Menu
Platform
Why JRNYHow it worksBlogFAQ
Request an account
All articles
Insurance7 min read

First notice of loss without the phone tree

Claim intake is a structured interview. Handled conversationally, and written straight into the claims system, it stops being the worst call of the customer week.

PR
Priya Raman
Insurance Lead · August 19, 2026

Nobody files a claim on a good day. The person on the other end of first notice of loss has just had an accident, a flood or a break-in, and the first thing most insurers do is ask them to choose from a menu of six options, none of which describes what happened to them.

The interview itself is not the problem. FNOL is one of the most structured conversations in the business: a fixed set of facts, a fixed set of validations, a fixed handoff into adjudication. The problem is the packaging.

What intake actually needs

Strip a property FNOL down and you get date and time of loss, cause of loss, location, a description in the customer’s own words, whether the property is habitable, whether emergency services attended, and contact preferences. Auto adds vehicles, drivers, injuries and police report numbers. Everything else is enrichment the insurer can do itself.

That list is short enough to hold in a conversation. What makes it feel long on the phone is that the customer cannot see the end of it and has no idea which answers matter.

Ask in the order the customer is thinking

The claims system wants the fields in schema order. The customer wants to tell you what happened. Those two orders are not the same, and the assistant should absorb the difference rather than pushing it onto the person.

In practice that means opening with a free-text description, extracting whatever the customer already gave you, and then asking only for what is still missing. A customer who says “a pipe burst in the upstairs bathroom last night and the kitchen ceiling has come down” has already answered cause, location, approximate time and severity. Asking them for those four fields again is the moment they decide the assistant is not listening.

The strongest predictor of abandonment we have found is not the number of questions. It is being asked something you already said.

Validate at the edge, not at the end

The classic failure of digital FNOL is a form that accepts everything and an adjuster who rejects it three days later. Validation belongs in the conversation, where it costs one extra turn: a policy that was not in force on the date of loss, a vehicle not listed on the policy, a cause of loss excluded by the product.

Not every one of these should be a refusal. Most are a reason to route to a human immediately with the reason attached, so the customer hears it from a person who can explain the consequence.

Write into the system of record, once

An intake that produces a transcript is not intake. The output has to be a claim in the claims system with the right fields populated, an FNOL reference the customer can quote, and any attachments — photos, documents, a police report — already associated with it.

The reference number matters more than it looks. It is the thing that turns “I told someone about this” into “the insurer has this”. Customers who leave the conversation with a reference number call back at roughly half the rate of those who do not.

Where a human belongs

Three categories go to a person straight away, every time: any injury, any total loss, and any conversation where the customer is distressed enough that the interview is not working. The third one is a judgement call, and we bias it heavily towards escalating.

The rest can be completed conversationally and reviewed by an adjuster who opens a claim that is already complete, already validated and already timestamped from the moment the customer first reported it — not from the moment somebody got round to typing it in.

Related reading

All articles
Insurance
Claims status: the question you get asked five times
Priya Raman · 5 min read
What day-of-travel guidance looks like when the AI can read the PSS
Airlines
What day-of-travel guidance looks like when the AI can read the PSS
Daniel Whitaker · 9 min read
Healthcare
Explaining a deductible in plain language
Elena Ruiz · 6 min read