AI Needs Designers to Help Build It, Not Just Use It

Most writing and thinking by designers about AI today is about how to integrate AI into the process. For example, using AI to analyse sentiment and research, to support ideation, to automate parts of the creative process, or to predict how different design solutions might perform in real-world use.¹

This is valuable work, but it’s too small. AI is not just a helpful tool for designers. We desperately need designers, and Design Thinking, to play leading role in defining and executing the vision for AI in our world.

Without Design Thinking, we risk having a lot of very clever technologies that don’t solve real world problems very well.

That’s exactly what’s happening now. Gartner predicted that by the end of 2025, 30% of GenAI projects would be abandoned. Hugging Face CEO Clem Delangue speaks of an LLM bubble. Google CEO Sundar Pichai reminds us not to trust everything AI tells us.²

These are, partly, design problems. Problems of clever technology where the use cases aren’t working. Problems of too many engines and not enough steering wheels.

Design Thinking teaches us to understand people’s needs deeply, building empathy, then iteratively explore and test solutions to these.

It’s the first bit that’s critical here: understanding people’s needs deeply. We’re not talking about reading a research paper (although that might be helpful). We’re talking about putting yourself in their shoes. Literally.

If you want to design an AI for a busy A&E team, you stand in the ward at 3am. You watch them trying to type notes with one hand while holding a patient’s hand with the other. You talk to lots of people, and watch them going about their tasks, to validate and deepen your knowledge. You learn perhaps that accuracy isn’t the only need - speed and cognitive load are critical too. Perhaps a voice-based model is going to be far safer than a textbox interface. We can test that to find out.

This is what provides the unique insight we can use to design products that work better for people.

A great example is ChatGPT’s initial launch. The underlying technology had been around for several years.³ It was only when OpenAI had the insight to wrap the technology in a chat interface with conversational memory - a user experience innovation - that it became the fastest-growing consumer application in history.

There are zillions more design and user experience innovations in AI just waiting to be developed. Some will be game changing, some will be incremental. But they will all make a difference, and we all need them to get modern AI from a place where it’s a sort of wild, scary, unregulatable beast, into something that is safe, productive and profitable in our organisations and society.

In AI, everything needs to be designed:

• The goals of the models we are creating
• The datasets we are using to create them
• The training and evaluation processes
• The performance metrics that we want to capture and how we communicate these to different groups of people
• The systems we use to deploy them
• The APIs and integration experiences that other builders will use to work with the models
• The experiences we want end users to have when they use our models
• The user interfaces that enable people to use, control and understand models - Explainable AI
• The controls we put around AI to ensure we are using it safely, ethically, responsibly
• The representation of uncertainty - how does a probabilistic, rather than deterministic, system communicate with us?

These things need design because they all have different groups of users and stakeholders, with different needs and priorities. The design task is to find elegant solutions that solve these complex puzzles.

Of course, it’s not just designers who can do design. Everyone working in AI can bring design processes to their work, to make it better, just as everyone can bring ethical thinking to make things better.

But there is an amazing, pressing need and opportunity here for design specialists, who have spent their careers working out how to solve difficult human problems in a certain way, to play a pivotal role in the future of AI.

We need to get in the driving seat to help build the steering wheels.