AI for UX: A DesignLab course

Problem: I wanted to take an intuitive step towards learning about AI, and how it can be utilized within UX design, strategy and execution.

Audience: Fellow DesignLab classmates and a Lead Mentor.

My Role/Contributions: In addition to working on my personal project, I contributed and shared new insights and observations within peer group sessions.

Tools & Technology: ChatGPT, Claude, MS Copilot, Perplexity, Dall-E, Midjourney, Firefly, Figma.

Outcome: Developed a much better understanding of AI’s role within design. This learning was showcased within a fitness app project, where AI assisted in design, research, ideation and strategic implementation.

Duration: 1 month course, 40 hours of content learning.

Similar to how I transitioned into UX Design, I wanted to legitimately learn how AI fits into the design process. So I took another course through DesignLab. I learned a lot, applied my findings to an exciting project, and had a great time connecting with other UX Designers along the way.

Some of my key takeaways:

  • AI should be embraced as a design assistant. And really, who wouldn’t want an assistant?

  • AI output should always be vetted, refined, and expanded by a human.

  • AI can significantly improve both the speed and cost-efficiency of building a product.

Project Objective

Design and develop an AI-powered fitness app that brings together personalized workout planning, nutrition tracking, and community engagement in one seamless experience.

The client approved the use of AI tools to help accelerate research, ideation, prototyping, and workflow design. The goal was to demonstrate clear efficiency gains throughout the process.

Below is a summary of the AI-assisted processes used to shape this app. This is not a fully developed, end-to-end project, but rather a focused exploration into how AI can support design workflows.

Part 1: AI Design Foundations & Discovery

After learning the basics of large language models, along with key ethical considerations like bias and responsible use, I started exploring image generation. I quickly discovered that specificity is essential to getting strong results. The more explicit the input, the better the output.

AI image generation proved especially helpful as a space to brainstorm and explore creative directions. I began by generating early logo ideas for the fitness app, focusing on visual cues related to strength and health.

Combine these two logos. Keep the left side as a heart shape, and merge it with the right side flames, that appears in the other logo.

Create a modern minimalist logo for fitness app, in black and gold, stresses strength and fuel, include a heart.

I liked the initial shape of this logo attempt because it captured a sense of fuel and energy. My goal was to combine the silhouette of a flame with the shape of a heart, something that could visually represent both motion and care. At this stage, though, it still looked a bit too much like an ice cream cone, which sent the wrong message for a fitness brand.

I tried additionl variations on the “heart” and “flame” concept, and found that rapid image generation became a really effective brainstorming tool. I started thinking more about the contrast between rendered and minimal graphics, and how those styles impact a product’s voice.

The flames still looked a bit like wheat, but I felt the logo was heading in a promising direction. I also realized that regenerating an image with new instructions was often more effective than trying to edit an existing one.

Not perfect, and still more work required, but I now have plenty to build on!

Mood Board

Building a mood board came with mixed emotions. On the positive side, several boards could be generated almost instantly. But at the same time, even some of the best-looking results had weird or inconsistent details. AI struggled with human anatomy, often generating extra fingers, elongated arms, or awkward poses that became distracting.

Still, the purpose of a mood board is to quickly communicate a general vibe, not a polished product. In that way, AI really shined. It saved a lot of time and helped me test visual directions quickly.

My goal was to create a fitness app that felt inviting and inclusive. One challenge was finding the right balance between “modern and sophisticated” and visual diversity. Many AI-generated images leaned heavily on over-posed models, lots of skin, and a look that didn’t feel welcoming. At one point, I joked that I needed to put everyone in long sleeves, just to make it feel approachable.

I used Midjourney for the first round of mood board exploration.

Design a mood board for a sophisticated fitness app, stressing energy and health, tech inspired. Use both male and female, ages 25-40, use the color scheme of black and gold. Include photos and UI elements. Wear long sleeve tops and T-shirts.

Rest Day Reminder: "Recovery is part of the process. Take a deep breath, stretch it out, and let your body recharge for tomorrow’s wins!"

I decided to try Adobe Firefly and found the results more to my liking, especially in how it handled digital components like UI elements. AI’s biggest strength continued to be its speed.

