Building a Solution for Payment & Subscription Tracking with AI Tools
01 Context
The way we build products is changing rapidly. Over the past couple of years, I’ve been using different AI tools in my daily work, but this time I decided to push further and build a product from scratch with their help.
I believe the role of product designers will keep evolving. Just as UX/UI designers once grew into product designers by taking on more of the business side. I see us moving toward becoming true generalists.
Another big shift is happening in how we solve everyday problems. Education is becoming increasingly personalized, and I think product design will follow the same path. Instead of relying only on external services, we’ll start creating small, personal tools, like apps built by ourselves, for ourselves, tailored to our specific needs.
Almost like having our own personal App Store 😀
02 Goals
I set out with three main goals for this project:
Build from 0 to 1. Take an idea and turn it into a complete product, combining all steps in one path.
Experiment and learn. To test new tools and explore what I could learn along the way.
Execute independently, and launch a product that I built entirely on my own, from concept to release.
The only restriction I set for myself was speed: I had to launch within a week, which translates to roughly 48 working hours in total.
03 Approach
Defining the pain point.
I realized I had no clear picture of how much I’d be charged in the coming weeks. Subscriptions, loans, shopping installments, my mobile bill, insurance, and other regular payments were all scattered across different cards: one for daily expenses, one for subscriptions, a credit card for extras, and another for business costs.
There are services that consolidate this information, but many cost over $120 a year, and I didn’t want to pay just to see my own data in one place.
What I needed was a simple tool that brought everything together. Even if I had to enter some payments manually, that was fine, as long as it gave me clarity.
In this case, I was the user of my own product, which made the problem crystal clear and gave me a direct path forward.
04 Research & Exploration
04.1 Research
Before diving into AI, I gathered all my notes, sketched ideas, and structured them into one place.
Then I turned to prompts: starting with ChatGPT (I tested Claude too) and Perplexity for research.
This gave me:
A market overview
A competitor list
A first draft of possible features
I reviewed competitors, took screenshots, and made notes to identify gaps and opportunities.
You can check my Figma Page.
04.2 Journey Mapping & Ideation
I kept things analog here: sticky notes on the wall. This helped me brainstorm and quickly grow a list of ideas without overthinking.
04.3 Prototyping with Figma Make
Next, I returned to AI to refine prompts and used Figma Make to create the first prototype.
It was a mix of good and bad: limited in some ways, but surprisingly effective in others. After a few iterations, I landed on something that covered the core features: creating and saving expenses, and viewing everything in context.
It wasn’t perfect, but it worked. A better version of my first draft, and a glimpse of this new AI-powered design reality.
✅ Idea approved by the creator (me).
05 Design
As a designer, this is the part I enjoyed the most.
Moodboard & Visual Direction
I started with a moodboard, spending a few hours gathering inspiration, references, color palettes, and visual ideas, and saving them in Cosmos. This helped me define a visual style and set the tone for the product.
For illustrations, I experimented with AI. I used ChatGPT to generate visuals (I know MidJourney could produce higher quality, but it would have taken more time). For this stage, I chose speed and clarity over perfection.
I found a watercolor reference I liked and used it for the images I collected
Iteration & Feedback
Once I had some initial drafts, I went through multiple iterations. Then I used UX Pilot to get a second opinion. One of the most valuable insights was the idea of adding Reports — a feature I hadn’t initially planned but that made a lot of sense for tracking payments.
This led to another round of mockups, and I landed on a version I liked — something strong enough to serve as the foundation for a real app.
☀️ My Figma file (password: inbelivwetrust)
📍 Currently I'm here:
A pre-final design that feels ready to evolve into an actual product.
06 Key Takeaways So Far
AI is a real accelerator, not a magic wand. Tools helped me move faster, but they didn’t replace design decisions. I still had to direct, refine, and make trade-offs.
“Good enough” beats “perfect” when building momentum. Progress mattered more than polish at this stage (yeah, it's sounds like a cliché).
The niche is very specific. I don’t expect big results without investing in marketing or ads, and that’s okay. For now, the focus is on solving my own problem and proving the concept.
-- Next Steps
I’ve spent a little over 20 hours so far. The next phase will focus on execution:
Build the app based on my designs (still deciding between Cursor or Claude Code).
Add animations.
Integrate authorization.
Design and implement the Analytics screen.
Integrate Stripe for payments (or via App Store, haven’t tried either yet).
Figure out how to test everything (tools like Testdriver.ai might help).
Create a landing page.
Nice to have:
Integrate Plaid for direct bank account connections.
Dark Theme.
