Risk automation boosts underwriting accuracy to 90%

Insurers are reallocating AI spending from claims processing to continuous risk automation. Companies are using computer vision, telematics, connected sensors and predictive models to update risk profiles during a policy term. Some underwriting workflows saw data intake accuracy rise from roughly 75% to over 90% after adopting these tools.

Claims automation sped first notice of loss, fraud detection and photo-based damage assessment, increasing straight-through processing from about 10–15% to roughly 70–90% among heavy adopters. Those gains reduced processing times and operating costs but did not change how often losses occur or correct mispriced policies written on incomplete data.

Risk automation treats risk as a constant signal rather than a single snapshot. Underwriting systems now ingest live feeds from IoT devices, vehicle telematics, satellite imagery and external databases. Computer vision converts inspection photos into structured condition reports. Telematics record driver behavior and vehicle condition. Predictive analytics combine historical loss data with current signals to produce forward-looking risk scores.

AIG reported lifts in data intake accuracy from about 75% to over 90% on some underwriting workflows after applying AI to intake. Vendor case studies have shown machine-learning accuracy gains as large as 54% in specific uses. Consultant analyses indicate AI-driven underwriting improvements reduced loss ratios by about three percentage points in lines that incorporated unstructured condition data. Firms using advanced analytics recorded combined ratios roughly six points lower than slower adopters between 2022 and 2024.

Operational effects include faster decisioning and shorter cycle times. Automated underwriting that once required days can collapse to minutes in some workflows. Reported metrics show underwriting cycle time fell about 31% and risk assessment accuracy rose 43% for complex policies. For commercial fleets and large portfolios, continuous monitoring provides steady risk signals that underwriters can act on between renewals, enabling earlier interventions and pricing adjustments.

Vendors now produce photo-backed condition reports in minutes, replacing subjective inspector notes with structured, component-level findings. Those structured inputs feed models that predict deterioration or hazardous conditions before claims occur, changing how underwriting, claims and risk teams coordinate.

Regulatory and adoption signals are visible in industry surveys and deployment data. A 2025 survey of 93 health insurers found 84% using AI or machine learning in operations, with fraud detection a common use. Overall AI deployments in insurance grew 87% year over year, and agentic AI systems accounted for about one in five public deployments.

Consultants project insurers will need different skills, data priorities and organizational designs as firms adopt continuous risk intelligence rather than operating only on annual snapshots and reactive claims processes.