That said, there were still some ongoing issues. It struggled with hands and fingers, and often overcompensated when I requested images that included age or diversity. Instead of adjusting what I had, it would often start over entirely. And no matter what I tried, I still couldn’t get the models to actually exercise.

Design a mood board for a sophisticated fitness app, stressing energy and health. tech inspired. Use both male and female, ages 35-45, use the color scheme of black and gold. Include photos and UI elements. Wear long sleeve tops and T-shirts. Use a radiant glow. Show people exercising. Generate as a modern illustration. Positive and encouraging.

Brand Voice

I’ll admit that I might be a little too easily impressed by AI’s writing suggestions. But more often than not, they’re genuinely helpful. This is a clear example of using AI as an assistant, where I stay in the driver’s seat and focus on guiding the insight, shaping the tone, and making additional edits.

I did have to remind ChatGPT to stop using em dashes, several times.

“Give me examples of the branding voice for a mobile fitness app. The mobile fitness app is called FitFuel. It’s goal is to provide positive fitness experience for the target ages 25-45. The app encourages health, fitness and happiness. It is encouraging and supportive and the goal is to make people excited about exercising”.

Tagline & Mission Statement

Fuel Your Fitness. Ignite Your Energy. FitFuel makes exercise exciting, empowering, and enjoyable. No guilt, no pressure; just movement that feels good and helps you thrive.

Push Notifications (Encouragement Without Guilt)

Welcome Message (App Onboarding)

"Hey, Amanda! Welcome to FitFuel, where workouts feel less like a chore and more like a celebration of movement. Let’s fuel your body, spark your energy, and make fitness fun! Ready to get started?"

Achievement Celebration (Gamification & Motivation)

"You did it! That’s one more workout fueling your strength, stamina, and happiness. Consistency is and you’re proving every day that you’ve got what it takes!"

Personas (ChatGPT vs. Microsoft Copilot)

I’ve always prided myself on being able to design personas that feel real and genuinely useful. To me, they are an art form of their own. When crafted with care, a persona doesn’t just describe a user, it creates the importance of empathy. Of course, building one properly takes time. That said, with the right direction, AI can be a powerful partner in creating personas that are both efficient and immediately usable.

Persona first attempt: ChatGPT

Generate a user persona targeting someone who uses a fitness app. Display in a designed layout, including a photo”. Include a quote, use black and gold color scheme.

Bias Audit Worksheet: Evaluating personas

For the purpose of learning about prompts, I intentionally began with something fairly generic. Unsurprisingly, the results reflected that choice: “Generic In, Generic Out.” I learned that a good prompt is like a solid design brief. The more specific you are, the better the outcome. So don’t scrimp on the details!

Revised persona: Copilot

Generate a persona for a fitness app user. Should stress positivity and health. Include diversity in age (20–65), body types, and cultural backgrounds. Avoid stereotypes. A mix of enjoying exercise and also hesitation.

There were definite improvements to the generic personas, using a more specific prompt. I believe that it would be even more evident among a survey of 5 or more personas.

For the future:

  • Most obviously, more Inclusivity and Accessibility needs to be introduced.

  • Take into consideration the cost of the proposed fitness app.

  • Discover any culture bias that exist with exercising, causing stress or hesitation.

Efficiency Metric Estimation

I’ll admit it: I can sometimes be a romantic designer. I like to take my time with a problem, sit in a coffee shop to mull things over, and create a variety of concepts. While these are all healthy and valuable contributions to my creative process, they can also become time traps.

When it comes to rough concept work, that is where AI really shines. With just a few tools, I can generate a wide range of ideas and imagery in minutes. This approach saves significant time compared to browsing across multiple sites, and it lets me focus more on shaping and refining the strongest directions

Part 2: Research

I’ve been fortunate to collaborate with some highly skilled UX Researchers throughout my career. What should be a quick discovery phase, though, often stretches into long scheduling conflicts paired with inflated research costs. I’ve found that AI helps remove those barriers, especially as a starting point. It makes research more accessible, streamlines the process, and turns insights into actionable results faster.

Here are some examples from my research plan for FitFuel:

Competitive Study

I want to do research on the following fitness apps; Strava, MyFitnessPal, Nike Training Club. Identify their core features, target audiences and their unique marketing strategies. Compare the offerings of all 3 apps, and intuitively uncover missing features or unique opportunities for potential customers. What is the fitness app market size, and does it cater to unique users, such as injured athletes, or low income users? What are some emerging trends, or unique needs, within the industry?

