What problem does it solve?
Commercial and specialty insurers receive far more broker submissions than their underwriters can read. Each one arrives as an email with attachments in different formats: an application, a schedule of locations, several years of loss runs, financials and supplementals. Before anyone can judge the risk, someone has to clear it (is it a duplicate, is another broker already on it), check it against appetite, rekey the data into the rating and policy systems and pull third party data.
That work is slow and falls on expensive people. Senior underwriters triage their own inboxes, so in appetite business waits behind business the insurer will decline, broker turnaround slips and some submissions are never looked at. Insurers treat speed as part of what they sell to brokers: Kinsale presents service as a competitive advantage to investors and reports an average submission clearance time of 9 minutes.
- Hiscox says extracting key data from email submissions is a manual process that can typically take up to three days in today's insurance operating model.Hiscox and Google Cloud Collaborate on AI in lead underwriting for the London Market (2023)
- Paragon Insurance Group's CTO told Kalepa that one of its programs, receiving around 50,000 submissions a year, was only looking at about 30% of its submissions.How Paragon Doubled Its Quote-to-Bind Rate and Achieved 99% Submission Accuracy with Kalepa (2026)
How does it work?
- Ingest. Submissions arrive in a shared mailbox, a broker portal or a placing platform. The system splits the email and attachments, classifies each document (application, loss run, schedule, financials) and reads scanned and native files.
- Extract to a schema. A model extracts the fields the underwriting workbench needs (insured, address, class of business, revenue, limits, loss history) into a fixed schema, with a confidence score and a pointer to the page each value came from.
- Clear and classify. The record is matched against existing accounts and open submissions to catch duplicates and broker conflicts, and the business is classified into the insurer's industry codes.
- Check appetite and enrich. Rules and models compare the risk to the written appetite and add internal history and approved third party data (firmographics, hazard scores, news).
- Rank and route. Each submission gets a priority (fit, likelihood to bind, broker importance) and goes to the right underwriter or team; clear declines are drafted for a human to confirm.
- Underwriter decides. The underwriter opens a decision ready file, corrects any extracted field (the correction is logged and fed back) and makes every quote or decline decision.
- Audience
- Back office
- Autonomy
- Supervised agent
- Adoption
- Early adopters
- Channels
- Email, API and system to system, Internal tools
What is it worth?
Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.
| KPI | Median | Reported range | Data points | Claimed by |
|---|---|---|---|---|
| Accuracy | Too few to pool | about 98% | 1 | 1 organization |
| Conversion uplift | Too few to pool | 2x | 1 | 1 organization |
| Cycle time reduction | Too few to pool | 50% | 1 | 1 vendor |
| Interactions handled | Not pooled | at least 15,000 | 1 | 1 vendor |
| Productivity gain | Too few to pool | 113% | 1 | 1 vendor |
| Cycle time | Not pooled | Not pooled: up to 1 days | 0plus 1 up to | 1 organization |
Value drivers: Speed and cycle time, Employee productivity, Revenue growth, Lower cost to serve.
Indicative value
A commercial insurer receiving 40,000 broker submissions a year
USD 320,000 to USD 1.9 million
Underwriting intake capacity released per year
How this is calculated
Formula: submissions * minutesPerSubmission / 60 * timeSavedShare * costPerHour. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Broker submissions per year submissions, submissions per year | 40,000 | 40,000 | The reference insurer. |
| Manual intake and triage time per submission minutesPerSubmission, minutes per submission | 20 | 40 | Editorial assumption for reading, clearance, appetite check and rekeying. Replace with a time study of your own intake. |
| Share of intake time the AI removes timeSavedShare, fraction of intake time | 0.4 | 0.7 | Editorial assumption. The evidence on this page measures elapsed time rather than effort (Sixfold reports a 50% cut in turnaround for Generali GC&C cyber submissions; AIG says submissions are prepared for underwriter review within one day, against a submission to quote process of about three to four weeks), so replace this with a time study from your own pilot. |
| Fully loaded cost of an underwriting or assistant hour costPerHour, USD per hour | 60 | 100 | Editorial assumption. Replace with your own fully loaded cost. |
What it leaves out: Capacity released, not cash saved, unless headcount or outsourcing changes. It leaves out the usually larger effect of quoting more in appetite business faster (higher bind ratios), the cost of the platform and integration, and data licences for enrichment.
