Industry guides
AI for insurance underwriting: submission intake, risk scoring and explainable pricing data on a governed lakehouse
Underwriters spend much of their time collecting and re-keying data from broker submissions. On DataNature, DocuContext extracts applications, schedules of values and loss runs, AI360 Studio scores appetite and risk using internal and third-party data, and a Voir underwriting assistant drafts a summary with citations. Every feature, model and decision is versioned and traceable in DataNature Atlas, which supports explainability expectations such as the NAIC AI bulletin and, for life and health pricing, the EU AI Act's high-risk rules.
Published . Last updated
On this page
Underwriting speed wins business, and underwriting quality protects the loss ratio. Both suffer when underwriters spend hours turning emails and spreadsheets into usable data. This guide shows how DataNature supports underwriting from submission to bind, with models that can be explained.
Where does underwriting time go?
- Reading broker emails and attachments to find the essentials.
- Re-keying applications, schedules of values and loss runs.
- Looking up prior policies, claims and third-party data.
- Checking appetite and authority before real underwriting starts.
How does the AI-supported workflow run?
- Intake. Submissions arrive by email or portal. DocuContext classifies attachments and extracts insured details, locations, values, coverages and loss history with confidence scores.
- Clearance. The platform checks for existing submissions and policies for the same insured and broker.
- Enrichment. Internal history and approved third-party data (property, hazard, firmographic) join the submission in governed tables.
- Appetite and risk scoring. AI360 Studio scores fit with appetite, expected loss and complexity, and ranks the queue so the best opportunities are worked first.
- Underwriter review. A Voir assistant drafts a summary with citations to the source documents, flags missing information and answers questions about guidelines.
- Decision and feedback. The underwriter decides; outcomes and later losses feed back into the models.
What about portfolio and accumulation?
Underwriting decisions affect the whole book. With exposures in the lakehouse, insurers can see accumulation by peril and region, check concentration against appetite before binding large risks, and run catastrophe scenarios. See catastrophe exposure monitoring.
How are models kept explainable?
| Requirement | How DataNature supports it |
|---|---|
| Know what data a model used | Feature definitions and training snapshots are versioned; Atlas lineage traces them to source |
| Explain individual decisions | Scores are stored with their top contributing factors |
| Test for unfair outcomes | Evaluation results by segment are stored with each model version |
| Control changes | Model registry with approvals before promotion |
| Restrict sensitive data | Atlas masks health and personal fields by role |
In the EU, AI used for risk assessment and pricing of natural persons in life and health insurance is classified as high-risk under the AI Act, which brings data governance, logging and human oversight duties. In the US, the NAIC Model Bulletin expects documented governance and testing of AI used in regulated decisions. P&C pricing is generally outside the EU AI Act's high-risk list, but sound governance is still expected by supervisors. This is general information, not legal advice.
How do you measure the impact?
Quote turnaround time, hit and bind ratios, underwriter time per submission, the share of submissions declined early for appetite, and loss ratio by segment over time.
Frequently asked questions
Does AI make underwriting decisions?
No. It prepares data, ranks work and explains risk factors; underwriters make the decisions within their authority.
Can we use third-party data?
Yes, through approved connectors or the Marketplace, with access scopes and audit applied like any other data.
Is this only for commercial lines?
No. Personal lines benefit from the same intake and scoring patterns, and life and health underwriting benefits from document extraction of medical evidence, with stricter controls.
Sources
- Holland & Knight: the implications and scope of the NAIC Model Bulletin
- Horváth: is your insurance company ready for the EU AI Act?
- DataNature: underwriting submission agent
How does DataNature Atlas keep this secure and compliant?
One set of controls supports banking, privacy and AI rules at the same time. DataNature Atlas is the platform's governance module:
- Knows your data: a catalog with automated PII detection and classification across Bronze, Silver and Gold.
- Controls access: role-based grants plus attribute-based policies, row filters and column masks (null, redact, partial, hash, year only).
- Governs AI: Voir agents are scoped principals; masks, row filters and PII guardrails apply to every tool call and answer.
- Proves it: lineage from source to metric to decision, and an audit log of who or what accessed which data.
- Stays local: Atlas runs inside your deployment, so policies, lineage and logs never leave your jurisdiction.
These controls support GDPR, PIPEDA, Quebec Law 25, CCPA, HIPAA and sector rules such as OSFI E-23. Read DataNature Atlas in depth or see security and compliance.
Platform components in this article
Learn more about each DataNature product and module mentioned here.
Products
See it on your own data.
Book a demoDemos and quotes are handled by Dlytica, the company behind DataNature.