AI in Insurance

The $1T insurance industry is straining under legacy systems, mounting data complexity, and rising customer expectations. Over 60% of operational costs still go to manual processes. 80% of valuable data, from adjuster notes to PDFs, remains underutilized. With half the workforce expected to retire by 2036, insurers are under pressure to modernize.

Insurance is a data business that has historically been terrible at using data. Underwriting relies on actuarial tables that are decades old. Claims processing is manual, slow, and fraud-prone. Distribution is fragmented across broker networks that resist digitization.

The cost of this inertia is staggering:

• Claims leakage (fraud, overpayment, process errors) costs the industry $80B+ annually in the US alone
• Underwriting loss ratios continue to deteriorate in lines like commercial property and specialty risk
• Customer satisfaction scores for insurance remain among the lowest of any industry

Meanwhile, 80% of the data that could inform better decisions, adjuster notes, medical records, satellite imagery, IoT sensor feeds, sits unstructured and unanalyzed.

• LLMs unlock unstructured data. For the first time, insurers can extract signal from adjuster notes, claim narratives, medical reports, and policy documents at scale, without armies of analysts.
• Core system replacement cycle. Legacy policy administration systems (PAS) built in the 1980s and 90s are reaching end-of-life. Carriers are actively looking for modern replacements.
• Workforce cliff. Half of the insurance workforce is expected to retire by 2036. AI is the only viable path to maintaining operational capacity.
• New risk categories. Cyber, climate, and parametric insurance don't fit legacy models. Native AI systems are better suited to price and manage these risks.

• AI underwriting engines. Systems that ingest heterogeneous data sources, structured and unstructured, to produce faster, more accurate risk assessments. Particularly valuable in commercial lines where submissions are complex.
• Claims automation. End-to-end claims handling from FNOL (first notice of loss) through settlement, with AI triage, fraud detection, and automated document review.
• Distribution intelligence. Tools that help MGAs and brokers identify, qualify, and bind risks faster, with AI-powered quote-to-bind workflows.
• Customer experience AI. Conversational agents that handle policyholder inquiries, policy changes, and renewals without human intervention.

Winning insurance AI companies are building on three layers:

• Data ingestion and normalization. Parsing and structuring unstructured documents (PDFs, images, handwritten notes) into machine-readable formats. Computer vision and OCR are table stakes; the differentiation is in accuracy and throughput.
• Risk intelligence engine. ML models trained on carrier-specific loss data, combined with LLM reasoning over claim narratives and underwriting submissions.
• Workflow orchestration. Integration with existing claims management systems, policy admin platforms, and broker portals, the connective tissue that lets AI outputs drive actual decisions.

The insurtech wave of 2015–2020 largely failed because it tried to become full-stack carriers without solving the underlying data and workflow problems. The current wave is different: it's targeting the operational infrastructure that carriers and MGAs already need.

Key players include:

• Legacy vendors (Guidewire, Duck Creek, Majesco), entrenched but slow, with open APIs creating integration opportunities
• Point-solution AI (Gradient AI, Tractable, Shift Technology), narrow but proven in specific workflows
• Next-gen carriers (Openly, Pie, Ethos), building greenfield on modern stacks

The biggest white space is in the mid-market MGA and specialty carrier segment, too large for point solutions, too nimble to wait for legacy vendor roadmaps.

We believe this moment marks a fundamental shift: from rule-based systems to risk-based, AI-native infrastructure that embeds intelligence across the insurance value chain.

We fund founders building deep, defensible systems that make insurance faster, more adaptive, and radically more efficient. Our focus is on companies that can land in a specific workflow (underwriting, claims, distribution) and expand to own the carrier relationship, not feature vendors, but infrastructure builders.