Technology
AI as an Engine for UX Design
When designers master a tool,
they expand their ability.
Assembly Line, Photoshop, Artificial Intelligence
The first flow assembly line was initiated at the factory of Richard Garrett & Sons in the 19th century. This new technology became the industry standard for mass production factories. It resulted in many factory workers losing their jobs to the machine. It also resulted in new jobs for controlling and maintaining these machine. If we go back a little less in the past, the introduction of Photoshop around 1990 caused a, somewhat smaller, but similar movement. People were afraid that 'real' photography and art were being replaced by a technology that could easily layer images together and adjust them too (including control/cmd+z of course which was hardly possible in the physical world).
Thinking back on both revolutionary techniques, it's difficult to imagine living in a world without them. And this links us to the rise of artificial intelligence.
I know, I know, this is potentially going to become way bigger and reality-changing than anything before. But it's like discovering fire, it's dangerous but it will also change everything. So rather than avoiding/boycotting/hating it, start learning to embrace it. Remember: AI doesn't just create on the spot content and 'art', it works in so many less visible layers like operational tasks, automations and administrative chores.
Turn that 'Annoying Issue' into Artificial Intelligence
(See what I did there)
It's really not that complicated. Think of all the tedious tasks you need to complete on a daily basis before you can do the actual work.
Some examples:
-
Taking the minutes for meetings
-
Creating summaries of your brainstorm
-
Uploading invoices for tech tools that don't have auto-invoice
-
Sourcing for hideous stock photos
-
Collecting top layer competitive benchmarking data
-
Planning your meetings and follow up meetings
-
Preparing sprint-retrospectives
Now try to sum up how many hours a week, or a day those tasks take.
And now imagine having those hours back?
AI for UX Design
Time to focus a bit more on a specific discipline: UX Design. We all know the basics of AI, but what tools or techniques can you specifically utilise to become a truly better and more efficient UX Designer?
Let's grab the Product Design process to guide us. Research > Define > (Design > Validate > Prototype) > Build > QA test > Launch
For each step you can optimise your workflow with AI, without losing your integrity and authenticity as a Designer. Trust me.
It only works when you view AI as a tool and not as a cheat code to fast forward to the last level of the game, only to discover you lack all the powerful gear and XP to defeat the Final Boss ⚔️
Here you can read a bit on how Figma implemented AI technology throughout all of their tools.
Okay! Let's get into the nitty gritty and let me provide you with some neat techniques to start with.
Research Phase
No brainer: Collecting statistics, market numbers and features for competitive benchmarking. You can do this with whatever LLM works best for you (like ChatGPT, Claude, Gemini, etc). Using this output as a starting point I have discovered useful examples of companies that I didn't even knew existed. This especially came in handy because I was researching cross-border competitors and my cross-border knowledge was slightly limited for that specific subject.
Another great and easy technique is to use the same LLM to create interview/survey templates with. Make sure to always provide it with a solid prompt (there is also an AI for prompt optimisation, of course) and always re-read and adjust the output to completely fit your specific needs.
Define Phase
During the 'define' phase we collect all of our research and use it to emphasise and form the user need. To do this more efficiently, you can also use AI to feed it with all your research and create extensive summaries and reports.
Some important notes here: When feeding AI with any confidential information, make sure to have your AI tool run locally. This way it doesn't share your information with the HQ (and help the overlords take over the world with your data).
Another important thing is to always define the 'user need' yourself, based on the AI generated report. This is where your authentic skills comes in. Interpreting the data from your experience and skills will potentially result in a different conclusion than the AI's conclusion.
Design Phase
Time to turn those insights and problem statements into designs. Start with low fidelity designs and follow the Design Process correctly before creating those high fidelity designs (even if it's tempting).
For creating quick wireframes you can use a tool like uxpilot.ai or Figma Make AI, based on your prompt and additional information it will create some basic wireframe ideas. It's definitely a long shot from a definite design, but it can help you start somewhere relevant.
Validate & Prototype
I'm putting Validation & Prototyping together because validating your designs and assumptions works better with having some nice prototypes.
For creating prototypes you can utilise Figma AI to quickly add interactions and link frames together. Apart from AI you can use smart animate as a life-hack for simple animations and transitions.
As for validation, remember the interview and survey templates I mentioned earlier? If you combine those format with your fresh prototype and continuous testing tools like Lyssna or Maze, you'll be able to accelerate your testing workflow.
Build, QA test and Launch
I will leave using AI for the building part up to developers to educate you on that. But QA testing is something I do want to mention. People might not say it, but every Designer is thinking it: Everybody hates doing QA. I wonder is QA engineers even like doing it? If so, good for them 🔥 But what if you are a Designer in a small and fast-paced company? With no QA engineer present? Chances are that you need to test the implementations (partly) yourself. This is where automation agents come into view.
To truly be able to water-tightly test your product, unfortunately there will need to be manual work done. There are some expensive tools that can do QA for you, following ticket guidelines closely. But because behaviour is a big part of the UX, a human eye is needed. How else to know if the flow feels good? So building a QA automation agent that can pick up testing the generic functionalities and edge cases of flows already takes away a big part of the drag that is testing. Take a look at n8n, which is free! But make sure to run it on a local server.
Try and think of all other kinds of automation agents you can build! The possibilities are endless!
These are just a few very basic tips to get you started. There are so many techniques and tools available as we speak. We should treat this revolution (as ethically as possible of course) as one of the main quests, venturing into a new adventure.
As AI is developing quickly on a daily basis, this blog might already be obsolete next week. Well, then at least we have some nice 'historic' insights for anthropology's sake 🥴