What problem does it solve?
Buying insurance online often means long forms with questions customers do not understand ("what is your property's construction type?"), and each confusing question can make a customer leave before the price. A form can show a price, but it cannot answer a question about what is and is not covered. Intermediaries still dominate some lines: Lemonade's 2025 annual report notes that homeowners insurance in the United States is sold primarily via agents.
A conversation can ask fewer, better questions, explain terms in plain language and answer questions about cover at the moment of purchase. The hard part is doing that while staying inside distribution rules: demands and needs testing, product information disclosure, suitability for investment based products and honest explanations of exclusions.
How does it work?
- Understand the need. The agent asks what the customer wants to protect and captures the demands and needs the law requires before recommending a product.
- Collect rating data conversationally. It asks only the rating questions the product needs, prefills what it can from approved data sources and explains why each question matters.
- Price through the rating engine. The agent calls the insurer's own rating and underwriting rules; it never calculates or negotiates price itself.
- Explain and compare cover. Retrieval over the product documents answers questions on limits, excesses and exclusions, and the required product information is shown before purchase.
- Bind and pay. The customer confirms the details, accepts the documents and pays through a secure payment link; the policy is issued and documents are sent.
- Hand over when needed. Advice requests, referrals from underwriting rules, vulnerable customers and complex needs go to a licensed human with the conversation so far.
- Audience
- Customer facing
- Autonomy
- Supervised agent
- Adoption
- Early adopters
- Channels
- Web chat, Mobile app, WhatsApp, Phone and voice
What is it worth?
Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.
No public deployment has disclosed a measurable outcome yet.
Value drivers: Revenue growth, Customer experience, Lower cost to serve, Inclusion and access.
Indicative value
A direct personal lines insurer with 200,000 online quote journeys a year
USD 320,000 to USD 2.2 million
Additional gross written premium per year
How this is calculated
Formula: quoteJourneys * baseConversion * relativeUplift * averagePremium. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Online quote journeys started per year quoteJourneys, journeys per year | 200,000 | 200,000 | The reference insurer. |
| Baseline conversion from quote start to purchase baseConversion, fraction of journeys | 0.08 | 0.12 | Editorial assumption for a direct channel. Replace with your own funnel data. |
| Relative conversion uplift from the conversational journey relativeUplift, fraction of baseline conversion | 0.05 | 0.15 | Editorial assumption; no insurer on this page publishes a controlled uplift. Measure it with an A/B test. |
| Average annual premium per new policy averagePremium, USD per policy | 400 | 600 | Editorial assumption. Replace with your own average premium. |
What it leaves out: Premium, not profit. It leaves out loss ratio effects of the new business, acquisition costs saved or added, the cost of running the agent and regulatory work, and the risk that a poorly designed flow lowers conversion.
Market estimates (analyst estimates, not deployments)
- Evident reports that sales and distribution matched underwriting and pricing with nine new AI use cases among the 30 insurers it tracked in the second quarter of 2026, and notes that Aviva, Liberty Mutual Insurance and Allianz now generate live quotes directly inside ChatGPT. Evident: Insurance Use Case Trends Q2 2026 (2026)
Who already uses it?
1 public deployment, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
Lemonade
United States · Insurance · 2025
Lemonade sells renters, homeowners, pet, car and life insurance through a chat with its bot AI Maya, which collects information, personalizes coverage, creates the quote and takes payment by asking a limited number of high impact questions. Its 2025 annual report says AI Maya and its APIs sell 98% of its policies. A second bot platform, CX.AI, resolves pre and post purchase requests such as coverage questions, adding a spouse, changing coverage amounts or payment methods and adding newly bought items, and handles over half of customer inquiries without human intervention.
- Containment rate: at least 50%, customer inquiries handled by CX.AI without human intervention, as reported in the 10-K for 2025
"Currently, over half of Lemonade’s customer inquiries are handled this way."
Claimed by: organization
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- Rating and underwriting rules exposed through an API
- Product documents (terms, product information documents) per product version
- A mapping of every rating question to plain language explanations and allowed answers
- Approved prefill data sources and their use conditions
Systems to integrate
- Rating engine and underwriting rules
- Policy administration for issuance
- Payment service provider
- Document generation and delivery
- CRM and handover to licensed sales staff
Complexity: High
The conversation is the easy part. Binding real policies needs the rating engine, underwriting rules, document generation, payments and policy issuance behind APIs, plus distribution compliance (demands and needs, product information, record keeping) built into the flow.
