UNIQA Hungary

Redesigning the sales process of a classic insurer going digital

The UNIQA Business liability flow on a laptop: the customisation step with the live basket total.
Client
UNIQA Hungary · freelance
Role
Design system · Sales flows · AI-Casco
Year
2025 – present
Team
Lóránt Valaczka · János Baksa, Design Team Lead
Duration
Ongoing, since 2025
Platform / scope
Responsive web

Top 5 insurer in Hungary · Since 1990 · 665,000 policies · Part of the UNIQA Group

Freelance project

My role

Freelance product designer. I built the design system and designed the new sales flows: MTPL, Business liability, and Casco with its AI features.

Problem

UNIQA sells most of its products through agents, and each online product was built its own way. Taking the next one online meant starting over.

Approach

Built a design system, standardised the purchase process on top of it, and moved the products over one by one.

Key insight

A design system was only the baseline; the real job was getting every flow onto one base, fast.

Impact
  • Business liability, agents-only before, now sells online on par with the agent network
  • MTPL’s new flow turned 9 percentage points more of its started quotes into policies, in its first week
  • A new product flow designed in weeks, ~85% from shared parts
  • AI joins the Casco purchase flow on the same standardised base
Timeline

2025–present

Skills

Design systems · Sales flows · Information design · AI assistant UX · MVP scoping

Three phones showing the UNIQA motor insurance flow: calculator, contract signing and price details.
The MTPL flow on the new structure: calculator, contract signing and price details.
Starting point

Every flow was its own project

UNIQA is where it had to grow up

I came to UNIQA from Cherrisk, its digital insurer, with the Design Lead I had worked with there, János Baksa. The first version of the sales flow was built at Cherrisk.

Each online product had its own form

Each online product had its own form, built its own way: different steps, different fields, even for the same question. Taking the next product online meant starting over. The brief was a design system. It turned out to be the smaller half of the job.

Before: each product had its own form. After: one structure, the first four on it.
Design system

A design system wasn’t enough

Every part defined once

We built the design system first, on the UNIQA Group International Design Guide; the components and the token architecture were reshaped around UNIQA Hungary’s digital needs.

It was a migration problem

At an insurer that has sold the same way for decades, nothing changes on its own: everything connects to everything else. We needed a plan where each flow goes faster than the last, because the last one left patterns behind. It looked like a design system job. It was a migration problem.

105 component sets, 757 variants, 263 tokens: each part defined once, used in every flow.
Sales flow framework

One structure for every purchase

Every page has the same two levels

Steps, and sections inside them. The structure sets the page; the design system gives the parts; only the content is the product’s own.

A finished section folds

One thing the Cherrisk version lacked: a section stayed open after you filled it in, so a long form never felt done. Now it folds to one line, and a long insurance form feels like a row of short ones.

About 85% of each flow is shared

We counted the component instances in both files.

Each new product starts mostly built.

The frame stays; only the blue slots change from product to product.
Products

Two products, one build

Business liability was the first on the structure

The first product was meant to be Travel. The framework was still a concept, Cherrisk’s Travel flow was finished and worked, so the business shipped that one, rebranded. Business liability was the first on the structure: for agents, then for the public. MTPL followed, one responsive flow for customers and agents alike.

Impact

What it did to sales

Opened the same flow to the public

Business liability had only ever been sold by agents. We built their screen first, then opened the same flow to the public.

The part that moved

MTPL’s new flow went live in October 2026. First week, against the old flow: 9 points more of started quotes signed, almost 30 points more started at all. Early data, and not the only reason. But the part we designed is the part that moved.

Business liability by channel; MTPL’s first week against the old flow. Figures: UNIQA’s product team.
~85%of every flow comes from shared parts
≈ 1:1direct online : agents, Business liability
~20%of people who start the Business liability calculator buy online
AI layer

AI on the same base

Casco is the hardest product to buy

Describe your car in detail, then pick between packages that look alike. I brought a list of where AI could help, and it became the roadmap.

Two rules hold it together

Every value AI fills in is marked, so you can check it. One switch turns AI off for the whole flow. The AI patterns are design system states, so any product can switch them on.

AI in this project

AI helped me too: Claude and ChatGPT turned the Casco policy documents into decision criteria.

Five AI touchpoints in the Casco flow, from the landing page to any field’s tooltip.
Testing

Casco: what testing told us

What ships first, and how it tested

The MVP keeps package comparison and photo autofill. Researchers walked 12 testers through a clickable prototype, each on their own car’s data. The registration scan I designed was one of the strongest results: people expected to photograph the document and liked scanning its QR code with a phone.

What we still have to work on

Three areas still need work: how easily people notice the AI, how concrete and trustworthy its answers are, and how clear its questions are. The fixes belong in the shared parts.

The planned package-choice wizard: NiQA asks a few questions, then the packages are compared.
Closing

How it went

My part, and what comes next

Role

Flow, screens and interactions: the design system, the sales flow framework and the AI-Casco touchpoints.

Team

Lóránt Valaczka · János Baksa, Design Team Lead, UNIQA.

Next

Testers liked the type finder and the needs assessment, so we will bring them forward in development.

The full storyPortfolio deckThe complete case, slide by slide.Open →

Next case