Insurance Underwriters
13-2053.00Median wage $81,370/yr105,420 employed (US)Rank #64 of 923 scored · top 7% by substitution
Review individual applications for insurance to evaluate degree of risk involved and determine acceptance of applications.
Sub-scores
0–100 · band = confidence interval from rater disagreement
Substitution55
Exposure57
Augmentation82
Substitution — the headline: capability discounted by cost, barriers and adoption.
Exposure — technical capability alone, regardless of whether anyone deploys it.
Augmentation — how much AI assists without replacing. High here + moderate substitution = a changing job, not a disappearing one.
Tasks on the substitution scale
7 rated tasks, binned by substitution score.
Position among all scored occupations
Distribution of 923 occupation scores; the marker is this occupation.
Tasks with substitution ≥ 70
0%
Run 1.0.0-draft.1 · computed 2026-08-05 · rater panel: claude-sonnet-5, claude-haiku-4-5-20251001 · intervals span rater disagreement.
Why this score
The five weighted dimensions of the composite, averaged across this occupation's tasks (importance-weighted, panel mean). Exact weights and formulas: /api/v1/methodology.
Task automatabilityw 35%59
panel mean rating 3.4/5 → substitution pressure 59/100
Technical feasibility todayw 20%52
panel mean rating 3.1/5 → substitution pressure 52/100
Cost vs. human wagew 15%66
panel mean rating 3.7/5 → substitution pressure 66/100
Adoption barriersw 20%inverted — strong barriers lower the score41
panel mean rating 3.4/5 (barrier strength) → substitution pressure 41/100
Sector adoption velocityw 10%57
panel mean rating 3.3/5 → substitution pressure 57/100
Task breakdown (7 tasks)
Substitution pressure per task, weighted by O*NET importance in the composite. Expand a task for the full rater audit trail — every rating, every model, every rationale.
Review company records to determine amount of insurance in force on single risk or group of closely related risks.
68CI 62–74 · exposure 70 · augmentation 88 · importance 4.0/5 · click for rater detail
Review company records to determine amount of insurance in force on single risk or group of closely related risks.
68| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Insurance is a high-digitization, information-intensive sector with rapid AI adoption; many major insurers have deployed automation for policy data extraction and risk aggregation in underwriting workflows. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Insurance is a data-heavy financial services sector with growing InsurTech adoption, but many carriers still operate on legacy systems limiting full automation of exposure aggregation. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI dramatically accelerates the routine record-review phase by flagging policy data and related risks, freeing underwriters to focus on judgment-heavy risk assessment and pricing decisions while maintaining human oversight of final determinations. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-powered dashboards and analytics tools significantly speed up underwriters' ability to review and aggregate risk exposure data, improving accuracy and turnaround time. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI can extract and analyze insurance policy data from company records with high accuracy using document processing and database queries, achieving significant time savings. However, determining the precise relationship between 'closely related risks' often requires contextual business judgment that may require human verification in edge cases. |
| Task automatability | claude-sonnet-5 | 4/5 | Retrieving and aggregating records to compute total insurance-in-force on a risk or group is a structured data lookup/aggregation task well suited to automation with existing databases and AI query tools, though edge cases in ambiguous risk grouping may require human judgment. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Regulatory oversight and audit trails require documented methodology and human sign-off on underwriting decisions; compliance frameworks and internal governance create friction, though no hard legal requirement prevents AI from performing the initial record review. |
| Adoption barriers | claude-sonnet-5 | 3/5 | While no license is required to compile records, underwriting decisions based on this data often need sign-off from a qualified underwriter due to liability and regulatory oversight of insurance risk assessment. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Document processing and database queries cost pennies per policy review compared to an underwriter's loaded wage of $50–70/hour, making AI deployment orders of magnitude cheaper at scale. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated data aggregation and reporting tools operate at a fraction of the cost of manual record review once integrated with policy administration systems. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed AI systems (OCR, NLP, document classification) reliably extract policy information from records in production insurance systems. Query-based lookup of insurance amounts is fully automated in many underwriting platforms, though complex risk grouping may still involve manual review. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Insurance policy administration systems and analytics tools already automate aggregation of exposure and accumulation data, but many insurers still rely on manual review for complex or legacy records, so reliability varies across the industry. |
Examine documents to determine degree of risk from factors such as applicant health, financial standing and value, and condition of property.
