Like most reinsurers, this firm’s leadership team was under pressure to demonstrate an AI strategy. They’d seen the demos, attended the conferences, and heard the promise: AI would transform pricing, automate reporting, and make better decisions possible across the business.
The reality was more sobering. Their data was fragmented across disconnected systems. Actuaries manually reconciled figures between platforms. Settlement processes took days. Month-end close took longer than it should. Teams were spending weeks after quarter-end validating data before analysis could even begin.
No amount of sophisticated tooling was going to solve it - AI models are only as good as the data they run on. Layer AI onto fragmented, manually-maintained data and you’re just scaling mistakes faster.
What we did
We helped the firm move forward on three parallel paths - because waiting until everything was perfect wasn’t an option either.
Path one: the data layer. We built the foundations - clean, structured, validated data flowing automatically between systems. This meant establishing standardised interfaces between actuarial, finance, risk, and investment systems, automating ingestion and validation, and creating a single source of truth that every function could draw from.
Path two: personal productivity tools. While the foundations were being built, we helped deploy AI tools for day-to-day work - with clear governance frameworks so people could start benefiting immediately without waiting for the full infrastructure.
Path three: structured pilots. We identified specific processes where the data was already clean enough and workflows were well-mapped, and ran targeted AI experiments. This gave the leadership team real evidence of what AI could deliver - and what foundations were still needed - without betting the whole programme on untested capabilities.
Why it matters
Most reinsurers want to use AI. Few have the operational foundation to do it well. The gap isn’t permanent - but closing it requires building foundations, not buying tools.