Verified
Lemonade
Scaling- Insurance
- 1,339 employees (Q2 2026)
- Bots built in-house
- 96% of claims first reported to AI Jim
- $1.07M premium per employee
Lemonade runs its insurance business on bots it built itself: AI Maya and its APIs sell 98% of policies, AI Jim takes 96% of first claim reports without a person, and CX.AI handles over half of customer inquiries (Dec 2025). Lemonade says it is not tied to one model provider, and it credits AI for nearly tripling in force premium per employee to $1.07 million since Q2 2022 and for cutting the cost of handling a Pet claim to $14 in 2025.
lemonade.comTheir approach to AI
- The platform
- Lemonade built its whole stack, which it calls Lemonade OS: AI Maya sells, AI Jim handles claims, CX.AI answers service requests, and Blender, Forensic Graph and the internal bot Cooper run behind them on one customer data layer, the Customer Cortex (2025 Form 10-K).
- The models
- Lemonade trains its own pricing and lifetime-value models, and Shai Wininger said in Jul 2026 that Lemonade is model agnostic, benchmarks each new frontier model per use case and runs several at once. OpenAI's Realtime API powered AI Maya's voice agent in Apr 2025; Lemonade has named no other model vendor.
- The reach
- Lemonade's bots handle most customer contact: AI Maya and APIs sell 98% of policies, AI Jim takes 96% of first notices of loss, about 55% of claims are fully automated, and CX.AI handles over half of inquiries without a person (Dec 2025).
- The catch
- Lemonade credits the fall in its claims handling cost ratio to AI and automation together, with no AI share, has not said how many employees use AI each week, and has not disclosed which large language models run its bots or what AI itself costs.
Snapshot
- Employees
- 1,339 (Q2 2026)
- 1,282 at Dec 31, 2025: 810 in the United States, the rest mainly in Israel (about 310) and the Netherlands
- Technology budget
- $93.9M (2025)
- Technology development expense; $30.0M in Q2 2026, up 34% year on year, partly on software for AI
- AI/ML headcount
- Not disclosed
- Some product development staff sit in Israel with the co-founders (2025 Form 10-K)
- Who owns AI
- The co-founders
- CEO Daniel Schreiber and President Shai Wininger, a former CTO; no chief AI officer disclosed
- Platform
- Built in-house
- AI Maya, AI Jim, CX.AI, Blender, Forensic Graph and Cooper on the Customer Cortex
- Models
- Own models plus LLMs, model agnostic
- Proprietary LTV pricing models; open-source and proprietary LLMs (Nov 2023); not tied to one provider (Jul 2026); OpenAI Realtime API for voice (Apr 2025)
- Reach
- 98% of policies sold by bots
- AI Jim takes 96% of first claim reports; CX.AI handles over half of inquiries (Dec 2025)
- Use cases
- Sales, service, claims, pricing, fraud, operations
- Over 100 business processes identified for generative AI in May 2023
Key findings
Lemonade's bots handle most customer contact: AI Maya and its APIs sell 98% of policies, AI Jim takes 96% of first notices of loss with no human involved, about 55% of claims were fully automated, and CX.AI handles over half of customer inquiries without a person, according to the 2025 Form 10-K (Dec 31, 2025).
Lemonade put in force premium per employee at $1.07 million in Q2 2026, up from $0.40 million in Q2 2022, under the heading that AI powers its efficiency gains; Lemonade said in Apr 2026 that it scaled its operation by embedding large language models in its stack, more than doubling in force premium from Q4 2022 while headcount fell 6% to 1,291, and it then hired in product and engineering to reach 1,339 employees in Q2 2026.
Lemonade credits AI for cutting the cost of handling a Pet claim from $44 in 2021 to $14 in 2025 (Feb 2026). Lemonade's loss adjustment expense ratio, the cost of handling claims as a share of premium, fell to 5% in Q2 2026 from 13% in Q3 2022, which Lemonade credits to AI and automation together without giving AI's share (Jul 2026).
Lemonade's President Shai Wininger said in Jul 2026 that Lemonade is model agnostic, benchmarks each new frontier model and can move a better one from evaluation to implementation within hours; the only outside model provider on record is OpenAI, whose Realtime API powered AI Maya's voice agent (Apr 2025), and Lemonade's website and app run on AWS, its sole third-party data center provider.
Lemonade has not said how many of its employees use AI each week, and it remains unprofitable, with a 2025 net loss of $165.5 million and its first positive adjusted EBITDA quarter expected in Q4 2026; CFO Tim Bixby hands over to Nick Stead on Jan 1, 2027, and an investor day on Nov 17, 2026 is set to cover Lemonade's AI capabilities.
Where Lemonade sits
- Stage
- Scaling
Lemonade has reported a firm-level figure it credits to AI and one use-case return it credits to AI, in Pet claims: its Q2 2026 investor presentation, under the heading that AI powers its efficiency gains, puts in force premium per employee at $1.07 million, about three times the $0.40 million of Q2 2022, which its Apr 2026 letter credits to embedding large language models in its stack, and its Feb 2026 letter credits AI for cutting the cost of handling a Pet claim from $44 in 2021 to $14 in 2025. Lemonade runs AI on one in-house platform, Lemonade OS, whose bots sell 98% of policies and take 96% of first claim reports, and its co-founders Daniel Schreiber and Shai Wininger report the results on earnings calls, which places Lemonade at Scaling. Lemonade falls short of Embedded because it credits its firm-level cost measures, the loss adjustment expense ratio and flat operating expenses, to AI and automation together with no AI share, and it has published no count of employees using AI each week; COO Adina Eckstein said only, in Jun 2026, that most new code is written with AI assistance.