67CI 65–70 · exposure 75 · augmentation 100 · importance 4.5/5 · click for rater detail
Examine documents to determine degree of risk from factors such as applicant health, financial standing and value, and condition of property.
67| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Insurance is a digitized, information-intensive sector with strong financial incentives for automation; major carriers have deployed document AI and risk-scoring systems in production, with measurable displacement of routine review work occurring industry-wide. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Insurance is a data/finance-adjacent sector with substantial AI adoption in underwriting automation, InsurTech platforms, and algorithmic risk models already in production. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI dramatically augments underwriter productivity by pre-processing documents, flagging anomalies, and suggesting risk scores, allowing underwriters to focus judgment on complex or novel cases rather than manual document review, while the human remains the final decision-maker. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI substantially augments underwriters by pre-processing documents, flagging risk factors, and generating risk scores, letting humans focus on judgment calls and exceptions. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI can automatically extract and analyze key risk factors from applicant documents (health records, financial statements, property assessments) with high accuracy, meeting the 50% time-saving threshold. Document classification, risk scoring, and flagging of outliers can be largely automated, though final underwriting decisions often require human judgment on edge cases. |
| Task automatability | claude-sonnet-5 | 4/5 | Document review and risk scoring against structured criteria is largely automatable with AI/ML models trained on underwriting data, especially for standard lines of insurance; complex or edge cases still need human judgment. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Regulatory and liability requirements create material friction: insurance regulators often mandate human underwriter accountability, and errors in risk assessment carry significant financial consequences, so organizations retain human sign-off requirements even where AI could technically decide autonomously. |
| Adoption barriers | claude-sonnet-5 | 3/5 | Some regulatory requirements mandate licensed underwriter sign-off for certain policy types and there is liability exposure for wrongful denial or mispricing, creating moderate friction despite no blanket licensing requirement for every review step. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | Document review and risk analysis automation is substantially cheaper than human labor: a single AI system handles thousands of applications annually at per-task costs an order of magnitude below underwriter wages, factoring in infrastructure and oversight. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated document extraction and risk-scoring pipelines cost a small fraction per application compared to a loaded underwriter salary, particularly for high-volume standard policies. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed AI products (document intelligence platforms, risk-scoring engines) reliably perform document review and risk factor extraction in production insurance environments. Systems like OCR + classification + risk models work at scale, though most insurers still require human sign-off on final underwriting decisions rather than fully autonomous approval. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Automated underwriting engines and AI-based risk assessment tools are already deployed at scale in life, auto, and property insurance, though complex commercial or high-value cases still route to human underwriters. |
Decrease value of policy when risk is substandard and specify applicable endorsements or apply rating to ensure safe, profitable distribution of risks, using reference materials.
64CI 45–82 · exposure 70 · augmentation 75 · importance 4.0/5 · click for rater detail
Decrease value of policy when risk is substandard and specify applicable endorsements or apply rating to ensure safe, profitable distribution of risks, using reference materials.
64| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Insurance sectors have rapidly deployed AI underwriting systems at scale over the past 3–5 years; adoption is now mainstream in large carriers and digital-native insurers, though smaller firms and specialty lines lag. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Insurance is adopting AI-assisted underwriting tools steadily, particularly in P&C and life lines, but adoption is uneven and many substandard-risk decisions remain manually reviewed. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI effectively assists human underwriters by pre-screening risks, flagging anomalies, and suggesting endorsements or ratings, substantially raising productivity even when humans retain final decision authority. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially speeds up retrieval of rating rules, comparable risk data, and endorsement options, letting underwriters focus judgment on edge cases while automating routine lookups. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | AI systems can reliably perform risk assessment using structured data, reference materials, and underwriting guidelines to determine policy adjustments and endorsements, meeting the ≥50% time-saving threshold with minimal human intervention needed. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can retrieve rating rules and reference materials and suggest rating adjustments or endorsements, but final judgment on substandard risk pricing and endorsement selection often requires nuanced case-specific judgment and accountability that current systems only partially replicate. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | While underwriting automation is widespread, regulatory oversight requirements, compliance obligations, and organizational preference for human sign-off on material policy changes create moderate friction against full substitution. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Underwriting decisions on risk pricing and endorsements are typically subject to regulatory oversight, actuarial sign-off requirements, and licensing rules that constrain fully autonomous AI determination of policy terms. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | AI inference and integration costs for automated underwriting are negligible compared to the loaded hourly wage of human underwriters, delivering orders-of-magnitude cost savings per policy processed. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | Automated rating engines reduce marginal cost significantly for standard cases, but substandard risk analysis still requires licensed underwriter oversight, keeping blended costs only moderately below human-only processing. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed underwriting AI platforms used by major insurers can process risk profiles and generate policy modifications with high accuracy; however, edge cases and novel risk combinations still occasionally require human review, preventing a perfect 5. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Underwriting workbench tools and rules engines with AI-assisted rating exist in production at many carriers, but they still require underwriter review for substandard/complex risks rather than fully autonomous decisioning. |
Write to field representatives, medical personnel, or others to obtain further information, quote rates, or explain company underwriting policies.
59CI 51–67 · exposure 58 · augmentation 88 · importance 4.3/5 · click for rater detail
Write to field representatives, medical personnel, or others to obtain further information, quote rates, or explain company underwriting policies.
59| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Insurance remains a moderately regulated, risk-averse sector with strong preference for human underwriter accountability; pilot AI projects exist, but production adoption of fully autonomous underwriting communication remains limited and cautious. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Insurance is a digitized, document-heavy industry adopting AI writing tools and copilots at a moderate pace, though core underwriting communication workflows are still transitioning from pilots to full production. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI excels at drafting letter templates, pulling relevant policy language, and surfacing required information to request—allowing underwriters to review, personalize, and approve much faster than writing from scratch, significantly boosting productivity while preserving human control. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI drafting assistants substantially speed up composing informational requests and rate/policy explanation letters while underwriters retain control over final content and figures. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can draft routine communications requesting information or explaining standard underwriting policies with significant time savings, but personalized relationship-building, negotiating rates, and handling unusual circumstances require human judgment and contextual nuance that limit full automation. |
| Task automatability | claude-sonnet-5 | 4/5 | Drafting communications to request info, quote rates, or explain policies is largely templated correspondence that LLMs can generate accurately given underwriting guidelines, with human review for edge cases and accuracy of quoted figures. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Regulatory requirements (insurance law, data privacy, rate-justification documentation) and liability concerns over inaccurate rate quotes or policy misstatements create meaningful friction; underwriters often must personally sign off on communications, and professional relationships with agents and medical personnel favor human trust. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement to draft correspondence, but underwriters remain accountable for rate quotes and policy statements, creating some liability-driven review friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI-generated communications cost pennies per message in inference and oversight, while a skilled underwriter's time costs $50–100+ per hour; even with modest oversight overhead, AI is substantially cheaper for high-volume routine correspondence. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Generating routine correspondence via AI is far cheaper than having an underwriter compose each letter, though some oversight cost remains to check accuracy of quoted rates and policy details. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | LLM-based systems can generate templated letters and emails reliably in production, but actual deployment for sensitive underwriting communication is limited; most organizations still rely on human underwriters to ensure accuracy, compliance, and relationship management in real workflows. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | AI writing assistants and insurance-specific copilots are deployed for drafting correspondence, but ensuring accurate rate quotes and policy explanations still requires integration with underwriting systems and human verification, limiting fully autonomous reliability today. |
Evaluate possibility of losses due to catastrophe or excessive insurance.
57CI 28–87 · exposure 58 · augmentation 75 · importance 4.2/5 · click for rater detail
Evaluate possibility of losses due to catastrophe or excessive insurance.
57| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Insurance and financial services sectors show strong, documented adoption of AI for underwriting and risk assessment; major insurers have deployed automated loss evaluation systems, and the industry is actively integrating machine learning models into production pipelines. Adoption is faster and broader than most other professional services. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Insurance is a data-rich financial sector adopting AI/analytics steadily for risk modeling, but full automation of catastrophe judgment remains in pilot/augmented stages. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI systems augment underwriter productivity by rapidly summarizing catastrophe exposure, flagging high-risk profiles, and generating preliminary loss estimates, allowing humans to focus on complex judgment calls and exception handling. The tools materially accelerate the underwriter's analytical capacity while preserving oversight. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-driven catastrophe models and predictive analytics substantially enhance underwriters' ability to assess excessive-loss scenarios, though humans still make final risk determinations. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Modern AI systems can analyze historical loss data, catastrophe models, and risk factors to assess loss probability and severity with high speed and consistency, meeting the ≥50% time-saving threshold. Large language models and machine learning models trained on underwriting datasets can process vast amounts of structured and unstructured risk information faster than human underwriters while maintaining comparable accuracy on routine assessments. |
| Task automatability | claude-sonnet-5 | 2/5 | Catastrophe risk evaluation involves synthesizing geospatial data, climate models, actuarial tables, and judgment calls about correlated tail risks, which current AI can support but not fully replace end-to-end at equal quality.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | While regulatory oversight of insurance operations exists, underwriting decisions themselves are not legally required to be performed by licensed humans in most jurisdictions, and many insurers already use automated systems for initial loss assessment. The main friction is organizational conservatism and customer preference for human judgment on high-stakes decisions rather than hard legal barriers. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Underwriting decisions on catastrophic exposure often require licensed underwriter sign-off and carry major liability/regulatory exposure, creating strong barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | AI-driven loss evaluation costs per assessment (model inference, data integration, minimal human review) are typically one or more orders of magnitude cheaper than employing underwriters to manually evaluate catastrophic loss exposure and build-out risk profiles from scratch. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Catastrophe modeling platforms carry significant licensing, data, and integration costs, and still require skilled underwriter oversight, so total cost is not dramatically below human-inclusive workflows. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed products including specialized insurance AI platforms (e.g., catastrophe modeling software, automated underwriting systems) reliably perform loss evaluation at scale in production environments. While some edge cases and novel risk scenarios may require human oversight, mainstream catastrophe risk and excessive loss evaluation is demonstrably performed by mature systems in operational use. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Catastrophe modeling software (e.g., RMS, AIR) exists and is widely used, but these are decision-support tools requiring underwriter interpretation, not autonomous evaluators of loss possibility. |
Decline excessive risks.
41CI 31–51 · exposure 38 · augmentation 75 · importance 4.4/5 · click for rater detail
Decline excessive risks.
41| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Major property-and-casualty and life insurers have deployed AI underwriting tools in production for years, and adoption is accelerating across tier-1 and tier-2 firms. The sector is highly digitized, capital-intensive, and data-rich, making it one of the faster-adopting industries for automation in this process. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Insurance is adopting AI-driven risk models and automated underwriting at a moderate pace, with pilots and partial deployment common in personal lines but slower in complex commercial risk decline decisions. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI risk-scoring systems routinely assist human underwriters by surfacing key risk factors, automating data pull and analysis, and prioritizing cases for review. This significantly speeds underwriter productivity and consistency while humans retain discretion on final decline decisions. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI risk models, predictive analytics, and decision-support tools significantly help underwriters identify and quantify excessive risk faster, improving decision quality and speed while the underwriter retains final authority. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can flag and analyze risk factors automatically from structured policy data and historical claims, reducing time significantly. However, the final judgment on whether a risk is 'excessive' often requires contextual reasoning about market conditions, competitive dynamics, and nuanced human judgment, so end-to-end automation with equal quality is only partial. |
| Task automatability | claude-sonnet-5 | 2/5 | Deciding to decline an excessive risk requires synthesizing judgment about ambiguous, high-stakes situations, applying risk appetite and business context AI cannot fully internalize; while risk scoring can flag cases, the final decline decision is not yet reliably automatable end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Insurance underwriting decisions are heavily regulated under fair lending, anti-discrimination, and state insurance law; explainability and human accountability are required in many jurisdictions. Many insurers maintain legal and compliance requirements that a human underwriter must review or sign off on decline decisions, creating a hard barrier. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Underwriting decisions, especially declines, carry regulatory scrutiny (anti-discrimination, fair lending/insurance laws) and often require documented human rationale and licensed underwriter accountability, creating strong barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI risk models deployed in underwriting are significantly cheaper per decision than full human underwriter review, especially at scale. Integration and model maintenance costs are modest compared to the loaded cost of experienced underwriters for repetitive risk assessment. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI-assisted risk scoring reduces underwriter time per case, but human review and sign-off remain necessary, keeping all-in costs only moderately below fully manual underwriting. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | AI risk-scoring systems exist in production at major insurers and can reliably identify high-risk applicants based on underwriting models. However, the subjectivity of 'excessive' and the need to balance business strategy means tools flag candidates for human review rather than autonomously declining them at consistent quality. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Underwriting risk-scoring and rules-engine products exist and flag high-risk applications, but few organizations let AI autonomously issue final decline decisions without underwriter review due to error and liability concerns. |
Authorize reinsurance of policy when risk is high.
