SHARE:

AI in UX redesign is starting to prove useful for a problem many Brazilian companies know all too well: that old system that does everything right and keeps the operation running, but looks like it came from another decade. Nobody wants to mess with something that works. Still, a cramped screen packed with menus and buttons from another era takes its toll on the time of the people who use it and on the patience of whoever has to train new hires.

A recent case helps show how this plays out in practice. It is the story of Platinum Health Keyrus, published by ITWeb this week. Platinum Health, a health plan provider in South Africa, needed to move its platform to the web without losing a single function, and the consulting firm Keyrus built an AI-powered workflow to make it happen. According to the outlet, more than 200 screens were redesigned, with output roughly three times higher per sprint, and all of it without a designer dedicated exclusively to the project.

Below, you will see how the work was organized, why standardization came before automation, and what lessons this story offers for legacy system modernization here in Brazil. Here is the big takeaway up front: AI sped up the process a lot, but it did not do anything on its own.

What happened at Platinum Health

According to ITWeb, Platinum Health describes itself as the largest closed medical scheme in South Africa. Closed, in this case, means it serves a specific group of people rather than the general public. Its operation covers primary and specialist care, chronic disease management, plan administration, and workplace health programs.

Behind all of that sits a platform that handles member registration, requests, and correspondence. These are sensitive processes, the kind where a failure does not stay inside the IT department. When a system like this loses a function along the way, the impact usually reaches the people who depend on the service.

The dilemma was clear. The functions had already proven their value and needed to stay exactly as they were. The interface, on the other hand, belonged to an older generation of software and did not take advantage of what modern web frameworks can do.

The solution was to swap the shell and keep the core. Every screen would first be built as a prototype in Figma, following the new design system, instead of being a copy of the old desktop layout. That choice matters. Copying the old screen into the browser would be faster at first, but it would drag along every bad habit of the old interface.

Receive the best innovation content in your email.

All the news, tips, trends, and resources you're looking for, delivered to your inbox.

By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.

The real bottleneck was time

According to the report, the biggest challenge was not technical. It was the calendar. Integrations, ongoing development, and fixed deadlines were all running in parallel, with very little room to maneuver.

Putting a Platinum Health designer on the migration full time simply was not realistic. And there was a catch: design and development were happening at the same time. So requests to tweak finished screens kept coming in while others were still waiting in line.

Anyone who has worked on a similar project will recognize the scene. Picture a small team keeping a company’s core system up and running. They spend the day putting out fires, handling tickets, and shipping integrations. When someone suggests redesigning the experience, the obvious question pops up: who is going to do this, if everyone is already swamped?

Situations like this help explain why so many modernization projects stall. If this topic is on your radar, it is also worth reading about automation and AI to modernize legacy systems, which covers other ways to get these initiatives moving.

Standardize before you automate

The logic of the project fits in one sentence: first you organize the work, then you let the machine repeat it. Sounds obvious. But it is exactly the step many people try to skip when they get excited about a new tool. As described by ITWeb, Keyrus split the process into three phases.

  • Manual phase: the first screens were built by hand in Figma, on purpose. The team defined how components, layouts, and the overall experience should behave on the web and recorded every decision in Markdown files, covering color, typography, spacing, icons, and even how to resolve conflicts between reference sources.
  • Prompt phase: those files were loaded into Claude Projects as permanent knowledge. With the Figma MCP server, Claude started creating screens directly in the tool, always following the same standard, without anyone having to repeat the rules with every request.
  • Custom skills phase: with Claude Code and tailor-made skills, each old screen got its own path to the new version. From then on, short prompts were enough to generate screens of the same quality.

Each tool had a clear role. Figma was the final home for the designs, so it made sense to build everything there, with no conversion steps in between. The Markdown files were plain text, which both people and the model could read and improve as the system matured.

This point deserves a closer look. A rules document written in plain text works like a living manual. Imagine a team that notices, after a few screens, that secondary buttons are starting to look too much like primary ones. All it takes is updating the rule in the file, and the next screens come out already fixed. The decision is on record for everyone, including whoever joins the project later.

And there is one detail that makes all the difference: every screen went through manual review. The team checked each result against the design principles to make sure no button or function got left behind. Generated output was never accepted without a check.

The numbers ITWeb highlighted

According to the outlet, output went from four or five screens per two-week sprint to about 16. That is close to 3.5 times the pace of manual work. On a project with more than 200 screens, that kind of gain makes a big dent in the schedule.

