A risk cockpit for a broker: the map shows where exposures sit, the dashboard says how much — premium, claims, sites — and every filter keeps the two in sync.

The context

At a broker such as Diot-Siaci, risk has an address. Portfolios, industrial sites, fleets, claims: they live in separate files and tools. An underwriter or account manager needs one view: where exposures concentrate, what a region weighs, how a filter (line of business, year, client) changes the picture. An expert GIS is not enough — it has to be a decision dashboard, readable in a meeting, with figures you can stand behind. The same problem faces an operations director, a risk manager or a branch network: objects in space, KPIs on top, one filter tying both.

What we delivered

  • Interactive map (choropleth + pins): risk concentration by territory, geolocated sites, statuses readable at a glance
  • Synced KPI strip: covered sites, aggregated premium, claim count, loss ratio — recomputed on the visible filter and viewport
  • Business filters: line (property, liability, fleet…), year, region, client, without leaving the screen
  • Drill-down: click a region or a site → record (exposure, claims history, linked files)
  • Export of the current view (PDF / Excel) for a committee or a renewal
  • Data pipeline: imports from existing files and APIs, no re-keying into yet another tool
  • Brick architecture: map, dashboard, export and data connector are separate — OCR on claim files, automated alerts or a portfolio assistant can be plugged in later

The bricks — reusable on your business

This is not a closed product. Each brick comes off the shelf: a Mapbox map for a branch network, a KPI dashboard for a sales cockpit, OCR for invoices or claim files, an assistant to query an e-commerce catalogue.

Risk cartographyChoropleth, pins, heat. Transfers to stores, worksites, a fleet, a property portfolio.
KPI dashboardIndicator cards, ranked lists, trend. The same foundation as a sales or finance cockpit.
Filters and drill-downOne selection, two surfaces (map + figures). A business-app pattern, not a map gadget.
Document OCRThe same brick as Azô and ComplyTrust: read a claim file or a contract and pin it to the site.
Automation / reportingExports and alerts (loss-ratio threshold, concentration) without rebuilding the sheet every week.
AI assistantQuery the portfolio in natural language — the same chatbot brick as on an e-commerce catalogue.

Outcomes

  • Map and indicators share the same filter: no gap between « the map » and « the report »
  • A territory unfolds into sites and claims, without switching tools
  • An export matches exactly the view shown in the meeting
  • New sources (file, API, documents) plug into the pipeline, not into the map
  • The map + KPI pair is reused as-is for another business: retail, logistics, property, a branch network

Tech stack

Mapbox GL JS · React · TypeScript · Dashboard KPI · GeoJSON · API REST · PostgreSQL · Python · OCR · Google Cloud Run