Atlas Editor: A node canvas for forms with thousands of rules

In short

Insurance forms are mostly logic. If this, ask that, skip the next five questions.

Atlas Editor is the canvas where our people build that logic themselves, as connected nodes, without waiting for a developer. It started as the tool for claims forms and spread across the company to other customer-facing forms. The biggest ones run to several thousand nodes.

Goal

Let the people who know the business build and change form logic themselves, safely.

Success metrics

Forms built and changed without developer help
Business areas using it
Time to build or change a form
Errors caught before a form goes live
Reuse of nodes and sections across forms
Satisfaction of the experts who use it every day

My role

I designed Atlas and owned it as its product owner: interviews, priorities, roadmap, and saying no to good ideas that would have to wait.

Who it’s for

Experts who use it for hours and know insurance far better than I do. We interviewed stakeholders and users and wrote personas we never bothered to make pretty. They were for the team, not for a portfolio.

Research board with an editor persona and notes on each user group — Personas and research notes. Not pretty, very useful.

Ideation workshop board on file structure and tenant linking — One of many workshops with business developers: from notes to decisions.

Decisions that shaped it

Show the logic, don’t hide it

The logic is the product, so we made it visible on a canvas. A form with thousands of rules is impossible to hold in your head as a list; as a map, you can see where every path goes and what changes if you move one question.

Same design system, tuned for experts

I started from If’s main design system and tuned it for a dense tool people work in all day: tighter layouts, a properties panel beside the canvas, and reusable nodes and sections, so experts build from parts they already trust.

Say no, then plan

Ideas and finished designs went onto a prioritisation board, then onto a roadmap. Plenty of good ideas had to wait, and because the trade-offs were visible to everyone, saying no became a shared decision.

User test observations board — First iterations and user testing.

A single node with its properties panel — One node: what the customer sees, the variable it saves, and whether it holds personal data.

A splitter node connected to two lists — A splitter: one answer decides which path comes next.

Prioritisation matrix and roadmap — Ideas and designs, prioritised and put on a roadmap.

The design

A form with thousands of nodes needs more than a canvas. Forms are organised by country and product, search covers every node in a flow, groups and sections keep big forms readable, and several nodes can be selected and changed at once.

Atlas file list organised by country and product — Forms organised by country and product.

Searching for nodes on the canvas — Search across every node in a flow.

Nodes grouped into labelled areas — Groups keep related questions together.

Several connected nodes selected at once — Select several nodes and change them together.

A flow divided into sections — Sections turn one huge flow into chapters.

Result

What started as a claims tool now builds customer-facing forms across the company. Zoomed out, part of one form looks like a big spider web.

Part of one real form in Atlas, zoomed out — Part of one real form. Yes, all of that is logic.

Looking back

Atlas taught me what systematic design really means. The canvas, the nodes and the design system behind them are tools other people build with every day. When the system is right, teams across the company create forms I’ll never see, and they still work.

Designing for experts was its own lesson. They don’t need things hidden or simplified; they need them clear, fast and predictable, because they spend whole days in the tool. And every form they build ends up in front of a customer, so their experience shapes ours.