925 Techies
← All case studies Insurance

Claims settled in seconds, not weeks

How carriers and MGAs automate intake, triage and straightforward claims, and give adjusters better tools for the complex ones.

3 secLemonade's first fully automated claim payment
18Anti-fraud checks run on that claim
0Forms filled in by the customer

Figures: Lemonade, 2016. Source and caveats below.

01 The problem

A claim is the moment an insurance customer finds out what they paid for. Traditional handling involves forms, phone calls, manual document review and days or weeks of waiting, and adjusters spend much of their time on data entry for simple, low-value claims.

Slow claims cost more. They generate more calls, more complaints and more leakage, while fraud is hard to spot when each adjuster only sees their own files.

02 Where AI fits

A

First notice of loss

A conversational intake by chat or voice gathers the facts, photos and documents and creates the claim file without a form.

B

Straight-through processing

Low-value, low-risk claims are checked against the policy and paid automatically. Everything else is routed to an adjuster.

C

Document and image extraction

Invoices, police reports, medical bills and damage photos are read into structured data.

D

Fraud detection

Models score each claim across the whole book for patterns no single adjuster would see.

E

Adjuster and underwriter copilots

Policy-coverage questions answered with the clause cited, plus drafted file notes and customer letters.

03 Reference architecture

The layers we put in place for a typical insurance deployment, from your systems of record up to the people who review the output.

  1. L1Policy admin and claims system (Guidewire, Duck Creek)
  2. L2Omnichannel intake agent
  3. L3Document and image extraction
  4. L4Coverage rules and fraud scoring
  5. L5Auto-pay path and adjuster queue
  6. L6Decision log for regulators

04 Real-world implementation

Public deploymentLemonade2016

In December 2016 the insurer Lemonade reported that its claims bot, AI Jim, had reviewed a theft claim, cross-referenced it with the policy, run 18 anti-fraud algorithms, approved it and sent payment instructions to the bank in three seconds, with no paperwork. The company has since reported that a substantial share of its claims are handled without human involvement.

3 secLemonade's first fully automated claim payment
18Anti-fraud checks run on that claim
0Forms filled in by the customer

Source: Lemonade company announcement (2016). Lemonade is not a 925 Techies client.

What to keep in mind: The three-second figure is a single best-case claim, published by the company. Simple claims in personal lines are the best fit for full automation. Complex and high-value claims still need experienced adjusters.

05 Business impact

Loss adjustment expense

Each claim that goes straight through removes handling cost, and adjusters are concentrated on claims where judgment changes the outcome.

Cycle time

Faster settlement reduces follow-up calls, rental and storage costs and litigation rates.

Leakage and fraud

Consistent coverage checks and book-wide fraud scoring reduce overpayment.

Retention

A fast, fair claim is the strongest predictor of renewal.

KPIs we baseline and track
Straight-through rateClaim cycle timeLoss adjustment expense per claimFraud detection rateCustomer satisfaction

06 Implementation roadmap

  1. Weeks 1–3

    Assess

    Analyze claim types by volume, value and complexity to find the straight-through candidates.

  2. Weeks 4–8

    Prove

    Replay historical claims through the model and compare decisions and payouts with what adjusters did.

  3. Weeks 9–20

    Build

    Integrate intake and decisioning with the claims system. Start with a low payout ceiling.

  4. Ongoing

    Operate

    Audit a sample of automated decisions, raise the ceiling as accuracy is demonstrated, add lines of business.

07 Guardrails

  • Payout ceilings on automated decisions
  • Denials are always made by a licensed adjuster
  • Testing for unfair discrimination, per state regulator guidance
  • Full decision record for every claim

08 Services involved

Start here

Exploring AI in insurance?

We can run a proof of concept on your own data in about four weeks, measured against the KPIs above.