Design's New Frontier: Navigating the AI Wave
Exploring how we will integrate AI into the solutions we design, and what you as a designer can do today to better prepare for an AI-powered tomorrow
We are witnessing an explosion in new AI tools. Every day brings a new research breakthrough, product launch, or high-profile demonstration from giants like Google, Microsoft, Meta, as well as smaller teams running at incredible speeds. This puts incredible new capabilities at our fingertips. Capabilities that used to require a team of domain experts can now be accessed by anyone, from tiny bootstrapped startups to enterprise giants.
From turbocharging productivity to transforming the creative process, this dramatic acceleration of progress may be remembered as a pivotal moment in tech history. Designers are uniquely positioned to be on the frontlines, shaping how these emerging technologies are integrated into our lives.
To achieve our potential, we must pay close attention to this revolution and understand the new opportunities and constraints it brings. We need to consider how they can influence the solutions we design.
Aren't these just chatbots?
It's tempting to dismiss these new AI tools as just being the latest iteration on the “everything is a chatbot” fad of 5 years ago. The hype cycle has come and gone before.
But while healthy skepticism is important, as designers we must also be open to how emerging technology can enable new possibilities. Used thoughtfully, models like Midjourney for generating images and ChatGPT for conversing in natural language point to how AI could soon transform human creativity and communication.
Moreover, these tools are improving at breakneck speed. For example, in just one year, ChatGPT evolved from a research experiment to a full-blown, enterprise-grade development platform that can "hear," "see," and "speak." Similarly, Midjourney went from a cool toy to a valuable tool for marketing teams and graphic designers
These models are already producing amazing results by themselves, but as a designer, I’m most excited about the possibilities of integrating them into the tools we build. Let’s explore.
How can we integrate AI into our designs?
With these changes happening so fast, let’s look at some thoughtful ways these tools are being integrated into existing tools.
Need to better understand context
With powerful AI capabilities widely available, it's crucial to deeply understand the user's context. These models excel at processing and reacting to user inputs in real-time, making product integration into daily routines more seamless and intuitive. Understanding the user's context helps designers anticipate needs and behaviors, ensuring relevant features that improve the user experience in unexpected ways.
Having a deep understanding of how are people currently solving a problem makes sure we approach integrating AI where it can make an impact on a real, hard problem, as opposed to just finding a place where we can integrate it just because it’s cool and novel, creating a superficial experience.
Example: Loom recognizes the importance of ensuring that the videos created on their platform are watched, to eliminate unnecessary meetings. Understanding that few users take the time to provide meaningful names and descriptions for their videos (based on my observations of the tool's adoption in multiple companies), Loom identified an opportunity. They now automatically generate a transcript for the video and create a name and description, providing viewers with essential context when the video is shared.
More signal, less noise
AI models are excellent at processing large amounts of data and serving the right information when prompted. By combining large data sets with small prompts from the current context, we can help users achieve optimal results without overwhelming them with endless dropdowns and modals. "Good design is as little design as possible."
Example: Shopify sidekick empowers business owners by giving them an expert agent to ask questions, which are answered based on the knowledge Shopify collected seeing thousands of businesses succeed and fail, it’s even connected to your account and can perform actions on your behalf based on the recommendations. Letting you get advice on demand from an AI agent that can act as a Shopify Expert
Easier onboarding
AI models allow us to analyze user input and create richer results, making the onboarding process more intuitive. Users can start creating amazing results as soon as they begin using a product, gradually exposing them to more advanced capabilities as they become comfortable with the tool.
Example: If you have an idea for an image or a graphic element that you want to create, you can just describe it with plain language to Midjourney, and it would immediately create stunning results. Over time, you can start using more advanced capabilities like how to get outputs in a specific style and size, or how to iterate on just parts of an image.
Even if we got the user to understand how the tool works, we need to help them get started from scratch when there are so many different options - this is usually called “the blank canvas problem”.
By processing inputs like natural language and images, these models can allow users to start creating amazing results as soon as they start using your product, and .
Example: Figma has a great online collaborative Whiteboard called Figjam. When you get started, they offer a ton of useful templates created by their team and the community to get you over this problem, however, that too can be overwhelming and you may not find what you need. They have recently integrated AI capabilities to help get you started by describing what you’re trying to create with plain language, and then generating that for you - it’s not going to get the job done on its own, but it’s a great way to get over the blank canvas problem while showcasing the different possible building blocks the tool uses.
An experience that evolves over time
As users interact with our tools, we can adapt the interface based on what we learn about them, the data they create, and their current goals. AI models make it easier to customize the experience and its outputs based on user inputs, creating a tailored experience "on the fly" without creating multiple versions of each view in our app. However, we must consider the increased unpredictability in user input and ensure users are comfortable with receiving outputs different from their expectations.
Example: Duolingo MAX is the company’s latest release, leveraging GPT4 to take what they know about your current progress, and simulate a conversation with a character in a real-life scenario. As you progress, so does the level of conversation. Additionally, Duolingo collects information from all the conversations its users are having with the product, to improve the effectiveness of the conversation for each user.
What can we do to better prepare?
We've discussed why these AI advances are so compelling and how they might shape the way we design products in the near future. To wrap up, let's explore some steps we as designers can take today to best prepare for an AI-powered tomorrow.
While the interfaces we design may look very different in a few years, the fundamentals of product design remain relevant, if not more so than before. Classic principles like NNG's 10 design heuristics still hold value and can help us achieve better results.
In addition, there are design tools that will become even more useful in this context, encouraging us to think about the broader context of our solutions before diving into UI details. Story mapping by James Buckhouse and Breadboarding as introduced by Ryan Singer can help us think abstractly about functions and affordances without getting too caught up in UI patterns and implementation details.
To ensure we leverage AI effectively, we must be selective in identifying problems worth pursuing and avoid building products or features with AI just because we can. Techniques like "Opportunity Discovery" and Jobs to be Done can help us focus on real problems and user needs.
Lastly, it's crucial to immerse ourselves in the existing set of AI tools, staying informed about their evolution and how other companies are using them. This knowledge helps us identify opportunities to build even better solutions using AI models.
I hope I was able to get you excited about how our craft might be impacted by this wave of innovation. I know I am.