Volume 12

Volume 12 - AI Governance and Responsible AI

Volume 12 decides whether an AI system should run, for whom and on what evidence, using the claims assistant that Volume 11 built for an illustrative insurer with 600 adjusters and 30,000 claims a month. Each chapter settles one question with numbers, from how much human review a denial needs and how large a bias audit must be to when an erasure is finished and how many clean red-team trials a gate requires. The volume closes the handbook with an assurance case in Goal Structuring Notation that keeps the assistant at 150 adjusters and widens it in dated steps as the open evidence closes.