How the stages are defined - Ownership
- Co-founders: CEO Daniel Schreiber and President Shai Wininger, a former CTO; no chief AI officer disclosed
- Platform
- Built: Lemonade OS: AI Maya, AI Jim, CX.AI, Blender and Cooper, built in-house
- Models
- Model agnostic: own pricing and LTV models plus frontier LLMs; OpenAI for Maya's voice (2025)
- Hosting
- AWS, Lemonade's sole third-party data center provider, runs its site and app; OpenAI's API served Maya's voice (Apr 2025); other model hosting undisclosed
- Reach
- 98% of policies sold by AI Maya and APIs (Dec 2025)
How they deploy
Built on bots from the start
Lemonade wrote in May 2023 that it bet on chat interfaces, AI and bots at its 2015 founding: its pricing comes from machine learning, its knowledge base sits in vector databases, its stack is API-based, and it sells almost all policies and takes in almost all claims through chat and AI. Lemonade called generative AI an accelerant along that course rather than a change of direction.
One proprietary stack
Lemonade runs a single in-house platform: customer-facing bots (AI Maya, AI Jim and CX.AI) and three internal applications (Blender, Forensic Graph and Cooper) all draw on the Customer Cortex, the layer where customer data is collected and analyzed, according to the 2025 Form 10-K. Lemonade said in Nov 2025 that this unified stack lets it plug in new AI technologies, often within hours of finishing safety and reliability testing.
Grow the business, scale the operation
Lemonade's stated plan for 2026 is to deploy its LTV models across channels, products and countries to drive gross profit growth, while using AI to grow revenue faster than costs and reach positive adjusted EBITDA (Feb 2026). Lemonade named three AI-enabled initiatives for 2026: a cross-sale platform, a more automated pricing engine, and a growth engine moving toward real-time autonomous optimization of spend, channels and creative. Daniel Schreiber wrote in Jul 2025 that Lemonade's AI-first approach should let it grow the business tenfold while barely growing its expense base.
AI-native against incumbents
Daniel Schreiber argued in a Mar 2026 post that Lemonade is an AI company that entered insurance, not an insurer that adopted AI, and that this shaped its data, expenses and talent. Lemonade backs the claim with its loss adjustment expense ratio, which it says incumbents publish and so offers one of the few like-for-like efficiency comparisons in insurance.
Pricing for AI drivers
Lemonade is positioning Lemonade Car for a future in which pricing reflects whether a human or AI is driving. Since Jan 2026 Lemonade has priced miles driven on Tesla's Full Self-Driving (Supervised) at about 50% of the human-driven rate, using vehicle data from Tesla's Fleet API shared with the customer's permission, and in Oct 2026 Lemonade added a 30% discount for older Tesla hardware running FSD v14 Lite.
- Deployment model
- Lemonade builds and runs its AI on its own platform rather than through a bought platform, and has disclosed no separate AI organization. Lemonade's customer-facing bots hand work to staff through Blender, its in-house insurance management system: AI Jim assigns the claims it cannot settle to human claims specialists by specialty, workload and schedule, and Blender shows them next steps, coverage determinations and fraud alerts. Cooper, an internal bot, handles repetitive back-office work such as processing paper checks, runs tens of thousands of tests on each software release, and handles most engineering task allocation, code deployment and quality assurance (2025 Form 10-K).
- Model providers
- Lemonade has disclosed few model vendors. Lemonade said in Nov 2023 that it had adopted large language models, both open source and proprietary, trained to answer customer emails, review pet medical records, evaluate satellite and other imagery and read home condition reports. Shai Wininger said on the Jul 29, 2026 earnings call that Lemonade is model agnostic and not tied to any one frontier model provider: its teams benchmark each new model against the others for capability and cost in each use case, can move a better one from evaluation to implementation within hours, and run several models at once. OpenAI's developer account said in Apr 2025 that Lemonade uses OpenAI's Realtime API to power AI Maya's voice agent, offering 24/7 multilingual phone support. Lemonade's own pricing models are proprietary: its LTV6 model, presented at the Nov 2022 investor day, was described as a roughly million-parameter AI that predicts churn, claims and cross-sales for each prospect, followed by LTV7 by May 2023, when Lemonade said LTV8 was due later that quarter.
- Data and infrastructure
- Lemonade offers its products through its website and app on Amazon Web Services data centers, which its 2025 Form 10-K calls its sole provider of third-party data centers; Lemonade has worked with AWS under a customer agreement filed with its 2020 IPO registration. The 10-K does not say where Lemonade's models run beyond that, and Lemonade has not disclosed whether any model runs outside AWS apart from the OpenAI API behind AI Maya's voice. Daniel Schreiber wrote in Jul 2025 that Lemonade has amassed over 10 billion data points from product tests, ads and customer journeys to feed its models, plus over a trillion exposure data points and millions of processed claims, and that it updates its models in real time. Lemonade Car uses telematics data and Tesla vehicle data from Tesla's Fleet API, shared with customer permission, and California's Proposition 103 limits Lemonade to miles-driven telematics in its largest car market.
