Insurance has one of the worst conversion problems in online services, and the cause is well understood: the quote form.
Twenty fields. No indication of how long it will take. No explanation of why any of it is needed. Requests for information the person has to go and find. And at the end, possibly, a number.
People start these forms and abandon them at a rate that would be alarming in any other industry. The information is necessary (you genuinely can't quote without it), but the format is hostile.
A conversation asks for the same things and gets finished
The insight is not that a chatbot asks fewer questions. It asks the same ones. It's that a conversation has a completely different psychological shape.
One question at a time. Visible progress. The ability to explain why something is being asked. The ability to accept an approximate answer and clarify. And no wall of fields creating an instant judgement about how long this is going to take.
Bot: What kind of cover are you looking for? Customer: Van insurance, for my business. Bot: Got it. Is the van used just by you, or do you have other drivers? Customer: Just me and occasionally my son. Bot: Fine. We'll need his details later for the quote, but let's get the basics first. What's the make and model?
That's the same data capture, and it feels like being helped rather than processed.
The questions before the quote
Before anyone will start a quote at all, they want a few things answered, and these are where a bot earns its place most easily:
- Do you cover [my situation]? Modified vehicles, business use, a listed building, a previous claim, a conviction.
- What does the policy actually cover?
- What's excluded?
- How does the excess work?
- Do I need to declare X?
That last one is asked constantly and is genuinely important. People under-declare because they don't understand what's material, and a policy voided at claim stage is the worst outcome for everyone. A bot explaining what needs declaring, from your own published guidance, is doing useful work.
The regulatory boundary
This needs to be firm and explicit, because insurance is regulated.
Fine: factual product information from your published documents. What a policy covers, what's excluded, how excess works, what the claims process is, what documents are needed, how your service works, what your fees are.
Not fine: any recommendation. Which policy someone should choose. Whether a level of cover is adequate for them. Whether they should claim. Whether they're over- or under-insured. Anything applying products to a person's circumstances.
The correct response to "which one should I go for?" is that it depends on the specifics and an adviser will go through it. Shall I get someone to call you?
Configure it to hand over readily and to state plainly that it provides information, not advice. Put that in the opening message.
Claims: process, never outcome
Claims questions are emotionally loaded and arrive at all hours, after an accident, a burglary, a flood.
A bot can genuinely help here by explaining what to do first, what to photograph, what to keep, who to notify, what documents are needed, and how long the process typically takes. That's practical guidance at a moment when people are stressed and can't find anything.
What it must never do is indicate whether something is covered or likely to be paid. Coverage decisions depend on the policy wording and the facts, and a bot suggesting a claim will be met (or worse, won't be) creates a serious problem.
Configure a clean line: process yes, outcome never, and route to the claims team.
Renewals
There's a retention use worth building. A large share of insurance contact around renewal time is people asking why the price went up, whether they can change the cover, and how to cancel auto-renewal.
Handled badly (or slowly), these become lost customers. Handled by a bot that explains the process clearly and routes to a person who can actually adjust the policy, they become conversations that retain business.
Renewal churn is worth examining closely. Much of it comes from customers who could not get through quickly enough to discuss an increase, not from customers who found a better price elsewhere. Those are very different problems with very different fixes.
Capture and routing
For a quote enquiry, capture whatever your quoting process needs, in a sensible order, and route by product line (motor, home, commercial, liability), because different advisers handle each.
Flag anything with a deadline: a policy expiring this week, a vehicle needing cover today, a contract requiring proof of insurance. Those convert at a much higher rate and should be called first.
Data handling
Insurance enquiries involve personal and sometimes sensitive information. Two rules.
Don't collect more than the quote needs, and don't collect medical or criminal detail in a chat window. Route those to a secure process.
And be explicit about what happens to the information. A short, clear line at the start about why details are being taken and who will see them costs nothing and prevents the abandonment that vagueness causes.
What setup involves
Point it at your product pages, policy summaries, and claims guidance. Upload policy wordings. That's where the exclusion answers actually live.
Set the advice boundary first. Set product-line routing. Set the deadline flag.
What to watch afterwards
Completion rate: the share of people who start a quote conversation and finish it, against your old form's completion rate. That comparison is usually stark, and it's the whole argument.
Then track enquiries captured outside office hours, because insurance shopping (like most admin nobody enjoys) happens in the evening.