Taking an assessment rather than buying one? This page is written for employers. Here is the page for candidates.

Claims and policyholder service · Entry level

How to assess a First Notice of Loss Agent

The FNOL call is two jobs performed simultaneously by two parts of the brain that pull against each other: it is an emotional-support conversation and a structured data-capture exercise, and the current process tests each one separately or neither. The competency interview asks about empathy in the abstract, in a calm room, with the candidate narrating their own past behaviour. The typing test measures capture with nothing emotional attached. Nobody watches what actually happens on the floor, which is that under emotional load the agent picks one and drops the other — either they become warm and the file arrives incomplete, or they get the fields filled and the claimant hangs up feeling processed. Worse, nothing observes the single most expensive sentence in this seat: the reassurance. "That'll definitely be covered" is spoken by junior agents every day because it makes an unbearable call bearable, and it is not theirs to say — coverage is the adjuster's determination, and pre-empting it sets up a decision that the insurer must either honour or retract.

The first notice of loss call is the emotional peak of the entire insurance relationship, and it is staffed by the most junior people in the operation. It happens after a house fire, a road collision, a bereavement, a flooded shop floor — very often on the worst day the caller has had that year — and the person answering has to do something structurally strange: be a human being about it while extracting a set of facts precise enough that an adjuster three weeks later can make a coverage decision without recontacting the claimant. Those two demands compete for the same ninety seconds of attention, repeatedly, for the length of the call.

Everything downstream inherits the quality of this transcript. A loss narrative captured in the claimant's own sequence — "I was going to the shops, well, first I'd dropped my daughter off, the lights were out at the junction" — is not a timeline, and an intake agent who transcribes it as spoken has produced a file that costs the adjuster a callback and the insurer a cycle. An agent who reorganises it into a sequence, checks the reordering back with the caller, and notes explicitly which parts the claimant was uncertain about has done something that materially shortens the claim. That is a skill, it is not correlated with warmth, and no interview has ever tried to observe it.

The regulated behaviour in this seat is narrower than in claims handling but it is real and it is the same in almost every jurisdiction. The NAIC's Unfair Claims Settlement Practices Act (Model 900), which most US states have adopted in substance, prohibits knowingly misrepresenting to claimants relevant facts or policy provisions relating to the coverages at issue, and separately prohibits failing to acknowledge with reasonable promptness pertinent communications about claims. The FNOL agent breaches the first not by lying but by comforting: telling a distressed caller the loss will be covered, or that the excess will be waived, or that a hire car is automatic, when none of that has been determined. It is almost always well-intentioned and it converts a straightforward claim into a complaint the moment the adjuster applies the policy as written. The second duty turns into the expectation-setting behaviour: the claimant should end the call knowing what happens next, who will contact them, and roughly when. In California the outer bound of that "next" is concrete — 10 CCR 2695.7(b) requires the insurer to accept or deny within forty calendar days of receiving proof of claim — which gives the assessment a factual reference point rather than a vibe.

There is a third behaviour, and it needs stating carefully because it is where this territory can go badly wrong. Loss narratives sometimes contain inconsistencies — a reported time that does not fit the documentation, a description of damage that does not match the photographs. Noticing and recording that is a genuine, valuable competency and it is assessable. What must never be assessed is a judgment about the claimant as a person. Any scenario that rewards a candidate for treating a caller as suspicious on the basis of hesitancy, manner, accent or circumstance is building a discrimination engine with a fraud label on it. The mark scheme should require that any escalation cite a specific fact in the file, and should score an escalation that cites the caller's demeanour as a failure, not a partial credit.

Savvanta runs this as a live call with an AI claimant who is upset, out of sequence, and who mentions in passing a detail that determines coverage — plus the claim file the candidate writes immediately afterwards, which is scored against what was actually said on the call. Licensure is out of scope: adjuster licensing requirements vary by state, and a simulation ranks handling ability and never verifies a credential. Nothing in the rubric touches accent, dialect or manner of speech.

What the job actually needs

capturing a complete and accurate loss narrative from a distressed caller
sequencing an out-of-order account into a timeline
setting expectations about what happens next without predicting an outcome
recognising and flagging an urgent safety or hardship need
recording a factual inconsistency without accusing anyone

How people fail in this seat

tells the caller "you're covered
don't worry" before anyone has read the policy
collects the fields in form order while the caller is trying to describe what happened
misses the detail that determines coverage because it arrived mid-sob
records the claimant's distress as a note instead of capturing the facts
closes the call without the claimant knowing who contacts them and when

What most employers do instead

A CV sift, a competency interview about going the extra mile for a customer, and occasionally a data-entry or typing test.

The assessment

About 28 minutes end to end.