Platforms and internal tools
AI Maya
AI Maya is Lemonade's onboarding and customer experience bot: it uses natural language to collect information, personalize coverage, produce quotes and take payment, asking a small number of high-impact questions. Lemonade says AI Maya and its APIs sell 98% of its policies (2025 Form 10-K), and AI Maya also answers the phone as a voice agent built on OpenAI's Realtime API (Apr 2025).
- Users
- 98% of policies sold through AI Maya and APIs (Dec 2025)
- Models
- OpenAI Realtime API for voice; other models not disclosed
AI Jim2016
AI Jim is Lemonade's claims bot. AI Jim took the first notice of loss with no human involved 96% of the time as of Dec 31, 2025, roughly 55% of claims were automated from start to finish, and AI Jim triages and assigns the rest to human claims experts. Lemonade says AI Jim paid a theft claim in 3 seconds about three months after Lemonade launched in fall 2016.
- Users
- 96% of first notices of loss (Dec 2025)
CX.AI
CX.AI is Lemonade's customer service bot platform, which uses natural language processing to understand and resolve requests such as coverage questions, adding a spouse, changing coverage amounts or payment methods, and adding new items. Lemonade says CX.AI handles over half of its customer inquiries without human intervention (2025 Form 10-K).
- Users
- Over half of customer inquiries (Dec 2025)
Blender
Blender is Lemonade's in-house insurance management platform for its customer experience, underwriting, claims, growth, marketing, finance and risk teams. Claims specialists see the claims AI Jim assigns them, with suggested next steps, coverage determinations and alerts of suspicious activity, and can dispatch and pay vendors from Blender.
Forensic Graph
Forensic Graph is Lemonade's fraud system, combining behavioral economics, large data sets and machine learning to predict, deter, detect and block fraud across the customer relationship by finding links between signals that humans would miss (2025 Form 10-K).
Cooper
Cooper is Lemonade's internal bot. Cooper processes paper checks for the customer experience team, runs tens of thousands of tests on each software release, analyzes NASA satellite imaging to spot wildfires and block ads and sales in affected areas, collates materials for regulatory filings, and handles most engineering task allocation, code deployment and quality assurance (2025 Form 10-K). A Rasa customer case study from 2019 quotes Lemonade's data science lead saying Lemonade built Cooper with Rasa's machine learning framework; Lemonade has not said whether Cooper still runs on it.
- Models
- Rasa's machine learning framework (2019); current models not disclosed
Customer Cortex
The Customer Cortex is the data layer under all six of Lemonade's applications, where data about customers is transmitted, analyzed and then used by AI Maya, AI Jim, CX.AI, Blender, Forensic Graph and Cooper (2025 Form 10-K).
Lemonade OS
Lemonade OS is Lemonade's name for the in-house technology stack it says it has built over the past decade for autonomy, from quotes to coverage changes and claims. Shai Wininger said on the Jul 29, 2026 earnings call that the growing use of Lemonade OS, which drives AI across the claims operation, continues a multiyear fall in Lemonade's claims handling costs.
Lemonade AgenticBeta waitlist by Oct 2026
Lemonade Agentic is a planned service on Lemonade OS for customers' personal AI agents, which would get quotes, change coverage and file claims directly in Lemonade's systems instead of through a website or form. Lemonade opened a beta waitlist and had not made the service generally available as of Oct 27, 2026; Lemonade names no AI vendor or protocol behind it.
LTV modelsLTV6 by Nov 2022
Lemonade's LTV models predict churn, claims and cross-sales for each prospect and guide targeting, pricing and marketing spend; new car business is written only if lifetime loss ratios estimated by the LTV model at the point of sale meet targets (May 2025). Lemonade presented LTV6, a roughly million-parameter model, at its Nov 2022 investor day.
Vendors and partners
Amazon Web ServicesBy 2020
Sole third-party data center provider for Lemonade's website and app · Lemonade filed its AWS customer agreement with its 2020 IPO registration and names AWS outages as a risk in its 2025 Form 10-K.
OpenAIBy Apr 2025
Realtime API behind AI Maya's voice agent · OpenAI's developer account said in Apr 2025 that the Realtime API's voice detection and low latency let Lemonade offer 24/7 multilingual phone support.
TeslaDec 2025
Fleet API vehicle data for Autonomous Car pricing · Tesla owners can connect their cars to the Lemonade app with permission, and Lemonade reads the data through Tesla's Fleet API to tell Full Self-Driving miles from manual ones (Jan 2026); Lemonade has not disclosed a formal partnership with Tesla.
RasaBy 2019
Machine learning framework behind the internal bot Cooper · Rasa's customer case study, dated by its reference to Apr 2019, says Lemonade built Cooper with Rasa; Lemonade has not confirmed current use.
