The iPhone wasn’t the first touchscreen phone. But it quickly displaced the Blackberry as the corporate workhorse because it was a design masterpiece. The iPhone’s smooth digital buttons and uncluttered, user-centered interface made it intuitive, easy, and fun to use.

The technology inside the iPhone was powerful. But it was the design and user experience that truly connected that power to people’s needs.

Fast forward to today, and product designers are facing a new challenge as companies adopt Generative AI (GenAI) — another game-changing technology whose success will depend heavily on how easy and intuitive it is to use. GenAI presents a more complicated challenge as well. Because corporate users aren’t just tapping a button, they’re tapping an intelligence.

For employees, this intelligence may take the shape of an AI assistant — an application that can free them from mundane, time-consuming tasks while improving how they do research, visualize data, write emails, and much more.

For employers, this ability to improve speed, accuracy, and productivity will increasingly mean the difference between a business that races ahead and one that falls behind.

For myself, as an AI product designer, it’s an incredibly exciting time to be helping both employees and businesses access the power of GenAI as efficiently as possible.

Interacting with AI

In many ways, interacting with GenAI is easier than with other technologies. GenAI “understands” human language after all. Yet because human language is messy, it also raises unique design challenges.

On a computer, you click “Email” and you access email. It’s a deterministic action that takes you where you want to go 100% of the time.

But using an AI assistant is more open-ended. Instead of just accessing email, you might want to summarize some emails and reply to others. Your interactions are more iterative. There’s more back-and-forth between user and AI, which means requests may succeed, say, 95% of the time instead of all the time.

Improving these interactions with AI is an ongoing challenge for a product designer, but we’ve come a very long way in a short time.

When GenAI exploded onto the scene in late 2022, the wonder at what it got right was often overshadowed by what it got wrong. These mistakes, or hallucinations, remain a very serious concern for businesses today.

So every product designer working on GenAI needs to spend a lot of time thinking about establishing trust. It starts with increasing the accuracy of AI responses. Connecting AI models and real-world data with systems like retrieval-augmented generation (RAG) helps mitigate hallucinations and increase trust.

But it isn’t enough to just improve the accuracy of AI. We also need to design interactions that give users a clear window into the original sources of information — and we need to design them in ways that are quick and unobtrusive.

How do we do this without forcing people into a detour of laborious fact-checking? One solution allows the user to hover a cursor over a word provided by the AI assistant to instantly reveal multiple citation links to source material like studies, documents, or publications like Scientific American.

Command R / R+ is uniquely trained to generate accurate citations.

Overcoming perceptions

We’ve made such quick progress improving GenAI that one of our challenges as designers is convincing both employers and their employees how easy and valuable it is to use these systems today. We recently did some research into perceptions of GenAI and it revealed that many people still think the technology is too technical. That it’s more suited for technical experts or developers. For a regular employee, the underlying concepts and value proposition of an AI assistant must be easy to understand right from the start and relevant to their daily work.

Anyone can use GenAI today, and that’s partly due to steadily improving product design. Our goal as product designers is to spark a Eureka moment in everyone who tries an AI assistant — to recreate those magical moments when people used a computer mouse or touchscreen for the first time.

If you have 5, 15, or 50 documents to process, an AI assistant provides an accurate and concise summary in minutes instead of the hours or days it would usually take. If you need to process a blizzard of data, an AI assistant transforms that data into a table or graph that instantly makes sense.

And this is just the beginning. For example, why have just one AI assistant when you can have several that are specifically designed for different tasks? Or perhaps you need an AI assistant that’s more proactive. Instead of merely responding to a request, it might suggest a request.

For a designer, this creates some exciting opportunities, while stirring up a flurry of new challenges. Designing a clear line between an AI assistant that is proactively helpful and one that is annoyingly intrusive is just one of the things that will keep us up at night in the months and years to come.

Simple design features like easy-to-spot calls to action and having the ability to search through your chat history and tag chats, can make the world of difference with the end user. We are constantly experimenting with new interactions to aid uptake and deliver more value.

Looking ahead

Interactions with AI assistants are expanding to visual and voice interactions as well. For example, you may want to have your AI assistant analyze images for meaning. Or you might prefer to speak to an AI assistant because you can convey more nuance verbally.

These possibilities and more make designing for AI as dynamic and fast-moving as AI itself. And this means it’s imperative that we work intimately with both the technology and end users to ensure our AI solutions are always relevant and effective.

Getting GenAI to work effectively can be challenging. But it’s also incredibly rewarding as we continue to create AI assistants and other AI solutions that are rapidly transforming how the world lives and works.

Building enterprise AI agents

A comprehensive blueprint for developing scalable AI agents in regulated industries.

Download the eBook

Designing AI assistants that workers will love to use