The systems it runs in

The intake end of a claims management system: a policy search that retrieves the policy in force, a guided notice-of-loss form that asks response-driven questions, the loss date, time and location, contact records for the caller and anyone else involved, the claim number the system issues, routing to an adjuster or queue, and flags for an urgent need and for a referral. Guidewire describes ClaimCenter's intake in those terms — fast digital claims intake integrated with policy search and retrieval, using wizard-based dynamic, response-driven questions, governing the claim from intake to closure. Duck Creek lists a distinct first notice of loss offering built around conversational intake and automated workflow orchestration. The claim file the adjuster reads three weeks later is a record in that system, not a document, and this design scores it as one.

Guidewire ClaimCenter
Duck Creek Claims

Any claims system with a policy lookup, a structured notice-of-loss intake, a claim record that carries a narrative and dated facts, and a route to an adjuster or an urgent-need queue. One departure from every real instance is deliberate and is kept whatever system the fixture is built in: the file is written after the call rather than during it, which trades fidelity for the removal of a concurrent-data-entry construct that would disadvantage candidates with motor or visual impairments while measuring nothing about intake quality. Rebuilt against the buyer's own intake form and loss types where they supply a sandbox.

Working speed is not scored. Speed is how this seat breaks. The cheap path on all three forks is the fast one and it is also the kind one: reassure, let the story be told in the order it arrives, and close. The unoccupancy detail is missed by an agent moving quickly through a form, and the coverage promise is what makes a hard call end sooner. Scoring handling time would rank the agent who produced the incomplete file above the one who asked the two questions that pinned the dates. Typing speed and concurrent data entry are excluded for the reasons the adverse_impact section gives, and calls per hour is a staffing and welfare question rather than a selection one.

What the candidate actually does

Task	What happens
The 6am call live_call · 13 min	An AI caller reporting an escape of water discovered that morning in a property they had returned to the night before. They are distressed, they tell the story out of sequence, and they circle back twice. Three things are planted. First, a coverage-determinative detail arrives as an aside explaining why nobody noticed sooner, that the house had been empty for about six weeks while they were staying at their mother's, and it is never repeated. Second, the caller asks directly, and then again more insistently, for confirmation that it is covered, which is the single sentence that makes an unbearable call bearable and which is not the agent's to say. Third, the caller says they have nowhere to sleep tonight, which is a real and urgent need that has to be routed without a coverage promise attached to it. The cheap path on all three is the same and is genuinely kind: reassure, and let the story be told in the order it arrives.
The file the adjuster reads three weeks later written_artifact · 10 min	The claim record, created immediately after the call and scored against the candidate's own transcript: the structured fields the notice of loss asks for, the routing of the hardship need and of any referral to the queue the brief names, and the loss narrative itself. The task is to produce a timeline rather than a transcript: the events in the order they happened, the facts separated from the caller's inferences, the unoccupancy detail captured with the dates it covers, and the things the caller was unsure about marked as uncertain rather than smoothed into assertion. The file is written after the call and not during it, which is a deliberate departure from the real job.
What to escalate, and on what judgment_scenario · 5 min	Four short items about whether and how to flag something onward. One is a genuine inconsistency between the reported time of discovery and a document in the file. One is a distractor that invites escalation because the caller sounded rehearsed and changed their wording. One is an urgent hardship need. One is a communication from the claimant that expects a response and is sitting unanswered. The distractor is scored as a failure and not as partial credit, which candidates are told after the fact and buyers are told before.

The mark scheme

Each criterion is scored 1 to 5 against written anchors, and every score is reported with the excerpt that earned it. A criterion marked floored is reported as a finding rather than averaged into the total. The first is open; open any other to read its anchors in full.

Does not confirm cover, waive an excess, or promise a benefit (NAIC Unfair Claims Settlement Practices Act, Model 900, section 4.A, knowingly misrepresenting to claimants and insureds relevant facts or policy provisions relating to coverages at issue)weight 0.25Says plainly that the coverage decision belongs to the adjuster and is not made on this call, says what will actually happen and by when instead, and …

1 Says or agrees at any point that the loss is covered, that the excess will be waived, or that alternative accommodation is included, including by answering a direct question with yes.

3 Deflects with a phrase like I could not say, and moves on, so the caller is no better informed, asks again later, and gets the same deflection.

5 Says plainly that the coverage decision belongs to the adjuster and is not made on this call, says what will actually happen and by when instead, and routes the accommodation need as an urgent need without attaching any coverage promise to it.

Captures the detail that determines coverageweight 0.2Picked up in the turn it was said or before the call ends, followed with at least one question that pins the start and end dates, and recorded in the …

1 The unoccupancy detail appears nowhere in the transcript after the caller says it, and nowhere in the file.

3 It appears in the file as a copied phrase, with no follow-up question on the call, so the dates it covers are unknown.

5 Picked up in the turn it was said or before the call ends, followed with at least one question that pins the start and end dates, and recorded in the file in the caller's own words with those dates.