Market estimates (analyst estimates, not deployments)
- Evident reports that the 30 insurers in its AI Index for Insurance announced 37 new AI use cases in the second quarter of 2026, that underwriting and pricing was the fastest growing application area, and that insurers prioritized tasks that structure and route information, such as submission intake and triage, over pricing and portfolio management. Evident: Insurance Use Case Trends Q2 2026 (2026)
Who already uses it?
9 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
American International Group
United States · Insurance · 2025
At its March 2025 investor day, AIG presented AIG Underwriter Assistance, a generative AI solution in production in Financial Lines that extracts data from broker and agent submissions, augments it with AIG and approved third party data, summarizes each submission and ranks submissions by appetite and propensity to bind, after which the underwriter analyzes the output and quotes. AIG describes the prior submission to quote process as taking about three to four weeks, with underwriters unable to review every submission, and says the assistant prepares submissions for review within one day. AIG frames the build around a human in the loop principle and was extending the same components to claims.
- Cycle time: up to 1 days, submission to underwriter ready file, in production lines
"AIG Underwriter Assistance Synthesizes and Prepares Submissions for Underwriter Review Within One Day"
Claimed by: organization
CNA Financial
United States · Insurance · 2025
In its fourth quarter 2025 earnings remarks, CNA said it had deployed a number of AI solutions across underwriting, claims and the back office over the past year and rolled out generative AI tools to every employee. Management reported faster triage, better submission responsiveness and measurable time savings, and framed further investment around risk selection, service quality and efficiency. No figures were disclosed.
No outcome disclosed.
Kinsale Capital Group
United States · Insurance · 2025
Kinsale, a US excess and surplus lines insurer that sources about 95% of its premium through wholesale brokers, told investors in January 2026 that AI driven routing improves the accuracy of submission routing and underwriter productivity, alongside an average submission clearance time of 9 minutes. Its 2025 annual report says it gave every employee an enterprise AI tool licence in 2025, that use is most prevalent in its IT, actuarial and analytical teams with selective use in underwriting, and that it also uses internally developed agents. Kinsale does not attribute the clearance time to AI, so no metric is recorded.
No outcome disclosed.
Skyward Specialty Insurance Group
United States · Insurance · 2025
Skyward Specialty announced in December 2025 that its partnership with Sixfold was entering its second year. The platform preprocesses submissions and generates recommendations on prioritization, appetite alignment and risk summarization and assessment, while underwriters stay in the loop to apply their judgment. The platform is live across six business units and more than 10 product lines, with an average deployment timeline of 8 to 10 weeks. The company presents the partnership as a step toward fully AI powered underwriting across its US property and casualty lines; no outcome figures were disclosed.
No outcome disclosed.
Hiscox
United Kingdom · Insurance · 2024
Hiscox London Market combined its own Hiscox AI Laboratories (Hailo) with Google Cloud's Gemini model to automate lead underwriting from email submission to quote in its sabotage and terrorism line. In scope risks are assessed by the model and the process generates an email to the broker with pricing and other data completed, ready for underwriter review. After a December 2023 proof of concept, in which Hiscox said the manual extraction step can take up to three days and quotes could be produced within three minutes, the model went live in August 2024. It initially covers renewals of existing US and Canadian sabotage and terrorism risks, excluding the New York and Chicago metro areas.
No outcome disclosed.
Generali Global Corporate & Commercial
Italy · Insurance · 2025
As its cyber book grew, Generali GC&C chose Sixfold as its first external AI partner and connected it to its cyber data sources and scoring system, so that all available risk information is structured in a dashboard for the underwriter against Generali's own guidelines. The vendor reports that over 90% of underwriters adopted the platform, that most cyber submissions now go through it with turnaround times for distribution cut by 50%, and that risk engineering reports take a few hours instead of about two days. Its Global Head of Operations and IT, Matthew Richardson, is quoted as saying that Sixfold's input is now required for every quote.
- Cycle time reduction: 50%, turnaround for distribution channels on cyber submissions
"Today, most Cyber submissions are accelerated through the Sixfold solution, cutting turnaround times for our distribution channels by 50%."
