What problem does it solve?
Where new claims are routed with a handful of rules and a queue, a simple glass claim and a complex injury claim can wait in the same line, experienced handlers spend time on claims that need no judgment, and claims can reach a handler without the authority to settle them. Every claim is read from scratch: emails, estimates, invoices, medical reports and photos, and complex claims can carry thousands of pages of expert evidence.
The cost shows up as slow settlement of easy claims, and as leakage on hard claims that did not reach the right specialist early enough. Straight through processing has been an ambition for years, but rules based automation stalls on unstructured documents, so many simple claims still pass through a handler.
How does it work?
- Read the claim. The AI extracts the facts from the intake record and every attached document (estimates, invoices, reports, photos) and writes a short claim summary.
- Check cover and completeness. It compares the loss with the policy wording and limits and lists what is missing before anyone starts work.
- Score and segment. Models score complexity, expected severity, fraud risk and recovery potential, and a rule table turns the scores into a handling path.
- Settle the simple ones. Claims inside strict limits (claim type, amount, clean fraud score, confirmed cover) are approved and paid automatically, or declined only where a rule is unambiguous and the decision is explained.
- Route the rest. Other claims go to the handler with the right skill, authority and capacity, with the summary, open questions and suggested next steps attached, and the AI keeps checking as new information arrives.
- Audience
- Back office
- Autonomy
- Supervised agent
- Adoption
- Early adopters
- Channels
- 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 |
|---|---|---|---|---|
| Automation rate | Too few to pool | about 55% Not pooled: up to 70% | 1plus 1 up to | 1 organization |
| Cycle time | Not pooled | Not pooled: up to 4 days | 0plus 1 up to | 1 organization |
Value drivers: Lower cost to serve, Speed and cycle time, Customer experience, Employee productivity, Risk and loss reduction.
Indicative value
A personal lines insurer that settles 200,000 claims a year
USD 1.3 million to USD 7.9 million
Claims handling expense avoided per year
How this is calculated
Formula: claims * stpShare * costPerSimpleClaim + claims * (1 - stpShare) * otherClaimsMinutesSaved / 60 * costPerHour. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Claims per year claims, claims per year | 200,000 | 200,000 | The reference insurer. |
| Share of claims newly settled straight through stpShare, fraction of claims | 0.1 | 0.3 | Conservative against the evidence on this page (Lemonade reports roughly 55% of claims automated end to end), because an incumbent starts from legacy systems and a broader mix of claim types. |
| Internal handling cost of a simple claim today costPerSimpleClaim, USD per claim | 40 | 100 | Editorial assumption covering handler time, checks and payment; replace with your own claims expense data. |
| Handler minutes saved on each remaining claim by the summary and routing otherClaimsMinutesSaved, minutes per claim | 5 | 15 | Editorial assumption; replace with a time study. |
| Fully loaded cost per handler hour costPerHour, USD per hour | 35 | 55 | Editorial assumption, replace with your own. |
What it leaves out: Handling expense only. It leaves out leakage reduction from better routing, customer retention from faster settlement, the risk of paying claims that a person would have questioned, and the cost of the platform, models and integration.
Who already uses it?
7 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
Allianz Partners
United States · Insurance · 2026
Allianz Partners' director of partnerships told a US travel advisor consortium in April 2026 that AI assistance now handles 65 to 70% of all its claims. She said this has cut claims turnaround time from about 14 days, the average for the consortium's member agencies, to three to four days, with some claims turned around in a matter of hours.
- Automation rate: up to 70%, Share of claims using AI assistance, stated at Signature Travel Network's Horizon Club, April 2026
"We are right now using AI assistance for 65 to 70% of all of our claims, which has brought our claims turnaround time down from about 14 days, which was the average for Signature partners, to three to four days,"
Claimed by: organization - Cycle time: up to 4 days, Claims turnaround time with AI assistance, for Signature Travel Network member agencies
"We are right now using AI assistance for 65 to 70% of all of our claims, which has brought our claims turnaround time down from about 14 days, which was the average for Signature partners, to three to four days,"
Claimed by: organization
Travelers
United States · Insurance · 2026
In its 2025 annual report to shareholders, Travelers says it launched a natural language generative AI voice agent that takes first notice of loss by phone, used first for auto damage claims and planned to expand to more lines of business and claim interactions. The same letter reports that more than half of all claims are eligible for straight through digital processing, which customers choose about two thirds of the time, and that another 15% of claims are handled with advanced digital tools. No outcome figures for the voice agent itself were published; the company describes early adoption and feedback as positive.
