IntelliBooks — AI in Insurance

Ask a property insurer where a given risk is located and you'll usually get a postcode. Ask where it actually is — the specific building, its elevation, its distance from the coast or the river — and you'll often get a shrug. Most insurers geocode exposure to the postcode centroid: a single point that represents an area which might span kilometres. For everyday admin, fine. For catastrophe modelling, flood rating, and accumulation management, it's a quiet, systematic error that costs real money. The centroid problem A postcode centroid places every risk in that postcode at the same point — the geographic middle. But hazards don't respect postcode boundaries, and they're often sharply local. Flood risk can change completely across a single street depending on which side of a contour you're on. Wildfire risk depends on the specific slope and vegetation. Storm surge depends on metres of elevation. Snap all those risks to one centroid and you've thrown aw...

Weather data is abundant, largely free, historically deep, and directly relevant to a large share of insurance claims. It is also, at most insurers, joined to almost nothing. That combination — enormous relevance, negligible cost, near-zero utilisation — makes it one of the clearest missed opportunities in insurance data. What weather data could be answering Claim validation. A claim reports wind damage on a specific date at a specific address. Was there wind at that location on that date? This is answerable in seconds from public data. Done systematically at first notice of loss, it's one of the highest-yield fraud and error checks available — and it requires no model, just a join. Event response. A hailstorm crosses three districts. Which of your policies are in the footprint? Insurers who can answer immediately mobilise adjusters, warn customers, and set reserves within hours. Insurers who can't find out as the claims arrive over the following fortnight. Reservin...