Leadership
AI, data and platform leaders
Executives
Board
AI org map
Founder-led, with AI built into one in-house platform rather than run by a separate AI unit
Lemonade has no chief AI officer or separate AI organization on record. Lemonade's co-founders, Chairman and CEO Daniel Schreiber and President Shai Wininger, a former CTO, speak for its AI strategy, and the bots and internal tools share one stack built by its product and engineering teams. Lemonade's management team also includes COO Adina Eckstein, Chief Insurance Officer John Peters, Chief Business Officer Maya Prosor, Chief Product Officer Michal Langer and CFO Tim Bixby, who hands over to Nick Stead on Jan 1, 2027.
Lemonade does not disclose how many of its 1,282 employees (Dec 31, 2025) work on AI, data science or engineering. Lemonade's 2025 Form 10-K puts about 310 full-time employees in Israel, including the co-founders and some product development, help desk and online sales staff, with 810 in the United States and the rest mainly in the Netherlands. Lemonade's technology development expense, which covers product, engineering, design and quality assurance staff, was $93.9 million in 2025. Lemonade's headcount rose 5% year on year to 1,339 at Jun 30, 2026, which Nick Stead put down to net hiring in product and engineering, while its Q2 2026 investor presentation described total headcount as materially unchanged since Q2 2023.
Reporting lines
Chairman and CEO
Daniel Schreiber
Insurance
John Peters, Chief Insurance Officer
Reporting line not published
Lemonade does not publish its reporting lines; placing each executive under the CEO follows their titles in the 2025 Form 10-K and on Lemonade's governance page, not a disclosed structure. Lemonade's information security program is run by a Chief Information Security Officer, not named in the 2025 Form 10-K, who reports to an internal security committee and briefs the board's audit committee.
Where the AI workforce sits
Employees in the United States
810 of 1,282 employees at Dec 31, 2025. Lemonade does not break out an AI, data science or engineering headcount, so these are shares of the whole workforce.
~63%Employees in Israel, including product development staff
About 310 Lemonade employees at Dec 31, 2025, including the co-founders and some product development, help desk and online sales staff (2025 Form 10-K).
~24%Employees elsewhere, mainly the Netherlands
Estimate by subtraction from the 2025 Form 10-K figures.
~13%
Named people
Use cases
Lemonade runs AI in production across sales, customer service, claims, fraud, pricing and back-office work, through AI Maya, CX.AI, AI Jim, Forensic Graph, its LTV models and the internal bot Cooper. Lemonade reports a business return credited to AI for one of them, Pet claims, through the cost of handling each claim. At the firm level, Lemonade credits AI with lifting in force premium per employee to $1.07 million in Q2 2026 from $0.40 million in Q2 2022. Lemonade credits the fall in its loss adjustment expense ratio, the cost of handling claims as a share of premium, to AI and automation together, without saying how much is AI's.
| Use case | Area | Status | ROI |
|---|---|---|---|
| AI Jim claims intake and automation AI Jim takes the first notice of loss, pays or declines simple claims within seconds, and triages and assigns the rest to human claims experts in Blender; roughly 55% of claims were fully automated at Dec 31, 2025. Lemonade's Q2 2026 shareholder letter says advances in AI and automation across its claims operation, including higher instant claim rates in Pet and Renters, let it absorb substantial growth with minimal extra claims handling expense, as its loss adjustment expense ratio fell from 13% in Q3 2022 to 5% in Q2 2026. | Claims | In production | |
| Pet claims and AI medical records review Lemonade uses AI to review and validate pet medical records so that Pet claims can be decided almost instantly, and Pet's frequent claims give its models fast feedback. | Pet | In production | Cost of handling a Pet claim down from $44 in 2021 to $14 in 2025, credited to AI (Feb 2026)Source |
| AI growth engine and LTV model Lemonade's proprietary LTV model and AI-driven growth engine, which Daniel Schreiber described in Jul 2026 as some 50 machine learning algorithms working together, set targeting and allocate marketing spend in real time across channels, products and countries. Lemonade says growth spend rose about 200% from Q1 2023 to Q1 2026 while its modelled LTV/CAC ratio, an expectation rather than a realized return, held at or above 3x. | Growth | In production | |
| AI Maya onboarding and quoting AI Maya guides prospects through a short chat that collects information, personalizes coverage, quotes and takes payment, and together with Lemonade's APIs sells 98% of its policies. | Sales | In production | |
| CX.AI service bot CX.AI resolves customer requests such as coverage questions and policy changes without a person, and handles over half of Lemonade's customer inquiries. | Customer service | In production | |
| AI Maya voice agent AI Maya answers phone calls as a voice agent built on OpenAI's Realtime API, giving Lemonade customers 24/7 multilingual phone support. | Customer service | In production | |
| Machine learning underwriting and pricing Lemonade applies machine learning to data from each customer interaction to classify and price risk (2025 Form 10-K), and said in Feb 2026 that Europe's regulatory environment allows rapid iteration in pricing models and that telematics-driven loss prediction is improving its Car results. | Underwriting and pricing | In production | |
| Autonomous Car pricing for Tesla Full Self-Driving Lemonade reads a customer's Tesla data through Tesla's Fleet API with permission, separates Full Self-Driving miles from manual ones in its usage-based risk models and charges about 50% less for FSD (Supervised) miles, or 30% less on older hardware running FSD v14 Lite (Oct 2026). | Car | In production | |
