Why the Claim File Is Becoming Harder to Manage, Even as Claims Systems Get Better

The industry has made real progress storing claim information. The unsolved problem is maintaining context across all of it.

Claims technology has improved considerably.

Modern claims management systems can centralize notes, automate diaries, route assignments, store documents, track payments, connect with vendors, and give supervisors much better visibility than the claims platforms of the past.

Yet talk to claims professionals and a strange contradiction appears.

The software is getting better. The claim itself is getting harder to understand.

An adjuster can have more information available than ever and still spend twenty minutes figuring out what happened over the last two weeks.

That is because the modern claim file is no longer contained entirely inside the claims system.

The system may remain the official record, but the work surrounding a claim now stretches across email, medical portals, legal teams, vendor platforms, employer systems, documents, phone calls, messaging tools, and specialized services.

The claim file used to be easier to define

Historically, the claim file had fairly obvious boundaries.

There were notes. Correspondence. Medical reports. Payment records. Investigation. Legal documents. Perhaps a physical folder and eventually an electronic version of the same material.

Today, an adjuster may manage one claim through:

• The primary claims management system
• Microsoft Outlook
• A medical bill review platform
• Utilization review
• Pharmacy systems
• Defense counsel
• Nurse case management
• Provider portals
• Imaging portals
• Investigation vendors
• Return-to-work vendors
• Employer communication
• Document repositories
• Internal spreadsheets
• Teams or Slack
• State regulatory systems

Each tool may perform its job well.

The difficulty emerges between them.

Better claims systems create better systems of record

The modern claims platform is designed to create structure.

Important fields can be standardized. Diaries can be assigned. Payments can be tracked. Documents can be categorized. Notes can be searched. Supervisors can see caseloads and financial information across the organization.

That infrastructure is extremely valuable.

But a system of record can only organize what reaches it.

Consider a routine medical development.

A physician sends a report changing the injured worker’s restrictions.

The report arrives by email.

The adjuster downloads it.

The document is uploaded to the claim.

The work status is updated.

The employer receives an email asking whether modified duty is available.

The employer responds that it needs to speak with the supervisor.

A nurse case manager separately emails that the next appointment has moved.

Defense counsel sends another message asking whether the work status affects an upcoming deposition.

The claims system may eventually contain all of those events.

The actual situation exists across several conversations while it is unfolding.

The inbox has become part of the claim infrastructure

Email remains one of the most important operating systems in insurance.

Medical reports arrive there. Attorneys communicate there. Employers ask questions there. Vendors send updates there. Injured workers may submit documents there.

This creates a gap between receiving information and incorporating it into the official file.

A claims professional has to decide:

• Which claim does this belong to?
• Does the attachment matter?
• Does a field need to be updated?
• Should a note be entered?
• Does someone need a response?
• Should a diary be created?
• Does this change the current strategy?
• Does another stakeholder need to know?

The claims system cannot manage information it has not yet received.

That makes the adjuster the integration layer.

More specialized vendors create more handoffs

Insurance has become increasingly specialized.

That can improve the quality of individual services. A carrier or TPA can use dedicated experts for bill review, medical management, investigations, legal defense, pharmacy, utilization review, and other parts of the claim.

The tradeoff is coordination.

Every additional participant creates another place where information can originate.

A nurse case manager may know that the worker is struggling to schedule a specialist.

Defense counsel may know that a deposition changed the litigation posture.

The employer may know that modified work will become available next Monday.

The medical report may document a new restriction.

The adjuster needs all four pieces to understand the claim.

They may arrive through four different channels.

The challenge is no longer a lack of information. It is bringing the information together quickly enough to act on it.

Documents have become easier to store and harder to absorb

Digital claims eliminated many of the physical constraints of paper files.

They also made it possible for a claim to accumulate an enormous amount of information.

A long-running workers’ compensation claim may contain:

• Thousands of pages of medical records
• Repeated medical productions
• Diagnostic imaging reports
• Legal correspondence
• Deposition transcripts
• Claim notes
• Wage documentation
• Benefit notices
• Investigation reports
• Bills
• Utilization decisions
• Emails
• Settlement documents

Storage is cheap.

Human attention is not.

An adjuster does not need to know everything in every document. They need to know which information changes the current understanding of the claim.

That is a fundamentally different information problem.

Recent work examining large language models in claims environments highlights the same divide. A 2026 Casualty Actuarial Society paper puts it directly: actuaries rely primarily on structured numerical data, while valuable predictive information in unstructured text, including medical records, adjuster notes, and call transcripts, remains largely unused.

Structured data captures status better than reasoning

Claims systems are good at answering questions such as:

• What is the current reserve?
• Who is assigned?
• Is the claimant represented?
• What benefits have been paid?
• What is the next diary date?
• What documents are in the file?

Harder questions often matter more:

• Why did the reserve increase?
• What changed medically?
• Why has return to work stalled?
• What is defense counsel waiting for?
• Which issue is preventing settlement?
• What assumption was the prior adjuster making?
• Which missing record could change the claim strategy?

Some of that context appears in notes.

Some lives in correspondence.

Some exists only because the adjuster remembers the case.

The difference between status and reasoning becomes more important as claims grow longer and more complicated.

The same fact can exist in several versions

Fragmentation also creates inconsistency.

Imagine an injured worker’s work status changes on Monday.

The physician’s report says modified duty.

A claim note still reflects temporary total disability.

The employer spreadsheet shows “off work.”

Defense counsel’s litigation summary was prepared the previous week.

