Insurance underwriters

Screen applicants based on set criteria

Screening against defined criteria (credit score thresholds, age limits, health conditions, property characteristics) is a pure pattern-matching task. AI can apply rule sets and flag exceptions with near-perfect accuracy, requiring minimal human involvement.

BLS evidence: The duties section lists 'Screen applicants based on set criteria' as a task underwriters typically perform.

Use automated software to assess applicant risk and review recommendations

This task explicitly involves using automated software that already exists. AI can not only use such systems but increasingly generates the recommendations itself. The human role is reduced to reviewing batches of AI-flagged exceptions rather than processing each case.

BLS evidence: Underwriters 'use computer software to analyze risk' and 'take specific information about an applicant and enter it into a program,' then 'evaluate these recommendations and decide whether to approve or reject the application.'

Analyze information stated on insurance applications

AI systems excel at parsing structured and semi-structured application data, extracting relevant fields, and flagging inconsistencies or missing information. Modern LLMs can handle varied application formats and identify key risk indicators with minimal human review.

BLS evidence: The duties section lists 'Analyze information stated on insurance applications' as a primary task, and underwriters are described as evaluating insurance applications to decide whether to approve them.

Determine appropriate premiums and amounts of coverage

Premium calculation is highly algorithmic, based on risk scores, actuarial tables, and competitive positioning. AI can generate pricing recommendations end-to-end for standard cases. Humans primarily review outliers and adjust for market strategy, making this largely automated.

BLS evidence: The duties section explicitly states 'Determine appropriate premiums and amounts of coverage,' and the page notes that for approved applications, underwriters determine these amounts.

Determine the risk involved in insuring a client

AI risk models can process vast datasets of historical claims, actuarial tables, and applicant characteristics to generate risk scores. However, novel risk scenarios and edge cases still benefit from human judgment, placing this in the high-assistance range where AI does most work but humans validate.

BLS evidence: The duties section explicitly states 'Determine the risk involved in insuring a client,' and the page emphasizes that underwriters analyze risk factors appearing on applications.

Decide whether to offer insurance coverage to applicants

Final accept/reject decisions involve business judgment and risk appetite that companies may keep human-in-loop for accountability and regulatory comfort. AI can recommend with high accuracy, but the decision authority typically remains with humans who batch-review AI recommendations.

BLS evidence: The duties section lists 'Decide whether to offer insurance' as a primary task, and the page states underwriters 'evaluate insurance applications and decide whether to approve them.'

Evaluate complex insurance applications requiring analytical insight beyond automated recommendations

Complex cases requiring analytical insight beyond automation are precisely where human expertise remains valuable. However, AI can now handle increasingly sophisticated analysis (multi-factor risk correlation, unusual policy structures), making humans more productive even if still essential for final judgment.

BLS evidence: For specific and complex insurance types, such as workers' compensation, underwriters need to rely more on analytical insight beyond automated recommendations.

Contact field representatives, medical personnel, and others to obtain additional information

AI can draft inquiry emails and parse responses, but the interactive negotiation with field reps and medical personnel—especially when clarification or persuasion is needed—still requires human communication skills. AI assists substantially but humans remain load-bearing.

BLS evidence: The duties section states underwriters 'Contact field representatives, medical personnel, and others to obtain additional information,' and notes they may consult additional sources when decisions are difficult.

Travel to assess properties in person for property and casualty insurance

Physical site visits require travel, navigation of varied properties, assessment of structural conditions, and real-time judgment about hazards in unpredictable environments. While drones and computer vision can assist, the core task remains human-dependent.

BLS evidence: The page states 'Some property and casualty underwriters travel to assess properties in person.'