Lemonade
AI-driven underwriting and claims management
By the end of 2025, Lemonade's AI agents (Maya for underwriting, AI Jim for claims) handle about 96% of first notices of loss without a human and automate about 55% of cases, pulling the LAE ratio toward ~4% versus 13% three years earlier.
Key points
- AI-driven underwriting and claims management through Maya and AI Jim.
- In-house platform with customer cortex, risk machine learning, and pricing.
- About 96% of loss notices taken without a human, 55% automated.
- LAE ratio pulled toward about 4%, evidence level A confirmed.
Objective
Reduce the cost of claims management and refine risk pricing by automating underwriting and claims handling, in order to hold management ratios far below those of traditional insurers with a smaller headcount.
The deployment
Lemonade runs two AI agents. Maya handles conversational underwriting: it recommends coverage, generates a personalized quote, and collects payment in under 90 seconds, backed by the customer cortex, a machine learning system that estimates risk and price in real time. AI Jim handles claims: it takes the first notice of loss and resolves part of the cases end to end, without human intervention. As of the end of 2025, the company states that about 96% of first notices of loss are taken by AI without human intervention and that about 55% of claims are fully automated. This automation pulls the LAE ratio down: Lemonade cites an LAE ratio of about 4% for the AI-driven portion, and an overall LAE ratio that fell from 13% to 7% in three years, while claim volume grew by more than 2.5 times.
Results Proof A
Figures published in Lemonade's shareholder letters (filed as 8-K with the SEC) and echoed by the business press. Primary source of financial results, with several concordant quarters on the automation trajectory.
How it works
Documented architectureThe stack in detail
- outil Maya (agent IA de souscription) in-house conversational agent that recommends coverage, generates the quote, and collects payment in under 90 seconds
- outil AI Jim (agent IA de sinistres) takes about 96% of first notices of loss in chat and settles about 55% of cases end to end
- plateforme Customer cortex Lemonade's proprietary machine learning system that estimates risk and price in real time; exact models not published
- infra Telemetrie continue first-party more than 90% of customers equipped; feeds real-time pricing and model retraining
How it runs, concretely
For ops teams-
1Conversational underwriting AI
Maya converses with the prospect, estimates risk through the customer cortex, and generates a quote paid in under 90 seconds.
-
2Claim report AI
AI Jim takes the first notice of loss in chat and collects the information.
-
3Automatic resolution AI
For about 55% of cases, AI Jim verifies and settles the claim end to end, without a human.
-
4Human escalation claims team
Complex or suspicious cases go to a human handler.
Risk data and the outcome of past claims. If the quality of claim labels degrades, pricing and automation become less reliable.
How your customers perceive this type of use
Sourced studiesC'est la famille la moins acceptee : 68% des Americains jugent inacceptable un score financier personnel calcule par algorithme et 67% l'analyse video automatisee d'entretiens d'embauche (Pew Research, 2018). La demande d'explication et de recours est massive : 83% veulent savoir quelles donnees l'IA utilise et 91% veulent pouvoir corriger des donnees erronees (Consumer Reports, 2024). A l'echelle mondiale, seuls 46% se disent prets a faire confiance aux systemes d'IA et 70% jugent une regulation necessaire (KPMG / Universite de Melbourne, 2025).
Acceptance conditions
- Transparence sur les donnees utilisees : 83% des Americains la reclament (Consumer Reports 2024)
- Droit de correction des donnees erronees : 91% le demandent (Consumer Reports 2024)
- Explication de la logique de decision : 44% des consommateurs sont plus enclins a utiliser un agent IA si sa logique est clairement expliquee (Salesforce 2024)
- L'acceptabilite depend du contexte de la decision : 50% des Americains jugent equitable un score de risque criminel pour la liberation conditionnelle, contre 32% pour un score financier applique aux consommateurs (Pew Research 2018)
Red lines
- La decision opaque et sans recours sur l'emploi, le credit ou le logement : 45% tres mal a l'aise pour l'embauche, 39% pour le pret, 39% pour le logement (Consumer Reports 2024)
- Le scoring des personnes a partir de donnees comportementales : 68% le jugent inacceptable pour les offres financieres (Pew Research 2018)
Sources: Pew Research Center 2018 · Consumer Reports 2024 · KPMG / Universite de Melbourne 2025 · Salesforce 2024
How to replicate
Inference, not sourcedData prerequisites
- History of claims with outcome and cost
- First-party risk data (underwriting, telemetry)
- Logging of customer interactions
Org prerequisites
- Insurer or MGA license
- Actuarial function to validate the pricing models
- Compliance framework for automated decision-making (GDPR, AI Act)
Possible stack
- Custom platform (the Lemonade path)
- Conversational AI + rules engine + actuarial model
- insurtech solutions for automated claims management
The plan, step by step
- Step 1Frame the data: history of claims with outcome and cost, underwriting data, mapping of automatable casesDeliverable: Labeled dataset + list of claim types that are candidates for automation
- Step 2Automate the first notice of loss: a conversational agent collects the information, the human keeps the settlementDeliverable: Claim report chat in production, AI intake rate measured
- Step 3Build the risk scoring and automatic settlement rules on simple cases, with actuarial validationDeliverable: Model validated by the actuarial function + documented GDPR/AI Act compliance framework (oversight, contestation)
- Step 4Open end-to-end settlement on a limited scope (small claims, low fraud risk), with systematic human escalation on doubtful casesDeliverable: Share of cases settled without a human measured + working escalation path
- Step 5Extend claim type by claim type and track the economic effectDeliverable: LAE ratio / automation rate dashboard compared to the reference period
First step: Automate the first notice of loss first (information intake in chat) before aiming for end-to-end settlement.
The same pattern, proven in other industries
Sources
- S1 Lemonade, Inc. - Form 8-K - Shareholder Letter Q3 2025 Primary archive pending
- S2 LMND Shareholder Letter Q3 2025 (PDF) Primary archive pending
- S3 Lemonade and Porch Show AI Is Rewriting Insurance Math Established press archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.