No outcome disclosed.
Lemonade
United States · Insurance · 2025
Lemonade's claims bot AI Jim takes the first notice of loss in a chat with the customer, pays or declines simple claims within seconds and assigns the claims it may not settle, or has concerns about, to human claims experts based on their specialty, workload and schedule. A separate system, Forensic Graph, uses machine learning to predict, detect and block fraud across the customer engagement. The annual report states that AI Jim took the first notice of loss without human intervention 96% of the time and that roughly 55% of claims were automated from start to finish, both as of December 31, 2025.
- Containment rate: 96%, First notice of loss taken without human intervention, as of December 31, 2025
"AI Jim is our claims bot, and, as of December 31, 2025, 96% of the time, it is AI Jim that will take the first notice of loss from a Lemonade customer without human intervention"
Claimed by: organization - Automation rate: about 55%, Share of claims automated end to end, as of December 31, 2025
"As of December 31, 2025, roughly 55% of our claims were automated, resulting in instant or near-instant processing from start to finish."
Claimed by: organization
Sedgwick
United States · Insurance · 2025
Sedgwick, a global claims administrator, integrated Sidekick Agent into the workflows and screens of its own claims management systems. Built on Azure OpenAI Service and Azure AI Document Intelligence, it gives examiners claim insights, the day's top priorities, forecasts of anticipated claim trajectories, analytics on claim durations and reserves, and guidance on the next steps in the claim life cycle, and supports quality assurance for consistency and compliance. It follows an earlier version of Sidekick built on ChatGPT technology. No outcome figures were published.
No outcome disclosed.
Hiscox
United Kingdom · Insurance · 2025
Hiscox is rolling out Microsoft 365 Copilot to its more than 3,000 employees after a trial. A senior technical claims underwriter in the UK claims team uses it to identify and record the key information of a new claim, to summarise long expert medical evidence and legal advice, and to pull the progress of a claim from several emails and compose an update to a broker or customer. He says that recording a new claim now takes him as little as 10 minutes instead of up to an hour. This is one user's experience, not a measured program result.
No outcome disclosed.
Tokio Marine & Nichido Fire Insurance
Japan · Insurance · 2025
Tokio Marine & Nichido Fire Insurance uses Shift Technology's claims intake and claims fraud detection solutions, extended with generative AI that extracts data from structured and unstructured sources such as images and documents. The system highlights the points handlers should check for consistency across estimates, damage photos and claim statements, which makes reviews more efficient and more standardized, including during the surge of claims after large disasters, and it helps detect suspicious claims, which tend to rise after such events. No figures were published.
No outcome disclosed.
Admiral Seguros
Spain · Insurance · 2021
Admiral Seguros, the Spanish business of Admiral Group, sends motor claimants a link to a web app in which they photograph the damage, and Tractable's AI produces the repair estimate within minutes. Tractable reports that Admiral Seguros processed 12,000 touchless claims this way in 2021, that 90% of claim estimates were processed without human appraisers and that 98% of claims were completed in less than 15 minutes.
- Interactions handled: 12,000, Touchless claims in 2021
"In 2021, Admiral Seguros processed 12,000 touchless claims using Tractable AI."
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
- Historical claims with handling path, outcome, severity and leakage findings to train and test the scores
- Policy wordings, limits and excesses in a form the system can check against
- Handler skills, authority levels and capacity for routing
- Labelled documents per claim type for extraction testing
Systems to integrate
- Claims management system for claim data, reserves, tasks and payments
- Policy administration system for cover, limits and excess
- Document intake and extraction for estimates, invoices and reports
- Fraud detection and subrogation models as inputs to the routing rules
- Payment system for automated settlement
Complexity: High
Routing on a summary is achievable quickly; paying claims without a person is not. Straight through settlement needs reliable document extraction, machine readable policy wording and limits, fraud screening in the same flow, payment integration and a governance model that claims, compliance and actuarial teams accept.
- 1
Map today's handling paths and their cost
Take a year of claims and group them by type, severity and handling path. The simple, high volume groups are the straight through candidates; the complex ones are where routing and summaries pay off.
- 2
Summaries and routing before settlement
Start with claim summaries and routing suggestions that handlers accept or correct. It builds the labelled data and trust you need before any claim is paid without a person.
- 3
Define the straight through envelope
Write down per claim type the maximum amount, required documents, cover checks and fraud score threshold for automatic settlement, and have claims, compliance and actuarial sign it off.
