AllRisk — One Page, Three Risks

📣 Context

AllRisk is a Portuguese insurance broker that wants to do the opposite of what the sector usually does: speak plainly, with no small print and no endless forms. The site had to show, in seconds, which protection makes sense for each person and each company.

The audience runs from people with a predictable office routine, to people who drive, train and travel every day, to people who work around heavy machinery. One list of products could not speak to all three. And behind the site, the brief went further: a back office the AllRisk team could run on its own, and AI that works on the broker's documents and on the activities of its clients.

🏆 What Was Delivered

🥋 My Role & the Team

As Director of Brand & Product at Digital Streetz, I led the project with a team of eight, working directly with AllRisk's founders, CEO and CFO. I owned the concept of the risk slider, the UX/UI, the design system and the art direction for each level, and directed the build of the WordPress theme, the back office and the AI layer.

🧭 Key Decisions

• A question instead of a catalogue. The site opens with one question, "what is your risk level?", and a slider with three positions changes the whole page: the opening copy, the everyday situations, the coverage, the testimonials, the accent colour, the logo and the background video. One page for individuals and one for companies make six experiences in the same site.
• Each level speaks in its own voice. Low: your life is calm and we want it to stay that way. Medium: you have challenges every day and we are ready for them. High: your routine demands more and we give you that coverage. Individuals are addressed informally, companies in the plural, with little insurance jargon.
• The level travels with the lead. The chosen risk level goes with every contact request, so whoever replies already knows which profile they are talking to.
• The client runs the content. Each level, coverage card, situation and testimonial is its own content type in WordPress, so the AllRisk team edits every level's copy without a developer.

🤖 AI That Shows Its Source

Insurance runs on documents and legal duties, and a wrong answer costs trust. So the rule for the AI layer was simple: the model helps find and classify; the answer comes from a source you can check.

• An assistant that answers from documents (RAG). AllRisk's PDF documents are indexed, and every answer is built from the passages found in them, not from the memory of the model.
• Content extraction. The same documents are turned into text and structured fields the team can use.
• SeguroIA. A plugin that tells which insurance is mandatory for an activity. The person describes the activity in their own words or gives the CAE code; a language model helps classify it, and the result comes from a local base of mandatory insurance, not from the model.

🖼️ The Work

The visitor drags the slider, clicks the bar, or picks a level in a fixed selector that follows the scroll. Below it, a list of everyday situations, from "your biggest risk at work is spilled coffee on the keyboard" upwards, lets them check they picked the right level, and the button at the end changes its wording with the level.

The visual system has light and dark themes. Surfaces are neutral and cool, and the accent belongs to the risk level, green, amber or red, each with a darker version in the light theme to keep the contrast. The back office also includes a contents calculator that stores proposals, sends them to the client and the team, exports to CSV and prints as PDF.

🧠 Takeaways

• A question beats a catalogue. When audiences differ this much, letting people say who they are first makes every message after it relevant.
• AI earns trust when it shows its source. In a regulated business, the model should find, classify and structure; the facts should come from documents and data the client owns.