User experiences in an AI-driven world

PICARD: Someone invented a hut. Someone invented a bow, who taught others, who taught their children, who built a stronger hut, built a better bow, who taught their children. Now, Nuria, suppose one of your cave dwelling ancestors could see you as you are today. What would she think?

NURIA: I don't know.

PICARD: Put yourself in her place. You see, she cannot kill a hornbuck at a great distance. You can. You have a power she lacks.

NURIA: Only because I have a bow.

PICARD: She's never seen a bow. It doesn't exist in her world. To you, it's a simple tool. To her, it's magic.

I’ve always loved it when something incomprehensible happens with technology. I’m thinking of when someone shows a phenomenal website design or PowerPoint transition, and the first thing I want to do is create something like it to wow others with.

It’s like, for every new style we experience, another line is drawn on our maps, and the routes of creativity available to us expand. User interfaces are an area where this is readily apparent.

The fundamentals of a good user interface are:

  1. User-centred design: Focus on understanding and meeting user needs, goals, and preferences throughout the design process.

  2. Simplicity and clarity: Create intuitive interfaces with clear navigation and minimise cognitive load for users.

  3. Consistency: Maintain uniform design elements, interactions, and behaviours across the product to enhance learnability and usability.

The advent of generative AI has given us an entirely new paradigm, allowing us to stretch these principles to their maximum potential.

The question is: when AI is no longer the facilitator but the driver, how do we need to consume information?

Rules

Up until this point in time, our options were limited to the structured tools available to us. Our imagination was limited to the constraints of our technology.

Think about the intuition that guides tasks such as:

  • Prioritise this field because most users will need that…

  • Lay out this Power BI report/PowerPoint slide deck with critical elements on that top left so that…

  • Add this to a sub-menu, because this small user group only needs to check it every now and then…

There’s a certain stagnancy to it all. A “set and forget” approach to configuration.

Of course, this isn’t doing the gradual development of AI enough credit - algorithms have been around for a while. Think of YouTube, Facebook, Instagram, TikTok etc. I think of these as using complex algorithms to present information to users, but in a kind of templated form. A post is still a post, a video still a video, an ad still an ad, and so on. They are scored as relevant to you based on a collation of data about you, sure, but it’s still this pre-packaged, on-rails, factory made experience.

Magic

With models that can carry context, perform functions, understand unstructured information and operate across multiple modes in parallel, the game changes. Consider, if the system is truly capable of these things, you can start asking questions in design such as:

  • What inputs are needed/wanted from the user in this context?

  • Why is this particular user likely to be interested in this information at this point in time?

  • How is my structure blocking the system from presenting the right information?

I’m thinking of a helpful assistant saying, “I’ve sorted out the rest, but I thought I’d bring this to your attention”, as if you are the President receiving a daily briefing. We’re in a time where we can start to embrace the concept of ‘invisible interfaces’ - minimalist, intuitive, predictive interfaces which show what is needed, when it is needed, in ways that delight the user.

In saying that, the core principles are the same - user-centred, simple and consistent. We merely need to build on these:

  1. User centred: with hyper-personalised AI, more than ever.

  2. Simple: with high contextual knowledge, impeccably intuitive.

  3. Consistent: in design language, yes. In what appears, no.

AI can plan out what you see based on what you need and the inputs it needs to help you better. In a way, it’s taking you back into reality whilst it deals with the digital realm, untangling you.

Example - skiing technique app

Applying these capabilities to a hypothetical skiing technique application, let’s compare a traditional and AI-driven app:

App Comparison

Capability Traditional App AI-Driven App
Sign-up experience Sign-up with Google/Facebook/Email Same + enable read of profile so that AI can learn about your capability.
Goal-Setting Complete a tab-by-tab dialogue to set goals/preferences AI uses profile information + conversational or media-driven analysis to derive skill level and goals
Feedback and Improvement Only possible based on tags/categories filtered via submenus Direct feedback based on individual skiing technique
Training Content Library Complete video/text library Relevant library, including generative content, tailored to the user.

From this example it’s clear that the capabilities of the app and the UI needs are completely distinct. We’re moving away from a smorgasbord to a refined table d'hôte experience.

Principles for a new approach

In a way, we are merely appending the toolkit we’ve been provided onto our existing best practices. The fundamentals will continue to apply for the sake of intuitiveness and accessibility. What we have here though is more colours for our palette. How do we best apply these colours?

For that I’d posit the following principles:

  1. Always consider how you are freeing or constraining the user. Both freedom and constraint can be good or bad in different circumstances. A free text field can have complete freedom but may introduce effort or disharmony to the app.

  2. Consider what information the AI will have ‘on-hand’ at any given time. This is integral to making the experience as seamless as possible. AI is amazing, yet logical: you want it to build up off something.

  3. Delight the user. Give the AI functions that will enable it to surprise the user with rich insights and immersion. In putting together the blueprint, consider how you can arm the AI to do this.

Happy designing!

Next
Next

Exploring the business applications of multi-modality