Key finding from simulated user feedback

I want to better understand the users of fitness apps. Develop 5 questions that I can ask fitness app users, with an interview guide to research these questions. Include a survey. Generate 10 realistic user responses to the questions about fitness apps, and the challenges the users may face. Any bias? Hallucinations? Analyze the user responses to identify key themes, pain points and unique opportunities.

Persona development and
bias mitigation

Generate 3 personas for fitness app users. They should have a positive attitude, be a casual user, and have the desire to be healthy and happy through fitness. Generate a photo of each user and check the personas for bias related to age, gender and body type. Suggest more inclusive alternatives.

First attempt

Bias review & regeneration

Part 3: Execution & Testing

The goal of this phase was to turn the fitness app research into visuals, and begin testing interactions. This included visual design, integrated workflows, usability testing, and ensuring ethical compliance. Since this work was AI-generated, the purpose was to create quick and insightful results. It is not a finished product, but a meaningful contribution to the design process.

I was especially interested in experimenting with AI prototypes, to see how usable they could be in practice. My personal goal is to spend more time shaping UX strategy, instead of spending the majority of my time inside design programs. AI allowed me to shift that balance, allowing me to focus on the bigger picture, while still producing tangible outputs quickly.

Generate a high-fidelity workout planner screen with onboarding, workout progress tracking, AI recommendations, and a calendar view that follows FitFuel's energetic, inclusive, tech-forward brand style, positive and encouraging tone, using the color scheme of gold and black.

Onboarding flow

Recommended workouts and planning

Generated microcopy

Generate 10 pieces of microcopy for the FitFuel app. They can be motivational messages, error states, onboarding prompts, completion messages, etc. The tone should be encouraging, Supportive, non-judgmental, aligned to FitFuel’s branding tone.

AI Workflow

Input: User goals, dietary restrictions → AI Processing: Nutrient calculation, recipe matching → Output: 7-day meal plan with shopping list, follows FitFuel's energetic, inclusive, tech-forward brand style, positive and encouraging tone, using the color scheme of gold and black.

Testing usability with AI personas

Create a usability testing plan with 5 tasks and follow-up questions for the FitFuel app's onboarding and workout planning features.

A review of potential bias in the usability testing plan

Check for any bias in the research questions, also check for hallucinations.

Part 4: AI-Powered GTM Strategy

The objective of this final phase was to guide AI in generating the content, for a Go-To-Market strategy, for the FitFuel fitness app. While the outputs were not always 100% to my liking, I made a deliberate choice to revise only through prompts, rather than dismantling and rebuilding everything in Figma. This approach allowed me to test AI’s strengths in handling complex design and strategy tasks with minimal manual intervention.

Throughout the process, I experienced many surprises. One of the most impressive outcomes was seeing a fully presented product concept come together in a fraction of the usual production time. The speed and scope of these results left me excited about the future of AI in UX design, particularly as a partner in accelerating strategic design work.

I want to create a Go-to-market Strategy. Create content for a 10-slide pitch for FitFuel, highlighting the app’s value proposition, market gaps, unique opportunities and positive business model.

Investor pitch deck

KPI’s

Suggest 8 key performance indicators for measuring success of a fitness app with AI-powered personalization features. Include target values and appropriate measurement intervals.

Diversity audit of AI-generated visuals & content

Review all AI-generated visuals and content for inclusive representation, for the FitFuel fitness app.

Time savings through AI integration

As a Designer, I can sometimes be guilty of falling in love with visuals. Not only with the creation of visual elements, but also including the search for the “perfect image”. AI generated may not result in the final outcome, but it does accelerate the process of brainstorming and developing concept artwork.

Conclusion

I learned a lot in this course and immediately began applying AI in my current UX work. AI really shines when pitching new product ideas, because it takes you from concept to reality quickly. Traditional ideation sessions can be hard to schedule and often expensive to run onsite, but AI can produce strong results with a fraction of the effort.

That said, even though AI can lay a solid foundation for a project, human insight and collaboration are still essential. The best outcomes happen when AI accelerates the groundwork and people step in to refine, validate, and bring empathy to the final design.