- 1
Start with a simple product
Renters, travel, pet or simple home cover with few rating factors and low advice needs are the right first products. Leave life and investment based products for later.
- 2
Put compliance in the flow, not the prompt
Demands and needs capture, required disclosures and document acceptance should be deterministic steps that cannot be skipped, with a record of each.
- 3
Let the rating engine own the price
The agent passes answers to the rating API and presents the result; it must not estimate, discount or negotiate prices.
- 4
Test for misselling
Build test conversations where customers ask leading questions ("so I'm covered for flood?") and check that exclusions are explained correctly.
- 5
Run an A/B test
Compare the conversational journey with the existing form on conversion, cancellations in the cooling off period and complaints before switching traffic.
Guardrails
- Price only from the rating engine; no free text price statements
- Required disclosures and demands and needs steps enforced by the flow
- Coverage answers only from the current product documents, with refusal when unsure
- Clear AI disclosure and a route to a licensed human
- Payment by secure link or tokenized card, never card numbers in the chat transcript
KPIs to instrument
- Conversion from quote start to purchase versus the form journey, by product
- Cancellations within the cooling off period
- Complaints and misselling indicators per 1,000 sales
- Handover rate to licensed staff and reasons
- Satisfaction at purchase
Human in the loop
Licensed sales staff take advice requests, underwriting referrals and vulnerable customers. Compliance reviews a sample of completed sales every month, and product owners approve every change to questions, explanations and flows.
Common failure modes
- The agent talks about price
- A model rounds, estimates or promises a discount. Keep all price statements tied to the rating engine output.
- Exclusions glossed over
- The agent reassures instead of explaining, which surfaces later as a declined claim. Test leading questions and cite the wording.
- Unsuitable sales
- The agent recommends a product without capturing needs. Enforce the demands and needs step.
- Invisible fairness issues
- Conversational data (language, typing style) leaks into underwriting or pricing. Keep the conversation layer separate from rating inputs.
What are the risks and rules?
EU AI Act
Depends on design
The conversational layer carries the Article 50 transparency duty. If the system assesses risk or sets prices for life or health insurance of natural persons, that part is high risk under Annex III point 5(c); pricing for property and casualty products is not listed.
Guidance
- Insurance Distribution Directive (IDD) (European Insurance and Occupational Pensions Authority, Europe). Sets the rules for how insurance is distributed in the EU, including demands and needs, product information and conduct, which also apply to AI sales journeys.
- Article 50, transparency obligations for providers and deployers of certain AI systems (European Union, Europe). People must be informed that they are interacting with an AI system, unless that is obvious from the context.
- Annex III, high risk AI systems referred to in Article 6(2) (European Union, Europe). Point 5(c) makes AI for life and health insurance risk assessment and pricing of natural persons high risk.
Controls to put in place
- Record of demands and needs, disclosures shown and documents accepted for every sale
- Rating engine version logged with each quote
- Monthly compliance sample of AI completed sales
- Complaints and cancellation monitoring by journey
- PCI DSS scope kept outside the conversational platform through tokenization
Frequently asked questions
- Do insurers really sell policies through chatbots?
- Some do. Lemonade's 2025 annual report says its bot AI Maya and its APIs sell 98% of its policies, with Maya collecting information, personalizing coverage, quoting and taking payment in a chat; the filing does not split that share between Maya and the APIs. Evident also notes that Aviva, Liberty Mutual Insurance and Allianz now generate live quotes directly inside ChatGPT.
- Can the AI give advice?
- Only within the distribution rules that apply. The cautious design keeps the agent to information and non advised sales on simple products, captures demands and needs in the flow, and hands advice requests to licensed staff.
- Is a sales chatbot high risk under the EU AI Act?
- The chat itself needs AI disclosure under Article 50. Any component that assesses risk or sets prices for individual life or health insurance is high risk under Annex III; for home, motor or travel pricing the Act does not list it as high risk.
How to cite this page
Blits.ai AI Use Case Library, "Conversational AI for insurance quote and buy", last verified 27 September 2026, https://www.blits.ai/ai-use-cases/conversational-insurance-quote-and-buy. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 27 September 2026: First published