Get started

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

Claims and policyholder service · Mid level

How to assess a Claims Handler

The technical paper tests file adjudication and the competency interview tests self-description; the job is a third thing neither of them touches, which is defending a correct decision out loud to the person it costs. These are genuinely separate competencies and they are rarely strong in the same person. A handler who reasons cleanly through wording but folds when the claimant gets upset will settle claims that should be declined, quietly and at scale, and their file notes will look perfect. A handler who is excellent on the phone and sloppy with the wording creates the mirror problem and a paper trail a regulator can read. Any claims director can split their own complaint volume into decisions that were wrong and decisions that were defensible but badly explained; it is worth asking which pile is bigger before designing the screen, because the current process is aimed almost entirely at the first pile. The specific unobserved act is the promise: "I'll get this approved for you," offered by a handler with no authority to approve it, is the single most common way a defensible claim becomes an unarguable one.

The claims handler owns the decision call. Somewhere in the life of a claim there is a conversation in which a person is told that their settlement is smaller than they expected, or that a portion is excluded, or that depreciation has been applied, or that the claim is declined outright — and the handler has to make that decision comprehensible to somebody who has an immediate financial interest in not accepting it. This is a distinct job from adjudicating the file, which is why this corpus separates the conversational half of insurance from the adjudication half in `risk-compliance-insurance-operations`. The handler must do both, but only one of them is currently screened.

The regulated behaviours here are unusually well specified, which makes the seat one of the most cleanly assessable in the corpus. Two are worth naming precisely. First, the duty to explain. NAIC Model 900 prohibits failing, in the case of claims denials or offers of compromise settlement, to promptly provide a reasonable and accurate explanation of the basis for that action; California puts the same duty in writing at 10 CCR 2695.7(b)(1), which requires a denial to be made in writing with a statement listing all bases for the denial and the factual and legal bases for each reason then within the insurer's knowledge. A handler who reads the exclusion clause aloud has not explained anything — they have performed the document at the claimant. A handler who says which fact in this file engages which provision, in ordinary words, has satisfied the duty and usually shortened the dispute. The difference is visible in a transcript and in a letter and invisible in an interview.

Second, the promise that is not the handler's to make. Coverage and settlement authority sit at defined levels, and a handler who tells a claimant on a call that they will get an approval, or a payment by a date, or a waiver of an excess, has created an expectation the insurer must either honour or publicly retract. Retracting it is where the regulatory exposure appears: NAIC Model 900's prohibition on knowingly misrepresenting relevant facts or policy provisions is the obvious hook, and in the UK the equivalent complaint reaches the Financial Ombudsman Service as a "you told me it was covered" case. The behaviour is a single sentence, it is spoken under a specific and predictable kind of pressure — the claimant is crying, or has said they cannot pay for the repairs, or has invoked a relative's illness — and it is trivially observable in a simulation that deliberately engineers that moment. It is essentially never observed before hire.

The third behaviour is the reverse failure, and the one that costs insurers most in aggregate: settling at the number that ends the argument. California's 2695.7(g) prohibits settlement offers that are unreasonably low, which is the regulator's side of the same problem; the insurer's side is the handler who pays above the policy position because the alternative is another twenty minutes of conflict. Both directions are gradeable against the same evidence, because the file states what the policy supports and the transcript states what was offered.

Savvanta assesses this in three stages that mirror the work: a structured judgment scenario over a real-shaped file (policy wording, an estimate, a photograph set, and one fact that materially changes the outcome), then a live call with an AI claimant who is articulate, upset and pushing hard for a commitment, then the written decision letter. The rubric scores whether the decision on the call matches the decision in the file, whether any commitment made was within authority, and whether the letter states the specific basis rather than a conclusion. It does not score how comfortable the call sounded, and it does not score accent, dialect or manner of speech. Licensure is a separate employer gate: adjuster licensing requirements vary by state, some states require prelicensing education or an exam, and a simulation ranks handling ability without ever verifying a credential.

What the job actually needs

How people fail in this seat

What most employers do instead

A CV screen for claim types and system names, a technical claims interview or scenario paper on policy interpretation, and a competency interview.