Claimed by: vendor - Employee adoption: at least 90%, cyber underwriters
"The results were immediate: over 90% of underwriters actively adopted the platform, reporting consistently high accuracy scores."
Claimed by: vendor
Paragon Insurance Group
United States · Insurance · 2025
Paragon runs about 25 specialty programs, some of which receive upward of 50,000 submissions a year, and its underwriters could review only about 30% of incoming submissions. With Kalepa, every submission is ingested, cleared and ranked by fit and likelihood to bind within minutes, with research from news, legal filings and third party data attached. Paragon's CTO says extraction is now around 98 to 99% accurate, better than the former manual operations team, and its E&S president says the quote to bind ratio doubled within the first year.
- Accuracy: about 98%, submission data extraction, after full inbox rollout
"We're somewhere around 98 to 99% accurate now - even more accurate than when we had an operations team doing this manually."
Claimed by: organization - Conversion uplift: 2x, quote to bind ratio, first year
"We're seeing a better quote-to-bind ratio. In the past year it has doubled from what it was before."
Claimed by: organization
AXIS Capital
Bermuda · Insurance · 2024
AXIS underwriters spent much of each new submission classifying the applicant into the right industry and writing a snapshot of its operations. AXIS started Sixfold with an underwriter facing dashboard and, after the pilot, integrated it into its automated clearance process, where it applies industry codes and matches cases against AXIS risk appetite. The vendor reports more than 15,000 applications analyzed in the first month and an implementation of about six weeks. AXIS's 2025 annual report separately describes AI tools used to empower underwriters and an AI Underwriting Working Group that monitors AI activity affecting underwriting.
- Interactions handled: at least 15,000, applications analyzed in the first month
"In the first month, AXIS analyzed more than 15,000 applications using Sixfold’s AI."
Claimed by: vendor
Markel
United Kingdom · Insurance · 2023
Before the change, the most senior underwriter in each Markel UK team triaged every incoming submission against appetite, and underwriters rekeyed risk data into several systems and pulled third party data by hand. With Cytora, broker submissions are digitized, enriched with external data, prioritized against Markel's underwriting and distribution strategy and routed to the right specialist as decision ready risks, with data flowing into the CRM and policy systems. The vendor reports a 113% productivity uplift (GWP per FTE) and a quote turnaround SLA for strategic partners cut from 24 hours to 2 hours.
- Productivity gain: 113%, gross written premium per underwriting FTE
"an uplift of 113% in productivity (GWP/FTE) in their underwriting teams"
Claimed by: vendor
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- A written appetite per line of business, precise enough to encode as rules
- A target data schema for each line (the fields the rating and policy systems need)
- A labelled sample of past submissions with the correct extracted values and outcomes (quoted, declined, bound)
- Account and broker master data for clearance and duplicate checks
Systems to integrate
- Submission mailboxes, broker portals or placing platforms
- Underwriting workbench or CRM where the file is opened
- Rating engine and policy administration system
- Approved third party data providers (firmographics, hazard and property data)
- Document storage for the original submission and extraction audit trail
Complexity: Medium
Extraction from messy broker documents now works well enough to use; the work is in the schema, the clearance logic, the appetite rules and the write back into the underwriting workbench and policy system. Reported rollouts are measured in weeks per team: Sixfold's case study gives about six weeks for AXIS, Skyward Specialty reports an average deployment timeline of 8 to 10 weeks, and Paragon had its full submission inbox running through Kalepa within 90 days.
- 1
Start with one line and its inbox
Pick a high volume line with a clear appetite (small commercial property, cyber, E&S casualty) and measure today's baseline: submissions per week, share reviewed, time to first response, quote and bind ratios.
- 2
Fix the schema before the model
Agree with underwriters which fields matter and how they are defined. Extraction quality is judged against this schema, so vague fields produce endless disputes about accuracy.
- 3
Measure field accuracy on real submissions
Run the extractor on a few hundred historical submissions and compare with the values underwriters actually used. Set a confidence threshold per field below which a human must confirm.
- 4
Encode appetite and clearance as reviewable rules
Keep hard rules (excluded classes, territories, limits) deterministic and visible to underwriting management; use models only for ranking and propensity, not for silent declines.
