McKinsey says AI can reshape P&C insurance economics through workflow redesign

McKinsey estimates gross written premiums have risen at roughly 4.9 percent a year since 2005 to an estimated $8.3 trillion in 2025, while profits before tax have climbed only about 4.3 percent to roughly $580 billion. In P&C, that gap is where the software decision lives, because the carriers that squeeze more value out of submission intake, triage, pricing, claims, fraud, and service will take operating leverage that scale alone has not delivered.

Why P&C is the right place to start

P&C is packed with high-volume, workflow-heavy decisions, which makes it the cleanest proving ground for AI that actually changes economics. Generative AI is already beginning to reshape document-heavy work such as policy issuance, submissions, and parts of claims handling and adjusting. Personal P&C carriers can use gen AI to automate claims processing and improve fraud detection with advanced analytics. That matters because fraud is not a side issue in motor and P&C lines. It is a recurring margin leak, and better fraud management is a serious cost and profit opportunity, not a hygiene project.

The broader market backdrop is not flattering either. Growth has been driven mainly by rate increases, with limited expansion into new risks. After the pandemic accelerated digital change, carriers needed to radically transform operating models and cost structures to stay competitive. In other words, carriers cannot keep buying growth with rate, then hope the expense ratio fixes itself later.

The workflows that actually move the P&C P&L

Submission intake is the first place AI can stop being decorative. If the carrier is still asking people to read every broker packet, extract every clause, and manually route every referral, the economics are broken before underwriting starts. Generative AI can sort, summarize, and classify submission content so underwriters spend time on the accounts that need judgment, not on the ones that need clerical work.

Pricing and underwriting are the next layer. Pricing is a primary differentiator for long-term value generation in P&C insurance, which means AI has to support better segmentation, better risk selection, and faster refresh cycles. The point is not to let a model quote everything automatically; the point is to put better decision support in front of the underwriter so the carrier can price more precisely and walk away from weak business faster.

Claims is where the workflow story gets real because the process is full of handoffs, document chasing, and exception handling. Technology can simplify claims through telematics and AI, and CCC Information Services is one of the providers building around that idea. In a 2026 McKinsey interview tied to claims technology, CCC chairman and CEO Githesh Ramamurthy said CCC was simplifying claim processes through telematics and AI. The market is moving toward fewer disconnected tools, more event-driven decisions, and more automation at the point of loss.

Fraud and service belong in the same operating model. ZwillGen’s August 2025 work on AI in claims-focused rules focused on underwriting, claims handling, fraud detection, and customer engagement. Those are the places where bad automation can create bad outcomes quickly. If the carrier cannot route suspicious claims, flag leakage, and respond to customers without creating new bottlenecks, then AI just speeds up the wrong process.

What the software stack has to do differently

The carriers that get real payoff will not treat AI as a layer on top of old workflows. They will build around three investments: core systems, orchestration, and decisioning. Core systems keep policy, billing, and claims data clean enough to trust. Orchestration moves work across intake, underwriting, claims, fraud, and service without losing the thread. Decisioning is where rules, models, and human review meet at the point of action.

In P&C, AI works as workflow infrastructure rather than as a shiny add-on. The useful version of AI in P&C is not a chatbot buried on a portal page. It is a system that can read a submission, route it, flag exceptions, call a fraud score, trigger a review, and leave an audit trail that compliance can live with.

The operating model also has to include accountability. Clean data, model controls, clear ownership, and measurable business cases are basic requirements. Without them, AI projects turn into a stack of disconnected pilots, each one promising efficiency while leaving loss adjustment expense, underwriting expense, and leakage basically unchanged.

Governance is now part of the product decision

The regulatory environment has caught up with the technology curve. The National Association of Insurance Commissioners updated its artificial intelligence page on April 3, 2026, stating that AI is now used across many industries, including insurance. Once AI starts influencing underwriting, claims handling, fraud detection, and customer engagement, governance stops being a policy memo and becomes a system requirement.

That is also why the software buyer’s brief has changed. Carriers should ask whether a platform can explain its decisions, surface human review where it matters, and preserve evidence for audits and disputes. If a vendor cannot show how a model was used in a claim, a referral, or a fraud escalation, the carrier is taking on risk without getting enough operational upside in return.

The real target is profitable growth

McKinsey’s 2025 global insurance report projects profitability should stabilize and the industry should shift its attention to profitable growth. The winners will be the carriers that redesign work so AI handles the repetitive, document-heavy, and pattern-based tasks while people focus on judgment, negotiation, and exception management.

The old model was to digitize the front end and leave the back office untouched. The better model is to rebuild the workflow itself so submission, pricing, claims, fraud, and service all run through the same decisioning fabric.