| Crash detection and car claims Most Lemonade Car claims start in the app, often through crash detection, which Lemonade says speeds the first notice of loss and improves triage (Jul 2026). | Car | In production | |
| Forensic Graph fraud detection Forensic Graph uses machine learning to find links between signals across the customer relationship to predict, detect and block fraud; Lemonade says it has helped avoid millions of dollars of potential losses without giving a figure or date. | Fraud | In production | |
| Wildfire detection and catastrophe claims Cooper analyzes NASA satellite imaging to spot wildfires and block ads and sales in affected areas, and during the Jan 2025 California wildfires Lemonade used AI and aerial imagery to settle hundreds of claims within an hour of submission. | Catastrophe | In production | |
| Cooper internal automation Cooper processes paper checks, runs tens of thousands of tests on each software release, collates regulatory filing materials, and handles most engineering task allocation, code deployment and quality assurance. | Operations and engineering | In production | |
| Large language models for emails and property reports Lemonade said in Nov 2023 that it had large language models trained to answer customer emails, evaluate satellite and other imagery and read home condition reports, while noting it had yet to deploy most of what it had built. | Operations | Pilot |
Adoption
Lemonade's AI reach is measured in customer contact rather than employee usage: AI Maya and its APIs sell 98% of policies, AI Jim takes 96% of first notices of loss with no human involved, about 55% of claims were fully automated and CX.AI handles over half of customer inquiries without a person (Dec 2025), up from close to half of claims settled by AI Jim in May 2023. COO Adina Eckstein said in Jun 2026 that more than 60% of Lemonade's claims are handled by AI, a different measure from the 55% fully automated at Dec 31, 2025, and that some customers already file claims through AI agents they run themselves. Inside Lemonade, claims, underwriting, customer experience, growth, marketing, finance and risk teams work in Blender, the Cooper bot handles most engineering task allocation, code deployment and quality assurance, and Eckstein said in Jun 2026 that most new code is written with AI assistance and that AI also runs in HR, finance and customer support. Lemonade has not published how many employees use generative AI or which coding assistants they use.
By the numbers
Rollout, first pilot to today
2016
Lemonade launched in fall 2016, and about three months later AI Jim paid a theft claim in 3 seconds, which Lemonade called a world record.
Nov 2022
Lemonade presented LTV6 at its investor day, a roughly million-parameter AI model that predicts churn, claims and cross-sales for each prospect.
May 2023
Lemonade said it had identified over 100 business processes that generative AI could automate, had prototyped dozens, and would put its first generative AI automations into production within weeks.
Nov 2023
Lemonade said it uses open-source and proprietary large language models trained to answer customer emails, review pet medical records, evaluate imagery and read home condition reports, and would end 2023 with roughly the same number of full-time employees as it started with.
Apr 2024
Lemonade reported a loss adjustment expense ratio of 7.6% for Q1 2024, nearly halved in two years, and headcount down 11% year on year as its book grew 22%.
Feb 2025
Lemonade described AI-powered review of pet medical records for near-instant claims decisions and its use of AI and aerial imagery to settle claims from the Jan 2025 California wildfires, after a Nov 2024 investor day with AI case studies.
Apr 2025
OpenAI said Lemonade uses its Realtime API to power AI Maya as a voice agent for 24/7 multilingual phone support.
Nov 2025
Lemonade said its loss adjustment expense ratio had fallen from 13% to 7% in three years, largely through AI-driven automation, and that its claims department shrank as claims volume rose more than 2.5 times.
Jan 2026
Lemonade launched Autonomous Car, using Tesla's Fleet API to price miles driven on Tesla's Full Self-Driving (Supervised) at about 50% of the human-driven rate.
Feb 2026
Lemonade's 2025 Form 10-K reported that AI Jim takes 96% of first notices of loss, about 55% of claims are fully automated, CX.AI handles over half of inquiries and AI Maya with APIs sells 98% of policies.
Jul 2026
Lemonade reported a record-low loss adjustment expense ratio of 5% for Q2 2026, ahead of its goal of halving the ratio again alongside the next doubling of in force premium.
Oct 2026
By Oct 27, 2026, Lemonade had opened a beta waitlist for Lemonade Agentic, a planned service on Lemonade OS through which customers' personal AI agents would buy and manage policies and file claims.
Economics & controls
Spend and ROI
Technology budget
Lemonade spent $93.9 million on technology development in 2025, up 9% from $85.8 million in 2024 and $88.8 million in 2023, mostly on product, engineering, design and quality assurance staff. Lemonade's technology development expense rose 34% to $30.0 million in Q2 2026 as employee-related costs grew on higher headcount and software expense rose 125%, or $1.5 million (Q2 2026 Form 10-Q); Nick Stead said on the Jul 29, 2026 call that the higher software costs support Lemonade's expanding AI capabilities. Lemonade does not report AI spending separately.
What AI itself costs
Lemonade has not disclosed what it pays for models, inference or AI tools. Lemonade's 2025 Form 10-K says it is making significant investments in AI and expects to need more to keep improving its AI technologies, and that those investments may not always improve efficiency or profitability.
