topp with our clients and would like to share some insights into some early principles to keep in mind when designing intelligent digital experiences already today.

Okay, let’s dig into the first handful of tips.

1 ) Understand your tech constraints to spot opportunities

From a purely rudimentary perspective, the backend of an AI is built by combining algorithms and data. By understanding what data we have and what algorithms we can use, we can gain insight in how to transform these building blocks into nuanced, new experiences. Keep in mind that it’s very rare to engineer all the technology from scratch, purpose built for the experience we want to design.

Approachable options include designing on top of existing AI platforms, such as what Amazon Alexa or IBM Watson provide. In this case, each platform provides a collection of APIs that come with their own inherent strengths and weaknesses. Unlike working with, for example, a particular CSS or javascript framework over another (which are relatively similar in their core capabilities), AI platforms vary at extremes in terms of what capabilities they might provide.

Knowing which data is available to you, and grokking the underlying engineering direction of the AI platform will enable you to work with the most powerful aspects of the AI, and even inform the direction of the platform if you happen to be working directly within an organisation building such a platform.

2 ) Design flexible, modular structures

Typically in most applications today we’ll have a good sense for the array of content and functionality that could populate widgets / boxes / modules / etc. With AI driven experiences a modular approach is particularly important because we need it to respond to to personalisation, context, and take on unanticipated content and interactions.

This modularity also applies to non-visual modalities like voice. What type of content will fit in a reply? What are the reply types? Basic information, further inquisition & clarification requests, third party content, conversational experiences? What are the boundaries of the questions you can ask? How will it handle errors? Creating a clear framework for users becomes imperative, as intent recognition is far from perfect today.

Essentially, today’s dynamic layout and responsive approaches are moving to a far more diverse grammar framework. Whether visual, tactile, ambient, or screen-based consider what the grammar of your system should be.

3 ) Build for continued service & product evolution

Building with AI, does not mean building Homo Sapien. When your AI-driven experience reaches the user, its in its early stage of evolution. It might perform unexpected ‘mistakes’ and perhaps even seem stupid. The nature of intelligent services and products is to learn and develop over time - not be 100% complete out of the box.

Vastly different from a traditional product - an intelligent service could not only be contextual and personalised to how it learns from the end user, it will also be vastly different over the course of time. Day 1 and day 100 will likely not look the same.

We need to envision futures of our product together, so we can have some measure of control. But at the same time, we need to enable basic, but still delightful, user-experiences from day 1 where teaching the system comes naturally.

4 ) Set the expectation bar low, know your goal

Maybe we’ve set the bar low with the evolutionary comment, but still you need to set the appropriate level of expectation for the end-user as well.

Often, chatbots are a particularly tricky instance of AI, where they are trying to simulate some human-level intelligence, but the follow-through comes up short. You’ll need to figure out the right balance of actual smarts and instructive copy to include.

But there are other ways to set expectations as well. Each type of AI for the user can imply what they should expect from it. Often, passive intelligent experiences which avoid all conversational or embodied framings go far beyond expectations.

Discover Weekly by Spotify, a playlist tailored to your tastes in certainly driven by an intelligent platform. And overall, it has been driving significant engagement in music discovery. However, this platform does not advertise the AI-ness of it. It simply works, and continues to get better.

Just because you’ve got a new gadget in your toolbox doesn’t mean everyone needs to know it’s an AI. Ensure that the application of the AI matches what you’re really trying to achieve.

Design better experiences, not just intelligent technology

No matter what level of intelligent experience you are creating, always have in mind that great experiences are designed for human beings.

Our AI systems should not be a result of how far technology can be pushed to resemble human beings, but how far they can be pushed to enable meaningful experiences for us.

Rikke Otte, and Emil Wasberger from Topp recently hosted a workshop at the Copenhagen Techfestival where they explored these topics hands-on with dozens of designers, engineers, and product owners. As this is an emerging area, these insights will continue to evolve. Please stay-tuned for updates and don’t hesitate to reach out for a direct conversation.






The rich text element allows you to create and format headings, paragraphs, blockquotes, images, and video all in one place instead of having to add and format them individually. Just double-click and easily create content.

  • hej
  • dig
I am a block quote

Thank you! Your download should begin shortly.
Oops! Something went wrong while submitting the form
Thank you! Your download should begin shortly.
Oops! Something went wrong while submitting the form
When you enter your email address you become a friend of Topp. A group we send occasional emails to. We won't be offended if you unsubscribe, but love if you stuck around.

Rikke Otte + James Haliburton

Visual Designer | CEO

Thank you! Your submission has been received!