AIG CEO Eric Andersen's warning that AI data center projects are 'maxing out' P&C insurer capacity is a direct message to the mid-market insurance market: you are now facing underwriting complexity you have never seen before, and your current processes cannot handle it. AI infrastructure—from project finance through operational phase—demands assessment of risks that sit at the intersection of technology, geopolitics, supply chain, environmental impact, and cyber exposure. A mid-market UK insurer following traditional underwriting workflows (manual due diligence, spreadsheet-based risk scoring, siloed compliance checking) is now competing for the same risks as carriers with dedicated AI infrastructure teams. For any firm regulated under PRA SS1/23 or subject to FCA Consumer Duty PS22/9, this capacity crunch is not a distant problem. It is happening now, and it is creating real market friction.
What Andersen's comment reveals is that the insurance sector has finally hit the inflection point where the pace of AI-driven risk creation exceeds the pace of AI-driven risk assessment. This is not new territory for tech-focused industries, but it is new for insurance, which has historically operated on the assumption that human expertise scales through hiring and training. It does not. Not in this context. The EU AI Act, already shaping how financial services firms think about algorithmic governance, is beginning to force insurers to answer uncomfortable questions: Can you explain your underwriting decision on an AI data center project? Can you prove your risk model accounts for emerging failure modes in LLM infrastructure? Can you justify your premium in light of the novelty of the actual exposures? Most mid-market carriers cannot answer these questions with confidence. The ones who can are building competitive moats—not through hiring, but through systematic AI integration.
Trovix's position on this is clear: the answer to the capacity problem is not more underwriters. It is not better spreadsheets. It is integrating AI into the entire underwriting decision chain in a way that is explainable, auditable, and compliant with existing regulatory frameworks (FRC ISA UK, ICO UK GDPR, ISO 42001). Other vendors in the market—Harvey for legal work, Luminance for document intelligence, even Microsoft Copilot for generic productivity—solve narrow slices of the problem. They are good at what they do. But they do not solve the systemic issue: how to take a complex, multi-disciplinary risk assessment (AI data center infrastructure) and convert it into a structured, repeatable, compliant underwriting decision that your audit team can defend. This is where purpose-built insurance AI integration—not just document intelligence or search tools—becomes the competitive lever. Trovix Audit exists precisely because insurers need to prove to their regulators and their boards that their AI-assisted underwriting decisions are not just faster, but more defensible than the manual alternative.
What should a mid-market insurance firm do starting today? First, map the new risk types coming into your book—AI infrastructure projects, their variants, their phase-specific exposures. Second, audit your current underwriting process for bottlenecks that are not about human judgment but about information gathering, cross-referencing, and compliance checking. Third, invest in AI integration that is built for regulated environments, not AI that is retrofitted from the consumer tech world. Fourth, ensure that your claims and compliance teams understand how your underwriting AI works, so they can operate it and defend it. Fifth, use Trovix Watch to stay ahead of emerging regulatory guidance on AI governance in financial services. The insurers who move first on systematic AI integration will capture the AI infrastructure book at profitable terms. The ones who wait will be managed into capacity constraints that management will eventually blame on market conditions rather than operational choice.
Source: Bloomberg News