Reported ROI
Lemonade credits AI with lifting output per employee. Lemonade's Q2 2026 investor presentation put in force premium per employee at $1.07 million, about three times the $0.40 million of Q2 2022, under the heading that AI powers its efficiency gains, and its Apr 2026 letter said Lemonade scaled its operation by embedding large language models in its stack, as in force premium grew from $625 million to $1.33 billion between Q4 2022 and Q1 2026 while employees fell from 1,367 to 1,291, before rising to 1,339 by Jun 30, 2026.
Claims and operating costs
Lemonade credits its firm-level cost measures to AI and automation together, without saying how much is AI's. Lemonade's loss adjustment expense ratio, the cost of handling claims as a share of gross earned premium, fell from 13% in Q3 2022 to 7% in Q3 2025, which its Nov 2025 letter put largely down to harnessing AI to drive automation, and to 5% in Q2 2026, which its Jul 2026 letter credits to advances in AI and automation that let Lemonade absorb growth with minimal extra claims handling expense; Lemonade compares it with about 9% at incumbents. Lemonade's operating expenses excluding growth spend rose only from $87 million in Q3 2022 to $95 million in Q3 2025 and $101 million in Q4 2025 as in force premium roughly doubled, on charts that credit AI and automation together.
Headcount
Lemonade had 1,282 employees at Dec 31, 2025 and 1,291 at the end of Q1 2026, against 1,367 at the end of 2022, then 1,339 at Jun 30, 2026 after net hiring in product and engineering, while customers grew to about 3.3 million in Q2 2026. Lemonade's Q4 2025 letter said it added 1.2 million customers since Q3 2022 while headcount fell 6%, and Lemonade's 2025 Form 10-K puts its customers per employee at about 2,300, against its estimate of 300 to 600 at five U.S. competitors whose counts include agents and brokers.
Path to profit
Lemonade lost $165.5 million in 2025 and $43.4 million in Q2 2026, with an adjusted EBITDA loss of $18.7 million in the quarter, and guides to its first positive adjusted EBITDA quarter in Q4 2026, about $8 million implied. Lemonade's 2026 plan pairs LTV-driven growth with using AI to grow revenue faster than costs; in Q2 2026 its total operating expense excluding losses and loss adjustment expense rose 41% to $182.2 million, driven by growth spend rising to $64.4 million from $49.7 million, $6.5 million of executive stock compensation and a one-time tax benefit in the prior-year quarter. Lemonade reports results for Q3 2026 on Nov 5, 2026.
Synthetic agents are financing, not AI
Lemonade's synthetic agents program is a financing arrangement, not an AI system: General Catalyst has funded part of Lemonade's customer acquisition spend since Jun 2023 in return for a share of the premiums, and from Jan 2027 Hannover Re takes over with up to $250 million of capital under a Jun 22, 2026 agreement, at a cost of capital Lemonade put at about 9.8% in its Jul 2026 letter.
Governance and risk
Board oversight
Lemonade's board oversees strategic risk, and its audit committee oversees risk assessment, including information security and cybersecurity, and receives periodic reports from Lemonade's Chief Information Security Officer (2026 proxy statement and 2025 Form 10-K). Lemonade has disclosed no board committee dedicated to AI; Prashant Ratanchandani, who leads engineering for generative AI products at Meta, joined the board on Oct 15, 2025 and sits on its audit committee.
Testing before deployment
Lemonade said in May 2023 that it would prioritize compliance and fairness in its use of AI and be careful where it put generative AI into production, and in Nov 2023 that it was taking pains to make its models fair, reliable and compliant before deploying them. Lemonade said in Nov 2025 that it integrates new AI technologies once safety and reliability testing is complete, often within hours.
People in the claims loop
Lemonade's AI Jim settles only the claims it is authorized to settle; it routes the rest, and any claim where it identifies concerns, to human claims experts, and many incidents are reviewed by a person before approval (2025 Form 10-K). Lemonade names inaccurate claim payments and mispriced policies from its proprietary AI among its main business risks.
Insurance AI regulation
Lemonade's 2025 Form 10-K lists the NAIC's AI principles and model bulletin, state guidance on unfair discrimination from big data and AI in California, Connecticut and New York, Colorado's Artificial Intelligence Act and California privacy law as constraints on its models. California's Proposition 103 limits Lemonade to miles-driven telematics in its largest car market, and Lemonade does not use behavioral telematics in three other states.
Europe and automated decisions
Lemonade's 2025 Form 10-K says the EU AI Act, with most requirements applying from Aug 2, 2026, requires AI literacy and transparency and treats AI in health and life insurance as high risk, which Lemonade does not offer in the EU. Lemonade also notes that the GDPR limits automated decisions with legal effects unless they are necessary for a contract or have explicit consent, and that the UK's 2025 Data Use and Access Act relaxed that limit for data outside special categories.
Autonomous driving risk
Lemonade's 2025 Form 10-K warns that its pricing for self-driving miles depends on data continuing to show improving safety for those systems, and that it may have to raise rates if the data shows more risk. Lemonade connects to Tesla vehicles only with the customer's permission.
Recent signals
Oct 27, 2026
Beinsure reported that Lemonade had opened a beta waitlist for Lemonade Agentic, a planned service on Lemonade OS that would let customers' personal AI agents get quotes, change coverage and file claims without a website or form, and that it was not yet generally available.