The technical paper tests file adjudication and the competency interview tests self-description; the job is a third thing neither of them touches, which is defending a correct decision out loud to the person it costs. These are genuinely separate competencies and they are rarely strong in the same person. A handler who reasons cleanly through wording but folds when the claimant gets upset will settle claims that should be declined, quietly and at scale, and their file notes will look perfect. A handler who is excellent on the phone and sloppy with the wording creates the mirror problem and a paper trail a regulator can read. Any claims director can split their own complaint volume into decisions that were wrong and decisions that were defensible but badly explained; it is worth asking which pile is bigger before designing the screen, because the current process is aimed almost entirely at the first pile. The specific unobserved act is the promise: "I'll get this approved for you," offered by a handler with no authority to approve it, is the single most common way a defensible claim becomes an unarguable one.

The assessment

About 38 minutes end to end.

The systems it runs in

A claims management system holding the claim file end to end: the policy and its wording, the exposures, the reserve with its history, the payment and settlement actions with the authority limit that gates them, the claim status, the diary and activity trail, and the correspondence step that produces the decision letter. Guidewire describes ClaimCenter as governing the entire claims lifecycle from intake to closure; Duck Creek describes its claims product as accelerating claims resolution through business-led workflows with configurable assignment rules. This design puts the adjudication into that record rather than into a text box, which is what makes the authority criterion observable: an authority limit is a control in the system, and whether the candidate worked inside it, asked for it to be raised, or attempted to act past it is a logged event rather than an inference from what they said.

Any claims system with a claim file, a reserve the handler can set and change, a settlement or payment action gated by an authority limit, a status, and a correspondence step. The file is an escape of water claim and the design already tells buyers that performance on it generalises less well to liability, motor injury or business interruption work than they expect; a buyer hiring for another line should have the file rebuilt in their own system rather than treat the score as transferable. Authority levels stay in the brief, as they are now, so a candidate new to the employer is not penalised for not knowing them.

Working speed is not scored. Both named failures in this seat are fast ones. Paying the contractor's estimate is quicker than separating the covered damage from the excluded cause, and saying the claim will be approved ends the call instantly. Scoring time to decision or claims closed per day would rank exactly the handler this design exists to catch, and the wage arithmetic that justifies thirty-eight minutes is the same arithmetic that says a wrong settlement costs more than a slow one. Claimant satisfaction and call duration are not scored either, for the reason the adverse_impact section gives — in this seat satisfaction and correctness frequently point in opposite directions.

What the candidate actually does

TaskWhat happens
The file, before anybody is on the phone
judgment_scenario · 12 min		A real-shaped escape of water file: policy schedule, the relevant wording extracts, a contractor's estimate, a described photograph set, the claimant's own statement, and the excess. Two things in the file materially change the answer. The estimate prices a full kitchen replacement where the photographs and the claimant's statement support damage to four units, and the wording covers damage caused by the escape of water but excludes the cost of repairing the failed pipe itself, which the estimate has lumped into a single line. The candidate records the decision on the claim — the coverage position, the reserve, the amount they will settle at, the basis for each element, and the one question that must be answered before the decision is final — with the brief's authority limit in force, so a figure above it can only be recorded as a referral rather than as a decision. The cheap path is to pay the estimate: it is the only number in the file, it is from a contractor, and nobody is arguing with it yet.
The decision call
live_call · 14 min		An AI claimant who is articulate, prepared, not abusive, and pushing hard. They escalate in four moves: they ask for the number, they say a neighbour's insurer paid for a whole kitchen, they say they cannot fund the shortfall and that their mother is living with them, and finally they ask the candidate to get it approved. Two forks run in opposite directions. The obvious one is the promise, the sentence that is not the handler's to make and that ends the call instantly. The less obvious one is the number, where the temptation is to drift toward the estimate figure to stop the argument; and the mirror temptation, in candidates who have learned that firmness is the tested virtue, is to offer below the position the file supports in order to leave negotiating room.
The decision letter
written_artifact · 12 min		The written decision the claimant receives after that call. The cheap version is a conclusion and a clause reference, which is what most denial and partial-settlement letters are, and which is what generates the appeal. The correct version lists each basis, names the fact in this file that engages it, separates what is being paid from what is not and why, states the excess and its effect, and tells the claimant how to challenge it.

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.