- 5
Put the output where underwriters work
Deliver the enriched, ranked file inside the existing workbench with a link to the source page for every value. A separate screen gets ignored.
- 6
Close the loop
Log every correction underwriters make and every quote, decline and bind outcome, and use them to retrain extraction and ranking each month.
Guardrails
- No automatic declines at launch; declines are drafted by the system and confirmed by an underwriter
- Every extracted value carries its source page and a confidence score
- Appetite rules owned and signed off by underwriting management, with version control
- Only approved third party data sources for enrichment, with licence and use conditions recorded
- Personal data in submissions (named individuals, claimant details) masked before it reaches a model that does not need it
KPIs to instrument
- Share of submissions reviewed within one business day, before and after
- Field level extraction accuracy on a weekly audited sample
- Time from submission receipt to first broker response
- Quote ratio and bind ratio for top ranked versus lower ranked submissions
- Underwriter corrections per submission
Human in the loop
Underwriters make every quote, decline and referral decision and confirm low confidence fields. Underwriting management owns the appetite rules and reviews a weekly sample of auto ranked and declined submissions to check that good business is not being buried.
Common failure modes
- Accuracy measured on the wrong thing
- A single overall accuracy number hides that the fields that drive pricing (revenue, total insured value, loss history) are the ones that fail. Measure accuracy per critical field.
- Ranking that encodes yesterday's appetite
- A propensity model trained on past binds keeps favouring classes the insurer is trying to exit. Retrain after every appetite change and let underwriting override the ranking.
- Silent declines and broker damage
- Automated declines sent without a human check lose broker trust quickly when one is wrong. Keep a person on every decline until the error rate is proven low.
- Workbench integration left for later
- If extracted data still has to be copied into the policy system, the time saving disappears. Plan the write back from the start.
What are the risks and rules?
EU AI Act
Minimal risk
Intake and triage for commercial insurance is not listed in Annex III, which covers risk assessment and pricing of natural persons in life and health insurance. It moves up to high risk only if the same pipeline is used to assess or price life or health cover for individuals.
Guidance
- Opinion on Artificial Intelligence governance and risk management (European Insurance and Occupational Pensions Authority, Europe). Sets out how existing insurance legislation (governance, risk management, data quality, human oversight) applies to AI systems that are not high risk under the AI Act.
- Annex III, high risk AI systems referred to in Article 6(2) (European Union, Europe). Point 5(c) lists life and health insurance risk assessment and pricing of natural persons; commercial lines triage is outside it.
Controls to put in place
- AI inventory entry per line with an accountable underwriting owner
- Documented appetite rules with version history and sign off
- Audit trail linking each ranked or declined submission to the data and rule versions used
- Weekly quality sample of extraction and triage outcomes
- Third party and model provider risk assessment under DORA for hosted AI services
Frequently asked questions
- How accurate is AI extraction from broker submissions?
- Good enough to use when measured per field and checked on low confidence values. Paragon's CTO says its submission extraction is around 98 to 99% accurate with Kalepa, better than its former manual team. That is a single figure from one deployment; accuracy can differ between applications, schedules and scanned loss runs, so measure it per field and per document type.
- Should the AI decline submissions on its own?
- Not at first. AIG and Hiscox describe AI that extracts, enriches or prices while an underwriter reviews the output before anything goes to the broker, and Skyward Specialty says its Sixfold platform ranks and assesses submissions while keeping underwriters in the loop. Hard appetite rules can draft declines, but a person should confirm them until the error rate is proven.
- What does it do for brokers?
- Faster answers. Sixfold reports that Generali GC&C cut turnaround for its distribution channels on cyber submissions by 50%, and AIG says its assistant prepares submissions for underwriter review within one day instead of a process it described as taking weeks.
- Is submission triage high risk under the EU AI Act?
- For commercial lines, no: Annex III covers life and health insurance risk assessment and pricing for natural persons. Insurance supervisors still expect governance, data quality and human oversight under existing rules, as EIOPA's 2025 opinion on AI governance sets out.
How to cite this page
Blits.ai AI Use Case Library, "AI for commercial underwriting submission intake and triage", last verified 27 September 2026, https://www.blits.ai/ai-use-cases/commercial-underwriting-submission-triage. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 27 September 2026: First published