Another figure stands out. The number of people contributing to design went from one to three, and none of them had to set aside their main development work. That kept integrations and other deadlines safe. According to the report, every screen followed a single documented standard, went through review, and had each original button and function confirmed.

In the article, Jordon Pillay, a data engineer at Keyrus, sums up the idea by saying the team wrote down the rules first and only then let AI apply them, and that the standards are what made the automation reliable and repeatable. Joos Nieuwoudt, head of software development at Platinum Health, told ITWeb that the approach was conceived inside the company and that Keyrus helped put it into practice. According to him, this removed a critical dependency early on without pulling developers away from deliverables they had already committed to.

What this case teaches Brazilian companies

The design system comes before the tool

The first lesson is that UX design with artificial intelligence pays off more when there is a clear design system behind it. Without defined rules, AI tends to produce screens that may look nice but differ from one another. With rules, it works like an assistant that follows the manual.

Documentation helps people and machines

The second lesson is about documentation. Writing decisions down in plain text, without complicated formats, serves people and the model at the same time. Newcomers to the team understand the reasoning behind the choices. The model gets a concrete guide. It is a habit that takes discipline, but it usually pays for itself over the course of a project.

The human eye is still the filter

The third lesson is that human review is still the quality control. AI speeds up production, but someone has to go through screen by screen and check that what matters is still there. If you want to understand how this relationship between professionals and tools is taking shape, the piece on artificial intelligence and UX design in the VCU internship offers an interesting perspective.

Tools we use daily

So where do you start with interface modernization? A sensible path is to take inventory of your existing screens, define the basic rules for color, typography, spacing, and components, and run a pilot with a real deadline. Imagine a logistics company with an old internal system. It could pick a small module, like delivery registration, design the first screens by hand, document every decision, and only then test automation on the rest. If the pilot exposes flaws in the rules, fixing them is still cheap.

Things to watch before copying the recipe

One thing needs to be clear: the content ran on ITWeb as material released by Keyrus itself. That does not invalidate the results, but it calls for a critical read. The numbers come from the company that provided the service, and the article gives no details about costs or the bumps along the way.

The biggest risk for anyone trying to replicate this is skipping standardization. Without solid rules, the speed of AI also speeds up inconsistencies. Instead of a cohesive set of screens, you could end up with dozens of small variations that need rework later. To understand how tools that speed up delivery can help or hurt, the article on artificial intelligence and low-code reinventing companies rounds out this discussion nicely.

Accessibility and user research cannot be left out either. The article does not mention testing with the people who use the system every day, and that is a step any serious project should include. Proper contrast, screen reader compatibility, and keyboard navigation can be built into the rules from the very first file. That way, AI applies these safeguards to every screen instead of leaving them for the end.

Listening to the people who will use the system is still irreplaceable. A screen can follow the standard to the letter and still confuse someone who needs it every single day.

In the end, AI applied the rules, but people created them. If your old system is begging for a fresh look, start by documenting your visual standard before picking any tool. And if you want to dig deeper, keep exploring our content on design systems, Figma, and AI for product here at Método Viral.

Picture of Rafael

Rafael

Operations

I transform internal processes into delivery machines — ensuring that every Viral Method client receives premium service and real results.

Fill out the form and our team will contact you within 24 hours.

Related publications

Andrea Iorio and Digital Transformation: How to Protect Your Website Against Cyber Threats

See how to protect your website through the Andrea Iorio case. He used Metodo Viral digital transformation to change the

From an Insecure Website to a Lead-Generation Machine: CRIAR Varejo’s Digital Rebirth

See how the Criar Varejo case can help your business go from an insecure website to a LeadGen Machine

How Much Does a Freelancer Charge to Create a Website in 2026?

How Much Does a Freelancer Charge to Create a Website in 2026? Avoid mistakes and high prices by knowing what

Receive the best innovation content in your email.

All the news, tips, trends, and resources you're looking for, delivered to your inbox.

By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.

Rafael

Online

Atendimento

Website Pricing Calculator

Find out how much the ideal website for your business costs

Website Pages

How many pages do you need?

Drag to select from 1 to 20 pages

In just 2 minutes, automatically find out how much a custom website for your business costs

More than 0+ companies have already calculated their quote

Fale com um consultor

Preencha o formulário e nossa equipe entrará em contato.