The file is a timeline, not a transcriptweight 0.2Reordered, read back to the caller for confirmation during the call, and items the caller was unsure about are marked as uncertain in the file rather …

1 Events are recorded in the order the caller said them, circle-backs included, so the reader has to reconstruct what happened when.

3 Reordered into a sequence, but the reordering was never checked back with the caller, and nothing distinguishes what the caller knew from what they assumed.

5 Reordered, read back to the caller for confirmation during the call, and items the caller was unsure about are marked as uncertain in the file rather than stated flat.

Leaves the caller with a next step they can hold (Model 900 section 4.B, failing to acknowledge with reasonable promptness pertinent communications with respect to claims; 10 CCR 2695.5(e), acknowledgement and first actions within fifteen calendar days; 10 CCR 2695.7(b), accept or deny within forty calendar days of proof of claim)weight 0.15States who will make contact, within what period, using the timescale in the brief, states what the caller needs to do before then, and checks the cal…

1 The call ends without the caller being told who will contact them or when, or with a timescale the brief does not support.

3 Gives a timescale but not a person or a team, or a person but no timescale, so the caller has half of what they need to chase it.

5 States who will make contact, within what period, using the timescale in the brief, states what the caller needs to do before then, and checks the caller has it.

Any escalation cites a fact in the file, never the caller's demeanourweight 0.2States the specific fact and the specific document it conflicts with, records it neutrally without characterising the caller, and routes it to the que…

1 Escalates or flags on the basis of the caller's hesitancy, emotional state, rehearsed manner or way of speaking. This is a 1 even where the genuine documentary inconsistency was also identified.

3 Notes the timing inconsistency but does not say what it conflicts with, or records it as a general concern about the claimant rather than about the record.

5 States the specific fact and the specific document it conflicts with, records it neutrally without characterising the caller, and routes it to the queue the brief names.

How it is scored

Weighted mean of the five criteria, each scored 1 to 5 against the anchors, reported with the time-marked transcript excerpt or file line that earned each score. The reassurance criterion is additionally reported as a raw count of the sentences in which any coverage, excess or benefit was affirmed, because that count is the single number a claims director in this seat wants and it should not be buried inside an average.

Integrity

monitored session with the call and the file written back to back
the file is scored against the candidate's own transcript, so a prepared template cannot fit it
one live follow-up question asking the candidate what the caller said about the period the property was empty
timing anomalies between call end and file submission
scenario variants rotated so the planted detail and the loss type differ between sittings

The log describes what happened. It does not produce a cheating verdict — the follow-up conversation is the control, because a statistical accusation is not something we would ask a reviewer to defend.

What you receive

full call transcript with the coverage question, the planted detail and the hardship disclosure time-marked
the claim record as submitted, being the structured intake fields, the loss narrative, and the routing of the hardship need and of any referral
the four escalation answers with their stated bases
per-criterion score with the excerpt or file line that earned it

Who decides

Required for the escalation criterion and recommended for the rest. A person must read every escalation the candidate wrote and decide whether its stated basis is a fact in the record or an impression of the caller, because that distinction is the difference between a useful competency and a discrimination engine, and it is not a judgment to leave to a score. The reviewer is looking at three things: the sentences in which coverage was discussed, the turn after the unoccupancy detail, and the written basis of any flag. The ranking entitles the buyer to conclude that this candidate did or did not capture the material detail, did or did not promise cover, and escalated on facts or on impressions, in this call. It does not establish that the candidate holds an adjuster licence, which several states require and which no simulation can verify, and it does not predict how the candidate performs on the twentieth call of a catastrophe surge day, which is a staffing and welfare question rather than a selection one.

What this does not measure

This is a voice role and nothing in the rubric reads accent, dialect, fluency, vocabulary or manner of speech. That statement carries more weight here than anywhere else in this family, because the folk theory of a good intake agent is that they sound warm, and warmth is exactly the impression that correlates with familiar speech patterns and with nothing about file quality. Every anchor above resolves to a question asked, a fact captured, a promise made or not made, or a line in a document. The design also does not score typing speed or concurrent data entry: the file is written after the call rather than during it, which is a deliberate departure from the real job that trades some fidelity for the removal of a keyboard-speed construct that would disadvantage candidates with motor or visual impairments while measuring nothing about intake quality. It does not score emotional composure as a trait, and the AI caller is distressed but never abusive to the candidate. The fraud-adjacent item is the most dangerous part of any assessment in this territory, and it is handled by making demeanour-based escalation a 1 rather than partial credit; buyers should be told this before deployment, because a buyer who wants a suspicion instinct measured is asking for something this design will not provide. The written task grades content and not spelling, punctuation or idiom, which matters in a seat staffed heavily offshore. Deployers should monitor pass rates by hiring channel and language background and should look specifically for a gap that appears only on the written file.

