An Enhancement of the (re)Design Process for a Complex, Multi-Actor Procurement Platform

Let's start with the "why"

I've been the lead designer at Multonion, a digital procurement platform, for two years, and I've wanted to redesign it for almost as long.

Unfortunately, it was never the right time, as there was always a different priority.

So I changed my strategy. Instead of waiting for the go ahead to focus on the redesign, I gave my evenings and weekends to it, with as little manual effort as possible (you see, I'm just a girl).

Working with Claude, I redesigned and built a fully working prototype in two weeks.

What the redesign set out to do

• Improve the UI. While the last designer did an excellent job translating complex user needs into an interface, I believed there was need for a bit of change. A simpler interface with lots of whitespace would create ease in the user's journey.
• Improve the experience. Procurement is complex, but it doesn't have to be difficult. A lot of the flows created difficulties for MSME owners to navigate and complete tasks. This was tested with a few users.
• Integrate AI into the workflows. AI should improve tedious processes and provide clarity and ease in achieving one's goals. I wanted users to have this option if they needed it.

Why I built it instead of designing it first

• To prove feasibility. There has been some pushback on the feasibility of some of the design ideas I have proposed due to technical constraints. For example, having an auto-save feature. I decided that building it would immediately show if a thing can work or not.
• To see how every flow works. Prototyping in Figma is a near nightmare for products as complex as Multonion. With multiple user types, flows and use cases, one action may require several prototype paths. This makes a full end-to-end test tedious and sometimes even confusing.

Above is the prototype data for the tender creation flow alone. There are over 20 flows.

• To keep the design system alive. Through the Figma MCP, tokens and components flowed between design and code, and a Claude agent gave developers exact specs directly. This significantly reduces development time.

Iterating in the medium of the product

These images tell the story better than paragraphs would. I didn't accept first outputs; I kept iterating until the work was right.

What this says about designing complex products

The traditional pipeline assumes building is expensive, so we simulate the product for as long as possible. AI has broken that assumption.

It is now faster to go from idea to prototype, and to iterate as often as is needed.

Worth noting though, is that this does not eradicate the need for design thinking. If anything, we need it now more than ever.

For two weeks I made design decisions at high speed and saw them realised immediately, directing an LLM with the same specificity I'd use to brief a design team. That's the emerging role of the AI-UX strategist: designing through the build.

For complex, multi-actor platforms, I don't think I'll go back.

Random screens from the project