I am going to share something I have not shared before. A real project, a complete redesign of a banking dashboard screen using AI at every stage of the process. No client or company names but the process exactly as it happened.
The problem
The screen was a financial summary dashboard that enterprise banking users saw when they logged in. It had gone three years without a significant redesign. The information was there but the layout was dense, the visual hierarchies were broken and user feedback consistently said the same thing: I cannot find what I need quickly.
I had two weeks to deliver a proposal. Normally that timeline would have given me enough time to research, wireframe and maybe get to a high-fidelity mockup. With AI I decided to try something different.
Research with Claude
The first thing I did was take the user interview notes the research team had compiled over months. They were over forty pages of transcriptions. I passed everything to Claude and asked it to identify the three main frustration patterns and the three tasks users mentioned most frequently.
In fifteen minutes I had a synthesis that would have taken me a full day of reading and analysis. It was not perfect, I had to adjust some conclusions that Claude had oversimplified, but as a starting point it was remarkable.
Visual exploration with Midjourney
With the insights clear I moved to explore visual directions. I asked Midjourney to generate financial dashboard concepts with different visual hierarchy approaches. Not to copy directly but to break my own mental patterns of what a banking dashboard should look like.
Of the twenty variations I generated there were three that showed me something I would not have explored on my own: an asymmetric content distribution that prioritized cash flow over general balance. That aligned perfectly with what users were asking for.
Design and system in Figma
I went back to Figma with a clear direction. I built the wireframe using the existing components from the client's design system, adjusting the distribution Midjourney had inspired but adapting it to the real constraints: the 12-column grid, the spacing tokens, the institutional color palette.
I used Claude to iterate the microcopy for each module. Chart labels, tooltips, empty state messages. Everything that is normally left for later I solved during the design phase.
The result
I delivered in one week what normally took me three. Not because I worked less but because every stage of the process was more efficient. Research was faster, visual exploration broader and content detail more complete from the start.
The client approved the proposal with minor changes. Users in testing reported finding the main information in less than half the time.
It was not AI that designed that screen. It was me using AI as an extension of my experience. Ten years of designing for banking gave me the judgment. AI gave me the speed.
That combination has no competition.