Oct 15, 2026
Lemonade filed the terms of its CFO transition: Tim Bixby stays CFO through Jan 1, 2027 under an Oct 12 transition agreement with no severance, then joins the board, and Nick Stead takes over as CFO.
Oct 13, 2026
Shai Wininger said Lemonade's Autonomous Car for older Tesla hardware is live in Oregon and Missouri, with 30% off miles driven with Full Self-Driving engaged.
Oct 8, 2026
Lemonade's Autonomous Car began giving Teslas with older HW3 hardware 30% off miles driven on FSD (Supervised) v14 Lite in Arizona, Colorado and Tennessee, alongside 50% off for HW4 cars on FSD v14.2 or newer.
Sep 2, 2026
Lemonade launched Lemonade Car and Autonomous Car in Missouri on the same day, its first state to get both at once, with 50% off miles driven on FSD (Supervised); Shai Wininger said Lemonade's pricing models see the lower risk in the data.
Aug 3, 2026
Lemonade brought Autonomous Car to Tennessee, with Shai Wininger saying its pricing models see the lower risk of FSD (Supervised) miles in the data.
Jul 29, 2026
On Lemonade's Q2 2026 call, Shai Wininger said Lemonade is model agnostic and runs several models at once, Daniel Schreiber said some 50 machine learning algorithms allocate its marketing spend, and Nick Stead said headcount rose 5% to 1,339 on product and engineering hiring.
Jul 29, 2026
Lemonade reported a record-low 5% loss adjustment expense ratio for Q2 2026, credited it to advances in AI and automation in claims, and said its Nov 17, 2026 investor day will cover its AI capabilities.
Jul 27, 2026
Lemonade's board approved moving CFO Tim Bixby to the board and appointing SVP Finance Nick Stead, 33, as CFO, both effective Jan 1, 2027.
Jun 15, 2026
COO Adina Eckstein said 98% of Lemonade's policies are sold through AI, more than 60% of its claims are handled by AI and most of its new code is written with AI assistance, and that some customers file claims through their own AI agents.
Apr 29, 2026
Lemonade said it ended Q1 2026 with just over $1 million of in force premium per employee after more than doubling premium and cutting team size 6% since Q4 2022 by embedding large language models in its stack.
Mar 2, 2026
Daniel Schreiber argued that incumbents cannot catch up by adding AI, saying Lemonade's exposure grew about 65% from 2022 to 2025 while its headcount declined, and put its loss adjustment expense ratio at 6%.
Feb 19, 2026
Lemonade said it will use AI in 2026 to grow revenue faster than expenses and reach positive adjusted EBITDA, and credited AI across claims and operations for cutting the cost of handling each Pet claim to $14 in 2025.
Jan 21, 2026
Lemonade launched Autonomous Car, feeding Tesla Fleet API data into its usage-based risk models to price Full Self-Driving (Supervised) miles at about half its human-driven rate.
Nov 5, 2025
Lemonade said its loss adjustment expense ratio had nearly halved in three years, largely because it harnessed AI to drive automation, and that it expects to halve the ratio again alongside the next doubling of the business.
Quick answers
- How does Lemonade deploy AI?
- Lemonade runs its insurance business on bots it built itself: AI Maya and its APIs sell 98% of policies, AI Jim takes 96% of first claim reports without a person, and CX.AI handles over half of customer inquiries (Dec 2025). Lemonade says it is not tied to one model provider, and it credits AI for nearly tripling in force premium per employee to $1.07 million since Q2 2022 and for cutting the cost of handling a Pet claim to $14 in 2025.
- What does Lemonade use AI for?
- Lemonade runs AI in production across sales, customer service, claims, fraud, pricing and back-office work, through AI Maya, CX.AI, AI Jim, Forensic Graph, its LTV models and the internal bot Cooper. Lemonade reports a business return credited to AI for one of them, Pet claims, through the cost of handling each claim. At the firm level, Lemonade credits AI with lifting in force premium per employee to $1.07 million in Q2 2026 from $0.40 million in Q2 2022. Lemonade credits the fall in its loss adjustment expense ratio, the cost of handling claims as a share of premium, to AI and automation together, without saying how much is AI's.
- Who leads AI at Lemonade?
- AI at Lemonade is led by Shai Wininger (Co-founder and President); Michal Langer (Chief Product Officer). Ownership of AI on the org chart: Co-founders: CEO Daniel Schreiber and President Shai Wininger, a former CTO; no chief AI officer disclosed.
- What AI platform does Lemonade use?
- Lemonade built its whole stack, which it calls Lemonade OS: AI Maya sells, AI Jim handles claims, CX.AI answers service requests, and Blender, Forensic Graph and the internal bot Cooper run behind them on one customer data layer, the Customer Cortex (2025 Form 10-K).
- Which AI models does Lemonade use?
- Lemonade trains its own pricing and lifetime-value models, and Shai Wininger said in Jul 2026 that Lemonade is model agnostic, benchmarks each new frontier model per use case and runs several at once. OpenAI's Realtime API powered AI Maya's voice agent in Apr 2025; Lemonade has named no other model vendor.
- How many Lemonade employees use AI?