The position taken on the call is the position in the fileweight 0.2The same position and the same number as the adjudication, stated plainly on the call, and where the candidate does move, they name the new fact that …
1 The amount or the coverage position stated on the call differs from the candidate's own written adjudication, with no reason given for the change.
3 The same position, but the candidate declines to state any number on the call, so the claimant is given the decision without the figure and has to wait for the letter to find out what it means.
5 The same position and the same number as the adjudication, stated plainly on the call, and where the candidate does move, they name the new fact that moved them.
Explains the basis rather than performing the wording (NAIC Model 900, section 4.L, failing in the case of claims denials or offers of compromise settlement to promptly provide a reasonable and accurate explanation of the basis for such actions; 10 CCR 2695.7(b)(1), a statement listing all bases and the factual and legal basis for each)weight 0.25States in ordinary words which specific fact in this file engages which provision, separates the part being paid from the part that is not, explains t…
1 Reads or paraphrases the exclusion and repeats it more slowly when challenged. The claimant is never told which fact in their own claim engages it.
3 Names the clause and the fact once, in the language of the policy, and moves on without checking whether the claimant followed it.
5 States in ordinary words which specific fact in this file engages which provision, separates the part being paid from the part that is not, explains the excess and its effect on the figure, and checks the claimant has understood before moving on.
Makes no commitment outside the stated authority (NAIC Model 900, section 4.A, knowingly misrepresenting to claimants and insureds relevant facts or policy provisions relating to coverages at issue)weight 0.2States exactly what the handler will do, names who decides what the handler cannot, gives a date by which the claimant will hear, and repeats it uncha…
1 Says the claim will be approved, that a payment will arrive by a date, or that the excess will be waived or reviewed favourably.
3 Offers an implied commitment with no substance, such as pushing for it or seeing what can be done, with no named decision-maker and no timescale, which the claimant will reasonably hear as a promise.
5 States exactly what the handler will do, names who decides what the handler cannot, gives a date by which the claimant will hear, and repeats it unchanged when the pressure increases.
Holds the number in both directions (10 CCR 2695.7(g), no settlement offer that is unreasonably low; NAIC Model 900, section 4.D, not attempting in good faith to effectuate prompt, fair and equitable settlement of claims in which liability has become reasonably clear)weight 0.2Holds the figure the file supports, engages with the specific argument the claimant made including the neighbour comparison, and names what evidence w…
1 Moves to the contractor's estimate figure after the claimant's third push, or opens below the position the file supports in order to leave room to negotiate.
3 Holds the figure, but only by declining to engage with the claimant's argument, so the claimant is left with a number and no answer to the point they raised.
5 Holds the figure the file supports, engages with the specific argument the claimant made including the neighbour comparison, and names what evidence would legitimately change it.
The letter states the bases, not the conclusion (10 CCR 2695.7(b)(1) and 2695.7(b)(3), all bases with the factual and legal basis for each, and notice of the route to review)weight 0.15Lists each basis with the fact from this file that engages it, shows how the figure was arrived at including the excess, and states how the claimant c…
1 States the outcome and cites a clause number, with no facts and no arithmetic, so a reader cannot tell why this claim produced this figure.
3 Gives the reasons in the policy's own language, accurately, but does not connect each one to the specific fact in this file, and omits the route to challenge it.
5 Lists each basis with the fact from this file that engages it, shows how the figure was arrived at including the excess, and states how the claimant can have the decision reviewed.

How it is scored

Weighted mean of the five criteria, each scored 1 to 5 against the anchors, reported with the file entry, transcript excerpt or letter paragraph that earned each score. The report shows three numbers side by side: the amount in the candidate's adjudication, the amount stated on the call, and the amount in the letter. Divergence between them is reported as a fact rather than folded into a score, because it is the fastest way for a claims director to see the failure this seat is actually hired against.

Integrity

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

Who decides

Required. Two decisions belong to a person. The first is whether a sentence on the call was a commitment: the 3 anchor covers the implied version, and whether a particular I will see what I can do crossed the line depends on what surrounded it, which a reviewer settles from a surfaced excerpt in seconds and an automated scorer settles badly. The second is whether the candidate's adjudication was defensible, because the scenario has a correct shape but not a single correct number, and a candidate who reached a different figure by sound reasoning must not be marked down for disagreeing with the model answer. The reviewer reads the adjudication, the time-marked commitments, and the paragraph of the letter that states the bases. The ranking entitles the buyer to conclude that this candidate adjudicated this file in this way and defended or abandoned that position under this specific pressure. It does not establish adjuster licensure, which varies by state, sometimes requires prelicensing education or an examination, and which no simulation verifies; and it does not certify competence in lines of business the scenario did not cover.

