Will AI Replace Claims Adjusters? Following the Claim From First Notice to Settlement

Short answer: the simple, documented, low-value claim is being settled without a person, and the disputed, complex or suspicious claim is not. The US Bureau of Labor Statistics projects claims adjusters, appraisers, examiners and investigators to decline 6 percent between 2025 and 2035, a loss of about 21,800 positions from a 2025 base of 389,700, with 2025 median pay of $78,020. That is one of the steeper declines among knowledge occupations, against 3.5 percent growth projected across all employment. The decline is real and it is concentrated at one specific stage of the claim.

Where Automation Enters the Claim Lifecycle

Following a claim from start to finish shows precisely where the pressure sits.

First notice of loss. Largely automated. A customer reports through an app, uploads photographs, and the system captures the details. This stage barely involves adjusters now.

Triage and segmentation. Automated and improving fast. The claim is routed by estimated complexity, value and fraud indicators. This is where the biggest change has happened, because it determines whether a human ever sees the file.

Damage assessment. Increasingly automated for straightforward property and motor claims. Photograph-based estimating produces a repair figure without a physical inspection.

Coverage determination. Partly automated for clear cases, and difficult for anything ambiguous, which is where policy wording actually matters.

Negotiation and settlement. Not automated where the claimant disputes anything.

Investigation. Not automated. Fraud, liability disputes and large losses require someone who can interview people and evaluate what they say.

The pattern is that the middle of the process has been hollowed out. The straightforward claim now flows from notification to payment with no adjuster involvement, which removes a large volume of routine files.

What Remains, and Why

The rows that remain manual share a feature: someone disagrees, or the facts are contested, or the causation is genuinely uncertain. Automated assessment works well when everyone accepts the same account of what happened.

A claim that no system settles

A homeowner reports water damage to a ground-floor room. Photographs show damaged flooring and staining on the lower walls. The automated estimate produces a repair figure within an hour.

An experienced adjuster looks at the same photographs and sees questions. The staining pattern rises from the floor rather than descending, which is more consistent with ingress than with a burst pipe. There is a hairline crack visible in the render outside the affected wall in one of the exterior shots. The policy excludes gradual damage and covers sudden escape of water, so the distinction between the two is the entire claim.

Resolving it requires an inspection, possibly a drainage survey, and a conversation with the policyholder about when they first noticed the problem. That conversation is delicate: an honest answer of “a few months ago” changes the outcome, and the policyholder knows it. There may also be a prior claim on the same property that suggests a pattern.

The adjuster’s job is to reach a defensible decision and to communicate it to someone who may be about to be told that a large repair is not covered. Getting that wrong produces a complaint, a regulatory referral or a reputational problem, all of which cost far more than the claim. This is why insurers keep experienced people on exactly these files while automating everything simpler.

Why insurers automate settlement even when it costs them money

There is a commercial logic behind straight-through settlement that is worth understanding, because it explains both the speed of adoption and its limits.

For low-value claims, the cost of investigating exceeds the amount in dispute. An insurer that spends an hour of adjuster time examining a small claim has already spent a meaningful fraction of the settlement. Paying quickly is cheaper than checking, and it produces a customer who renews.

That calculation flips as claim value rises. At some threshold, the cost of getting it wrong exceeds the cost of a person looking properly, and every insurer draws that line somewhere. Automation has pushed the line upward, because photographic assessment and automated triage made checking cheaper, so more claims now fall below it.

But the line still exists, and above it the work is not merely retained, it is intensified. The claims that reach a person are now the ones already filtered as complex, disputed or suspicious. An adjuster’s day contains a higher concentration of difficult files than it did five years ago, with fewer straightforward ones between them.

That has two consequences worth naming. The role is more demanding than it was, and the traditional way of building judgement, by handling hundreds of easy claims before meeting a hard one, has largely disappeared.

What to Know Before You Draw Conclusions

The decline is concentrated in volume roles. Motor and simple property adjusting is where the 21,800 positions come from. Specialist and investigative roles are not the source of the decline.

Regulation constrains automated decisions. Claims handling is regulated, with obligations around fair treatment, explanation and complaint handling. An automated declinature still needs an explainable basis and an appeal route with a person at the end of it.

Fraud detection creates work rather than removing it. Better flagging produces more referrals, and every referral needs investigation by someone who can interview and assess evidence.

Exposure measures are not employment forecasts. BLS published AI exposure categories with the 2025-35 projections and states plainly that exposure “does not imply job loss, productivity gains, automation probability, or wage effects.” Claims work scores high, and the projected decline of 6 percent is meaningful but far from elimination.

Catastrophe response still needs bodies. After a major weather event, insurers deploy large numbers of adjusters physically. That surge capacity has no automated substitute.

One more factor is worth weighing before drawing conclusions about your own position. Claims volumes are not fixed. Weather-related property claims have been rising in many markets, and more frequent severe events increase both the number of claims and their complexity, because widespread damage produces contractor shortages, disputed scopes of work and long settlement periods. That trend works against the automation-driven decline, and it is the main reason the projection is a modest 6 percent rather than something larger.

Where Claims Careers Are Growing

• Complex property and casualty. Causation, coverage disputes and large losses.
• Bodily injury and liability. Medical evidence, quantum assessment and negotiation.
• Fraud investigation. Growing referral volume, interviewing and evidence work.
• Catastrophe response. Surge deployment, field assessment, coordination.
• Claims governance and quality. Auditing automated decisions, handling complaints, and evidencing regulatory compliance, which is a new function in most insurers and currently understaffed in nearly all of them.

A Decision Framework for Adjusters

1. Handling high volumes of simple claims. The most exposed position, and the source of the projected decline. Move toward complex property, injury or investigation within the next two years.
2. Field adjuster. Comparatively secure because physical inspection persists, particularly after catastrophe events. Adding causation expertise strengthens this considerably.
3. Injury or liability specialist. Strong position. Negotiation, medical evidence and quantum are not close to automated.
4. Entering the profession. The volume work that used to teach the basics is disappearing. Target insurers or loss adjusting firms with genuine complex books, and expect to learn on harder files sooner.

The test across all four: what proportion of your files involve someone disagreeing with someone else? That proportion is your security, and it is worth raising deliberately.

Common mistakes right now

• Accepting an automated estimate on a claim where causation is ambiguous.
• Treating a fraud flag as a finding rather than as a reason to investigate.
• Neglecting interview technique, which is the least automatable skill in claims work.
• Assuming volume experience will still be available to the people you are supposed to train.

Building the Fluency Claims Governance Now Requires

Insurers running automated settlement need people who understand both the claims discipline and the systems making the decisions. Someone has to audit whether automated assessments are fair, spot where a model systematically underestimates a particular kind of damage, and answer a regulator asking how a declinature was reached.

That is claims expertise applied to a new object, and it is a growing function in most insurers. It requires understanding how these systems produce output, why they are confident about common cases and unreliable about unusual ones, and how to document a review so it stands up. Learning that in a structured sequence is faster than absorbing it from vendor material, and a certificate alongside your claims qualifications makes it visible when a firm is choosing who leads oversight. If you want a structured route in, explore Coursiv AI lessons and check current plan details on the official site.

FAQ

Your Next Step

Sort last quarter’s files into those where every party agreed on what happened and those where they did not. The first pile is the one being automated, and the second is the profession. If your work sits mostly in the first pile, ask your manager for referrals from the second, because that is where the remaining 368,000 claims roles in 2035 are going to be.