26CI 25–28 · exposure 25 · augmentation 75 · importance 3.9/5 · click for rater detail
Authorize reinsurance of policy when risk is high.
26| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Insurance remains a heavily regulated, human-centric sector with slow digital transformation in underwriting decisions. While analytical tools are adopted, autonomous authorization decisions face legal and organizational friction; adoption of full automation remains limited. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Insurance and financial services are moderately fast adopters of AI for risk modeling and underwriting support, though authorization decisions themselves see slower, more cautious adoption due to liability concerns. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI excels at augmenting underwriters through rapid risk scoring, historical pattern matching, and documentation retrieval, meaningfully accelerating the analysis phase while the underwriter retains authority and judgment on reinsurance authorization decisions. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can significantly enhance this task by rapidly synthesizing risk data, historical loss patterns, and market conditions to inform the underwriter's authorization decision. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Reinsurance authorization for high-risk policies requires judgment balancing regulatory constraints, actuarial data, and underwriting experience. While AI can score risk and flag candidates, final authorization demands human discretion and accountability that current systems cannot reliably replicate end-to-end at equal quality. |
| Task automatability | claude-sonnet-5 | 2/5 | Authorizing reinsurance on high-risk policies requires judgment, negotiation, and accountability that current AI cannot fully replicate end-to-end; AI can support risk analysis but not the authorization decision itself.$ |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Reinsurance authorization carries significant liability exposure and is typically bound by insurance regulations requiring licensed underwriters to sign off on high-risk decisions. Legal and regulatory frameworks mandate human responsibility and due diligence, creating hard barriers to full automation. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Reinsurance authorization involves significant liability, regulatory oversight, and typically requires a licensed/authorized underwriter's sign-off, creating strong structural barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI can reduce analysis and data-gathering costs but cannot eliminate the licensed underwriter, whose salary dominates the cost structure. Integration and oversight infrastructure further offset gains, keeping total cost comparable to or exceeding the human-only approach. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can cheaply generate risk analytics, but the authorization step still requires costly human expertise and liability acceptance, keeping overall cost comparable to human-driven processes. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | AI tools can assist in risk assessment and recommend reinsurance but do not autonomously authorize policies in production systems. Existing deployed products lack the legal standing and accountability framework to make binding authorization decisions without human underwriter review. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Products exist for risk scoring and data aggregation, but no deployed system autonomously authorizes reinsurance decisions in production; this remains a human sign-off step. |
Related occupations — Business & Financial Operations
How to read this
A high substitution score does not mean this job disappears — it means a large share of its current tasks face replacement pressure, so the mix of tasks is likely to change. High augmentation alongside substitution typically means the occupation reorganizes around the protected tasks. Wide confidence intervals mean the rater panel disagreed: treat those scores as open questions, not verdicts.
What would change this score
New model capabilities (automatability, feasibility), falling inference costs (cost ratio), regulation and licensing shifts (barriers), and measured sector adoption (velocity) all re-enter at every index release. Each release is recomputed, versioned and kept queryable — scores are claims with a date on them, not permanent labels.