The first notice of loss call asks one person to do two things that pull against each other, in the same ninety seconds, repeatedly, for the length of the call. It is an emotional-support conversation and it is a structured data-capture exercise, and under load almost everybody drops one of them. Either the agent becomes warm and the file arrives incomplete, or the fields get filled and the claimant hangs up feeling processed. Current screening tests each half in isolation or neither: a competency interview about empathy in a calm room, and sometimes a typing test with nothing emotional attached. This design puts both demands in the same conversation, which is the only way the trade-off becomes visible.

The reassurance fork is the reason this seat is worth assessing at all. NAIC Model 900 section 4.A makes it an unfair claims practice to knowingly misrepresent to claimants relevant facts or policy provisions relating to the coverages at issue, and the breach here is almost never a lie. It is comfort. A junior agent on a call with somebody who has water coming through a ceiling is asked a direct question, twice, and says yes, because saying yes is the humane thing available in that moment and because the consequence lands on somebody else three weeks later. The scenario makes the question direct and escalating, and it adds a genuine hardship, nowhere to sleep tonight, so that the pressure to promise something is not abstract. The 5 anchor requires the candidate to route the hardship as a real and urgent need while keeping the coverage promise out of it, which is the actual competency and which almost nobody has articulated before they are asked to do it.

The capture fork is quieter and is the one that costs the insurer most in aggregate. The unoccupancy detail arrives inside a sentence about something else, said once, as an explanation for why nobody noticed sooner. An agent who hears it, asks when and until when, and records it with dates has materially changed what happens to this claim. An agent who does not has produced a file that costs the adjuster a callback and the insurer a cycle. This is a skill, it is not correlated with warmth, and no interview has ever tried to observe it, which is why it carries a fifth of the weight here.

The timeline criterion is the one that most surprises buyers. A loss narrative told in the claimant's own sequence is not a timeline, and transcribing it as spoken is the default behaviour of a sympathetic agent who is trying not to interrupt. Reorganising it into a sequence, checking the reordering back with the caller, and noting explicitly which parts the claimant was uncertain about, is a separate and teachable competency that shortens the claim. Scoring the file against the candidate's own transcript is what makes it gradeable: a reviewer can see whether a fact in the file was actually said, and whether an uncertainty was quietly promoted into an assertion.

The escalation task needs the most careful handling in the corpus, and this design deals with it by refusing to give any credit for the wrong basis. Loss narratives do sometimes contain real inconsistencies, and noticing that a reported time does not fit a document is a genuine and valuable competency. Judging the claimant as a person is not, and a scenario that rewarded a candidate for treating a caller as suspicious on the basis of hesitancy, manner or circumstance would be building a discrimination engine with a fraud label on it. So the item set contains a deliberate distractor, the caller who sounded rehearsed, and escalating on it scores 1 even when the documentary inconsistency was also found. A buyer who wants a nose for fraud measured should be told plainly, before deployment, that this assessment will not measure it, and why.

At twenty-eight minutes this is the shortest design in the family, which is deliberate. It is the highest-volume, lowest-paid seat in claims, hired in cohorts, and an assessment that costs more than the recruiter's time will not be run. Every minute in it is spent on a fork; there is no warm-up, no personality inventory and no untimed reading.

Sources

Every figure on this page is traceable. Where a claim could not be sourced it is stated qualitatively instead.

US Bureau of Labor Statistics, Occupational Outlook Handbook, Claims Adjusters, Appraisers, Examiners, and Investigators, 2025, https://www.bls.gov/ooh/business-and-financial/claims-adjusters-appraisers-examiners-and-investigators.htm
IT and Business Process Association of the Philippines (IBPAP), The Philippine IT-BPM Industry Overview 2026, 2026, https://admin.ibpap.org/storage/hub-resources/7Hv9U2uoLJVx2kxyRMObNm6W6L0MR57l9dZfrez7.pdf
NAIC Unfair Claims Settlement Practices Act (Model 900), prohibiting knowingly misrepresenting to claimants and insureds relevant facts or policy provisions relating to coverages at issue, and failing to acknowledge with reasonable promptness pertinent communications with respect to claims, https://content.naic.org/sites/default/files/model-law-900.pdf
California Code of Regulations Title 10, section 2695.7(b), requiring an insurer to accept or deny a claim within forty calendar days of receiving proof of claim, https://www.law.cornell.edu/regulations/california/10-CCR-2695.7

See what the employer actually receives. A full report for one role, with every score shown beside the excerpt that earned it, conduct findings reported rather than averaged, and a reviewer sign-off required before any decision. No form.

Read a sample report Or talk to us about this role