- 4
Put fraud and recovery checks in the same flow
Every claim on the automatic path must pass the fraud and subrogation checks first; speed must not open a door for fraud or leave recoveries on the table.
- 5
Run in shadow mode, then widen
Let the system decide in parallel with handlers for a period, compare outcomes claim by claim, and switch on automatic settlement one claim type at a time.
- 6
Audit a sample for ever
Keep reviewing a random sample of automatically settled claims, and watch leakage and complaint trends per claim type.
Guardrails
- Automatic settlement only inside a signed off envelope per claim type (amount, documents, cover and fraud score)
- Automatic declines only where a rule is unambiguous, with an explanation and a route to a person
- Every automated decision logged with the inputs, scores, rules and model versions used
- Fraud and recovery screening before any automatic payment
- Handlers can override any routing or decision, and overrides feed back into testing
KPIs to instrument
- Share of claims settled straight through, per claim type
- Time from first notice of loss to payment for straight through and routed claims
- Share of routing suggestions handlers accept, and reasons for overrides
- Leakage found in audits of automated settlements
- Complaints and reopened claims after automated decisions
Human in the loop
Handlers work every claim outside the envelope and can reopen any automated decision. A quality team reviews a random sample of straight through settlements every week, and claims leadership approves each change to the envelope, the rules or the models.
Common failure modes
- Speed opens a door for fraud
- Automatic payment is exactly what fraudsters look for. Keep fraud scoring in the same flow and limit amounts per claim type.
- Extraction errors become payment errors
- A misread invoice total is paid automatically. Cross check extracted amounts against estimates and limits, and hold claims with low extraction confidence.
- Automated declines without recourse
- Customers receive a decline they cannot challenge. Explain every decline, avoid automatic declines where cover is a judgment, and give a route to a person.
- Routing that ignores capacity
- The best handler for every complex claim ends up with a queue of hundreds. Route on skill, authority and current workload together.
What are the risks and rules?
EU AI Act
Depends on design
Claims handling as such is not listed in Annex III. The same system becomes high risk when it is also used for risk assessment and pricing of natural persons in life and health insurance (point 5(c)), or when it is used by or on behalf of a public authority to grant, reduce, revoke or reclaim essential public assistance benefits and services, including healthcare services (point 5(a)). Otherwise the tier is minimal, so the design and the operator decide. Decisions on claims based solely on automated processing are also subject to Article 22 of the GDPR and the UK GDPR.
Rules that apply
Guidance
- Opinion on Artificial Intelligence governance and risk management (European Insurance and Occupational Pensions Authority, Europe). Addressed to national supervisors (August 2025); clarifies how existing insurance sector legislation applies to AI systems, with a risk based and proportionate approach to governance, fairness, explainability and human oversight.
- Annex III, high risk AI systems referred to in Article 6(2) (European Union, Europe). Lists life and health insurance pricing and risk assessment and public benefit eligibility as high risk, which matters when triage is used for those lines or schemes.
Controls to put in place
- Signed off straight through envelope per claim type, under change control
- Decision log with inputs, scores, rules and model versions for every automated settlement
- Weekly random audit of automated settlements with leakage and fairness checks
- Model validation and monitoring for drift in complexity, fraud and severity scores
- Customer route to a person for every automated decision
Frequently asked questions
- What share of claims can be settled straight through?
- It depends on the lines of business and the systems. Lemonade reports that roughly 55% of its claims were automated from start to finish at the end of 2025, and Travelers says more than half of its claims are eligible for straight through digital processing. Starting with a narrow set of simple claim types and widening from there limits the risk.
- Should an AI decline claims automatically?
- Rarely. Lemonade already pays or declines simple claims automatically within seconds; automatic declines are safe only where a rule is unambiguous, and the customer must get an explanation and a route to a person. Where cover is a matter of judgment, the AI should prepare the file and a handler should decide.
- Where does triage pay off if straight through settlement is not yet possible?
- In handler time and routing quality. Sedgwick gives examiners priorities, claim forecasts and next step guidance inside its claims systems. At Hiscox, a senior technical claims underwriter says that with Microsoft 365 Copilot, identifying and recording the key information of a new claim now takes him as little as 10 minutes instead of up to an hour.
How to cite this page
Blits.ai AI Use Case Library, "AI for claims triage and straight through processing", last verified 26 September 2026, https://www.blits.ai/ai-use-cases/claims-triage-and-straight-through-processing. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 27 September 2026: First published