What this does not measure

The design does not read accent, dialect, fluency or manner of speech, and no anchor is satisfiable by sounding authoritative or sympathetic. The criterion most at risk of leakage is the explanation criterion, because a fluent speaker can make a bad explanation sound good, so its anchors are written around whether a specific fact was connected to a specific provision, which a reviewer checks against the file rather than against their impression of the delivery. The design does not score how comfortable the call sounded, how long it lasted, or whether the claimant ended it satisfied, and it deliberately does not score claimant satisfaction at all, because in this seat satisfaction and correctness frequently point in opposite directions. It does not test knowledge of any particular insurer's authority levels or claims system: the authority limits are stated in the brief, so a candidate new to this employer is not penalised for not knowing them. Two limitations should be stated to buyers. First, the file is an escape of water claim, and performance on it generalises less well to liability, motor injury or business interruption work than buyers usually assume; a buyer hiring for a different line should ask for the file to be rebuilt rather than treat this score as transferable. Second, candidates whose prior training emphasised talking the claimant round rather than explaining the decision will score low on the explanation criterion for habitual reasons, and that is retrainable. Written English is graded for whether the basis is stated, never for idiom, spelling or punctuation, and deployers should monitor for a gap that appears only on the letter.

The claims handler is hired on file accuracy and fails on the call. That is the whole argument for this design, and it is testable against the buyer's own data before they buy anything: every claims operation can split its complaint volume into decisions that were wrong and decisions that were defensible but badly explained, and it is worth asking a claims director which pile is bigger before designing the screen. The current process, a technical scenario paper plus a competency interview, aims almost entirely at the first pile. The second pile is where the money is and nothing in the process observes it.

Sequencing the adjudication before the call is what makes the rest of the design work. Without it there is no reference point, and any judgment about whether the candidate caved is an impression. With it, the report can show the number the candidate computed, the number they said out loud when the claimant pushed, and the number in the letter, and a reviewer can see the drift directly. That drift is the single most diagnostic output of this assessment, and it is unavailable to any screen that assesses adjudication and communication as separate exercises, which is what the market currently does.

The regulated behaviours are unusually well specified here, which is why this seat is one of the cleanest in the corpus to assess. NAIC Model 900 section 4.L makes it an unfair claims practice to fail, on a denial or an offer of compromise settlement, to promptly provide a reasonable and accurate explanation of the basis for the action, and California puts the same duty in writing at 10 CCR 2695.7(b)(1), which requires a statement listing all bases and the factual and legal basis for each then within the insurer's knowledge. Reading the exclusion aloud does not satisfy that. It performs the document at the claimant, and it is the commonest thing a competent, nervous handler does when the call gets hard. The distinction between performing the wording and explaining which fact engages which provision is visible in a transcript, visible in a letter, and completely invisible in an interview.

The promise is the other named failure and it is the one that turns a defensible claim into an unarguable one. Coverage and settlement authority sit at defined levels, and a handler who says on a call that they will get it approved has created an expectation the insurer must either honour or publicly retract. The scenario engineers the moment precisely: the claimant is articulate, has a comparison to a neighbour, has a dependent relative in the house, and cannot fund the shortfall. None of that is manipulative, all of it is ordinary, and it is exactly the combination under which the sentence gets said. The 3 anchor is written around the implied version, seeing what can be done, because that is the form the failure actually takes and treating it as partial credit toward good practice would be wrong.

Scoring the number in both directions is a deliberate correction to how firmness gets trained. An assessment that only punished caving would select for handlers who lowball, which is the failure 10 CCR 2695.7(g) exists to address and which costs the insurer differently but not less. The 1 anchor on that criterion therefore covers both moving to the estimate and opening below the file position, and the 5 anchor requires the candidate to engage with the claimant's specific argument rather than merely to hold firm. Holding a number by refusing to discuss it is a 3, not a 5, and that distinction is the one that separates the top quartile in this seat from the merely unmovable.

At thirty-eight minutes this is the longest design in the family and it is justified by the wage. The median US claims adjuster wage recorded by BLS for May 2025 is $78,020, every opening in the occupation is replacement hiring, and a mis-hire is expensive twice over, in the ramp and in the claims they settle wrongly on the way out. The same forty minutes spent on an entry-level intake seat would not be worth running.

Sources

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

  1. 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
  2. California Code of Regulations Title 10, section 2695.7(b), (b)(1) and (g): accept or deny within forty calendar days; denials in writing with all bases and the factual and legal basis for each; no settlement offer that is unreasonably low, https://www.law.cornell.edu/regulations/california/10-CCR-2695.7
  3. NAIC Unfair Claims Settlement Practices Act (Model 900), prohibiting knowing misrepresentation of relevant facts or policy provisions, and failing in the case of claims denials or offers of compromise settlement to promptly provide a reasonable and accurate explanation of the basis for such action, https://content.naic.org/sites/default/files/model-law-900.pdf

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 reportOr talk to us about this role