Image by author

There’s no question that AI is revolutionizing design — hell; it’s already an indispensable tool that I constantly use to help me condense the time from idea to action in my professional and personal life.

Now, with the advent of GPT-4, the just-released multi-modal product from Open AI, the promise is even better results using both text and images, and eventually, audio and video.

Source: Open AI

Exciting, right? Yes, and while it theoretically changes everything, it can’t change one crucial factor: the human element.

I’m not talking about emulating humans — if you ask ChatGPT itself if it’s going to replace people, it tells you this:

Exactly! Excellent design solves uniquely human problems, and it takes people like you and me to figure out what those challenges are and who needs help surmounting them.

This is something we’re never going to be able to pawn off on the machines, so in this article, I’m going to give you a good example of how design goes wrong and how to fix it by asking better questions in your design process. After all, the biggest flaws in the machine are in our hands, from a lack of prep, discovery, and understanding of the audience and problem to ill-defined processes. Once we clarify the things we humans must do better, we can get to best AI practices and the art of writing better prompts to build your business and boost your bottom line.

For starters, in design, form should always follow function, not the other way around. As Steve Jobs once said:

He also said:

This is something AI absolutely can’t do for you — it can do it with you, but you are the leader. Grokking is as much about feeling into a problem as analyzing and strategizing your way to a solution.

For example, I’m currently working on an enterprise supply chain product with my team. At a recent meeting with a mix of stakeholders, from our internal designers to the client, I realized nobody at the table had a clear idea of how we’d go about gathering audience information. Who would facilitate interviews with customers and stakeholders? What exactly would we ask? How would we synthesize that information and turn it into a successful design brief?

We did know the basics of the problem we were solving, which had to do with supply chain operations at the managerial level. The core pain point had to do with manual administrative work. But beyond that, we needed help filling in the blanks to paint a better picture of our core target audience members.

So, as an experiment, I asked myself, what if we could add another team member to the team that could help us? Someone who came with zero recruiting fees and was available to start immediately?

A name came up immediately: ChatGPT.

More and more, this is the top name on anyone’s list to kick off the research and prep phase of a project. But don’t be like the rest of the lemmings that write lazy prompts and think that AI can fill in the blanks by turning billions of data points into something that’s on point for your audience. It can’t.

If I had a hundred bucks for every time someone asked me how to write chatGPT prompts recently, I’d be like a social media influencer with millions of followers, rolling in cash from sponsored posts and brand deals. But seriously, like most things in life, the answer is usually in the question.

The framework that has worked for me in solving most problems is this:

In the case of the enterprise supply chain solution, here’s how I used that template to create an avatar of our ideal customer:

Here’s what ChapGPT came up with:

Having talked to real operations managers in this role, I can say this is pretty damn accurate.

Next, we wanted to drill down into the problem so we could really understand what keeps our audience up at night. So, I primed the ask:

The AI gave me a list of pain points and descriptions:

Having done countless real interviews throughout my career, I can tell you that the information that ChatGPT gave me back was about 85% accurate and very actionable.

From here, I asked the AI to determine a list of features for a product that would support our ideal customer and help alleviate their pain points and challenges.

Finally, I engineered a few more prompts that shaped the inputs necessary for a solid draft of a design brief, and got this plan of action from ChatCPT:

Research The design and research team will conduct stakeholder interviews with operations and fleet managers to identify pain points and features needed in a logistics automation platform. The following is a list of stakeholders to be interviewed:

Questions The following are some questions that can provide insightful pain points and features that customers need to speed up their job:

Pain Points Based on the stakeholder interviews, the design and research team identified the following pain points for operations and fleet managers:

Features Based on the pain points identified, the design and research team recommends the following features for the logistics automation platform:

Design Considerations When designing the logistics automation platform, the design and research team should consider the following:

By considering these design factors, the design and research team can create a logistics automation platform that addresses the pain points of operations and fleet managers while delivering a positive user experience.

Voila! We now had enough information to properly brief our cross-functional team — simply by knowing the problems we’re solving and the audience we’re solving them for.

As a design leader with over 17 years of experience, I can confidently say that this output provided much more context and detail than I would typically receive back when I was an individual contributor working on design projects. This plan of action allowed us to quickly map out all the different ingredients that go into creating a successful prototype and a dynamic product lifecycle.

And by “us,” I mean my team and our newest member, ChatGPT, who helped envision an app for our designers to prototype.

And this is just the beginning.

Countless AI tools are starting to emerge that will change the face of the design process for good, from converting sketches and drawings to creating high-fidelity prototypes.

Screenshot from: https://uizard.io/wireframing/

Big tech companies are on a talent and acquisition spree, democratizing complex machine learning and AI platforms. With products like Lobe and OpenAI (maker of ChatGPT), UX designers can do things in a day that used to take months, millions of dollars in computing power, and a team of PhD mathematicians and engineers.

Remember:

And by using AI, I mean asking better questions in the relentless pursuit of clear communication. As we move into this new era of multimodal language models that leading players are dedicated to (i.e., Facebook, Google, Microsoft), we can’t skip over the fundamentals. Understanding how our own processes and communication affect the design of outputs and enhances the impact of outcomes for real humans is crucial.

AI can help move things along faster, but it can’t fix the underlying problem that too many design leaders have: not approaching problems with curiosity to get better at documenting hypotheses, prototyping and testing those hypotheses, and then inviting others to be a part of this process.

What all of this evolution in technology is doing is its own kind of prompt — it’s nudging us to become better UX professionals. If you want to push your business to greater success, it’s time to stop being mechanical about your processes. The better you are at asking questions and articulating problems, the more likely it is the products you design with AI’s help will better serve humanity — and your business.