UX for AI, AI for UX: 20 Tips to Design for AI & Let AI Design for You

The good news and the bad news about AI

AI is here to stay, and it’s only going to get better and more sophisticated. UX teams will get leaner over time, but there is still a need for thoughtful and ethical UX research and design considerations when it comes to designing AI apps.

At the same time, AI is not the same as automation ... .and a lot of companies are still very behind with basic digital automation, particularly industries like healthcare and finance where there are a lot of regulatory concerns to contend with. Read more about 2024 UX trends here.

Download our AI Guide: 25 AI Prompts for UX

10 Tips:  How to apply UX methods for better,  ethical AI design

1. Get very specific about the HUMAN problem you are trying to solve, rather than nerding out about the AI tech. Don’t use AI just for the sake of using AI. The UX design process still applies to identifying user and customer problems, and then prioritizing the RIGHT problem to solve. This can take time applying UX research methods and conducting discovery user research. Did you know that over 40% of startups fail because they build something that nobody needs? The same holds true with startups and companies experimenting with AI-based products and services. It’s important to validate HUMAN problems before being blinded by flashy, new tech and committing money and time into building something nobody wanted or needed in the first place. 

2. Ask: will adding AI capabilities help with product differentiation? Or can a more low-effort automated solution work just as well? More personalized solutions don’t always have to be powered by AI. More automated solutions don’t always have to be powered by AI.

For AI to add differentiation, ask:

• How would a human expert perform the task?

• How would you give the human expert feedback so they improve next time?

• What assumptions would users want to make of this expert?

3.Set expectations about what your AI product can and cannot do. Be realistic about your product or service’s capabilities and what it can and cannot provide for users so that trust is built. AI is not a magic solution that can do everything well.

4. Explain the algorithm process in terms of the results that users receive. Explain what data sources the algorithm is drawing from and its process to synthesize a result or answer for the user. In math classes, we always say show your work! In this case, explain the algorithm process to remain transparent and co-create improvements with users as the product becomes more sophisticated.

5.Communicate and differentiate  more confident results against less confident results. Where necessary showcase confidence levels of results and what results have the highest confidence level. Present secondary/tertiary options with less confidence in an altered manner to differentiate between more confident and less confident results. 

6. Decision-making should be human. AI doesn’t necessarily need to replace human contributions, but should simplify and enhance what people are capable of doing in their jobs. Design AI to co-create with human users and take feedback from users. 

7. Let AI evolve as users evolve over time. AI should be adaptable to changing user preferences and even changing societal preferences. Yet at the same time, it should not force a specific team or person’s world views on others. 

8. Personalized experiences require that AI adopts some of the user's biases or preferences. However, this can also be limiting in terms of helping users get out of their comfort zone.

9. Co-create AI products and services with a diverse team to avoid biases. AI is only as good as the data inputs it receives. A diverse team will ensure that all possible data inputs and considerations are thought through as the algorithm is being developed. UX research is crucial at this stage.

10. Avoid collecting user data, or let users control how much data is collected. Avoid collecting user data because users have a right to own their data. However, if it is necessary to collect data, explain why data is collected and how it is used. Give users control over how much data they can choose to let the AI algorithm collect. 

10 Tips: How to apply AI to UX methods

1.  AI must be prompted by experienced UX professionals so that the best results can be delivered and synthesized. UX professionals can select the right AI recommendations based on the context of the problem at hand and what’s needed at the moment. That discernment cannot be replaced by a robot.

2. Beware of hallucinations. Only UX experts with real world experience can fact-check and correct false assertions and assumptions made by AI outputs, even if those assertions sound convincing.

3. Use AI apps like ChatGPT and Midjourney to prompt AI to come up with outputs. Become thoughtful with how you craft prompts.

- Provide the right amount of context when prompting AI

- Ask for multiple options or outputs from AI

- Keep iterating on the prompt to see how outputs change

- Build a prompt library with your most effective, tested prompts

-Keep iterating! 

4. In design, AI can be used for idea generation or image generation for prototypes and visual design.

5. In research, AI can be used to help write interview questions, synthesize user interview notes into themes, and rewrite or edit the way findings are presented for greater clarity.

6. Ask ChatGPT to ask you follow up questions once you give them a prompt. This will help the algorithm catch missing context.

7. Iterate on your prompt by cherry picking or editing out passages of previous AI responses. This will help the AI delve deeper into areas of your interest. 

8. Remember that human responses can be unpredictable, so AI can never replace user research. Instead it can help with data collection and data synthesis of real user data. 

9. Start with small tasks that AI can perform each day, before prompting them for more complex tasks. This will help you get started if you are completely new to AI.

10. Study related professions to better understand how AI works and why  it works the way it does. This doesn’t mean you have to be an expert engineer, but you should know at a high level what causes AI to behave the way it does at a structural level. 

Common FAQs about AI and UX:

Can AI replace UX designers?

AI can help make the work that UX designers and UX researchers do more efficient and less time consuming. However, it wont be able to replace UX designers and UX researchers because ultimately design is about understanding human needs and emotions, and then delivering on them (and only humans can do that).

Can AI do UX design?

Yes, AI can help with idea generation, brainstorming research questions, synthesizing research, and re-articulating research and editing insights, findings, and recommendations.

Can AI replace UX?

AI cannot replace UX researchers and UX designers completing the end-to-end design process. AI is as limited as the inputs its given, and innovation in product design requires a deep understanding of human needs/emotions, cognitive psychology, and cognitive science.

How will AI change UX?

AI will streamline UX research and UX design practices over the next few years. This could mean leaner teams, but this could also mean more opportunity to create more products and services in an efficient manner.

Will AI replace UX researchers?

AI can help make the work that UX designers and UX researchers do more efficient and less time consuming. However, it wont be able to replace UX designers and UX researchers because ultimately design is about understanding human needs and emotions, and then delivering on them (and only humans can do that).