- Lemonade's bots handle most customer contact: AI Maya and APIs sell 98% of policies, AI Jim takes 96% of first notices of loss, about 55% of claims are fully automated, and CX.AI handles over half of inquiries without a person (Dec 2025).
- How far has AI spread at Lemonade?
- Lemonade has reported a firm-level figure it credits to AI and one use-case return it credits to AI, in Pet claims: its Q2 2026 investor presentation, under the heading that AI powers its efficiency gains, puts in force premium per employee at $1.07 million, about three times the $0.40 million of Q2 2022, which its Apr 2026 letter credits to embedding large language models in its stack, and its Feb 2026 letter credits AI for cutting the cost of handling a Pet claim from $44 in 2021 to $14 in 2025. Lemonade runs AI on one in-house platform, Lemonade OS, whose bots sell 98% of policies and take 96% of first claim reports, and its co-founders Daniel Schreiber and Shai Wininger report the results on earnings calls, which places Lemonade at Scaling. Lemonade falls short of Embedded because it credits its firm-level cost measures, the loss adjustment expense ratio and flat operating expenses, to AI and automation together with no AI share, and it has published no count of employees using AI each week; COO Adina Eckstein said only, in Jun 2026, that most new code is written with AI assistance.
Open questions
- Which large language models, beyond OpenAI's Realtime API for AI Maya's voice, run Lemonade's bots and internal tools now that Lemonade calls itself model agnostic, and does Lemonade host any open-weight models on AWS?
- When will Lemonade Agentic leave its beta waitlist, and which protocols will personal AI agents use to reach Lemonade OS?
- What does Lemonade spend on AI models, inference and tools within its $93.9 million of 2025 technology development expense?
- How many Lemonade employees work on AI, data science and machine learning, and who leads that work below the co-founders?
- What is Michal Langer's remit as Chief Product Officer, and does the product team own AI Maya, AI Jim and CX.AI?
- How many Lemonade employees use generative AI each week, and which AI coding tools write most of Lemonade's new code, as Adina Eckstein said in Jun 2026?
- How much has Forensic Graph saved in fraud losses, given that Lemonade says only that it has avoided millions of dollars of potential losses?
- How much of the fall in Lemonade's loss adjustment expense ratio, from 13% in Q3 2022 to 5% in Q2 2026, comes from AI Jim and Lemonade's other AI rather than other automation, and how much of the near tripling in in force premium per employee comes from AI rather than growth spend and pricing?
- What business return do Lemonade's deployed tools with no published figures deliver, such as CX.AI, AI Maya's voice agent, Cooper and the LTV growth engine, whose LTV/CAC ratio is a modelled expectation rather than a realized return?
- How much of Lemonade's gross loss ratio improvement comes from its AI pricing models rather than rate increases and reinsurance changes?
- How will Lemonade meet the EU AI Act and Colorado's Artificial Intelligence Act for its pricing and claims models?
- What will Lemonade disclose about its AI capabilities at its Nov 17, 2026 investor day?
Sources
Lemonade
- Investor relations home
- Management and board of directors
- Q2 2026 investor presentation (Jul 2026)
- Daniel Schreiber: Why incumbents won't catch up (Mar 2, 2026)
- Daniel Schreiber: Lemonade turns 10 (Jul 1, 2025)
- Autonomous Car insurance for Tesla FSD
- Lemonade Agentic beta waitlist
- Press release: Lemonade unveils Autonomous Car insurance (Jan 21, 2026)
- Press release: Autonomous Car comes to Tennessee (Aug 3, 2026)
- Press release: Car and Autonomous Car come to Missouri (Sep 2, 2026), via Cision
- Press release: Autonomous Car adds Tesla FSD v14 Lite (Oct 8, 2026)
- Press release: Meta's VP of AI Products joins the board (Oct 15, 2025)
- Press release: Q3 2026 results date (Oct 19, 2026)
Earnings calls and news
Shareholder letters
- Q2 2026 shareholder letter (Jul 29, 2026)
- Q1 2026 shareholder letter (Apr 29, 2026)
- Q4 2025 shareholder letter (Feb 19, 2026)
- Q3 2025 shareholder letter (Nov 5, 2025)
- Q1 2025 shareholder letter (May 6, 2025)
- Q4 2024 shareholder letter (Feb 25, 2025)
- Q1 2024 shareholder letter (Apr 30, 2024)
- Q3 2023 shareholder letter (Nov 1, 2023)
- Q1 2023 shareholder letter (May 3, 2023)
SEC filings
Partners, and executives on X
- OpenAI Developers: Lemonade uses the Realtime API to power AI Maya (Apr 9, 2025)
- Shai Wininger: 30% off for HW3 Teslas with FSD v14 Lite (Oct 8, 2026)
- Shai Wininger: Autonomous Car live in Oregon and Missouri (Oct 13, 2026)
- Lemonade: Tesla owners can connect their cars to the app (Dec 11, 2025)
- Rasa customer case study: Lemonade built Cooper with Rasa (2019)
Data confidence
- Confirmed
- Stated in a primary or reputable public source.
- Estimated
- Derived or approximated from public reporting.
- Unconfirmed
- Reported, but not yet corroborated.
Lemonade profile · Enterprise AI diffusion report by Imagine · facts checked against their sources on Nov 1, 2026