The return-to-work vendor has not yet received the new restrictions.

There is no single dramatic error.

The organization has several versions of the same claim at different points in time.

Someone has to reconcile them.

This is why data integration alone does not solve every claims problem. Two systems can technically exchange data while still operating on different timing, definitions, and workflows.

Claims become harder to understand as they age

A new claim may be relatively simple.

There is an injury, an initial report, some medical treatment, and a small number of open questions.

Twelve months later, the same file may include:

• Multiple treating providers
• Several work status changes
• Conflicting medical opinions
• Reserve adjustments
• Attorney involvement
• Missed appointments
• New diagnostic findings
• Prior injury records
• Supplemental reports
• Settlement discussions
• Changes in adjuster ownership

Each new development depends on what came before it.

The adjuster increasingly needs the history in order to understand the present.

This is why an old claim can take longer to review even when very little is happening today.

The difficulty lies in reconstructing how the case reached its current position.

Turnover exposes the context problem

The clearest test of a claim file is what happens when another adjuster inherits it.

If the file is truly understandable, the new adjuster should be able to determine:

• What happened
• Where the claim stands
• What changed recently
• What major decisions were made
• Why those decisions were made
• What remains unresolved
• What needs to happen next

In practice, experienced adjusters often have to reconstruct the story from hundreds of notes and documents.

This is the claim handoff problem at a broader level.

The original adjuster may have understood the file perfectly. The organization did not.

More technology can actually increase fragmentation

Insurers frequently respond to operational problems by adding another tool.

A new product may solve the targeted problem extremely well.

The organization now has one more application.

Without deliberate integration, new technology can create:

• Another login
• Another inbox
• Another data format
• Another notification stream
• Another vendor relationship
• Another place an adjuster has to check

This helps explain why modernization can feel slower than expected even when individual tools improve.

Deloitte’s 2026 Global Insurance Outlook makes the same observation, noting that many insurers struggle with fragmented, messy data sprawl and outdated systems, and that legacy system modernization remains a top focus area.

The problem is architectural as much as technological.

Claims organizations need individual systems to work well, and they need information to remain coherent as it moves among them.

AI creates an opportunity because it can work with unstructured information

Traditional claims technology works best when information is already structured.

A reserve is a number. A claim number is an identifier. A diary has a date. A payment has an amount.

Much of claims handling is language.

A medical report explains what changed clinically.

An attorney email describes litigation strategy.

A claim note explains why an action was taken.

An employer response provides context around return to work.

This is where newer AI systems have potential.

They can help extract useful information from documents, correspondence, and notes that historically required a person to read and interpret them one at a time.

That same research, Leveraging LLMs for Unstructured Claims Data Analysis, built a two-stage pipeline that extracts document-level detail and then synthesizes it into claim-level variables, covering reserving, ratemaking, and claims management. It is a proof of concept rather than a production system, but the direction is the relevant part.

The commercial opportunity extends beyond summarization.

A useful claims intelligence layer could help answer:

• What changed since the last review?
• Which new document affected the claim?
• What actions remain open?
• Are two systems showing conflicting information?
• Which claims have no meaningful activity?
• What decisions are approaching?
• What source supports this claim note or summary?
• What does the adjuster need to look at first?

These are context problems.

AI can also make the problem worse

Adding AI does not automatically create a coherent claim file.

An unreliable summary can become another version of the claim that someone has to verify.

A system may:

• Miss a significant medical development
• Merge two separate injury periods
• Present an old restriction as current
• Treat an attorney’s allegation as an established fact
• Lose the source behind a statement
• Produce a polished narrative that hides uncertainty

Claims teams therefore need more than fluent summaries.

They need source traceability, clear distinctions between documented facts and generated conclusions, appropriate human review, and a way to correct the system when the claim changes.

The same Deloitte outlook recommends that insurers prioritize data quality, integration, and master data management rather than treating AI as a layer that can be added on top. It also makes a practical point worth keeping: perfect data hygiene may not be essential for every AI project, but standardization and control are critical to avoid conflicting results and maintain trust.

AI becomes valuable when it reduces fragmentation rather than adding another layer to it.

The next generation of claims systems may be judged by context

The traditional claims system answers:

What is in the file?

The next generation will increasingly need to answer:

What does the file currently mean?

That requires several capabilities working together:

A current timeline

The system should be able to show the major events without forcing the adjuster to reconstruct them manually.

Source-linked information

Important statements should remain connected to the medical report, note, correspondence, or record that supports them.

Changes over time

The adjuster should be able to see what is different from the last review.

Open issues

Missing records, unanswered communications, pending medical decisions, and unresolved inconsistencies should be visible.

Clear ownership

The next action should have a responsible person and expected timing.

Cross-system awareness

Relevant information should not disappear merely because it arrived through email or lives in another application.

Human judgment

The system should help the claims professional understand the case without quietly making consequential decisions on their behalf.

Claims management is becoming an information coordination problem

Modern claims systems have solved many problems that used to consume adjuster time.

The difficult part is shifting.

Claims professionals increasingly work in an environment where the organization possesses almost every piece of information it needs, yet those pieces are distributed across too many locations and arrive at different times.

The result is a claim file that is technically complete while remaining difficult to understand.

Better core systems will help.

Better integrations will help.

AI will help in places where large volumes of unstructured information have historically required manual review.

The larger objective is continuity.

A claims professional should be able to understand what happened, what changed, why it matters, and what needs to happen next without reconstructing the case from the beginning.

That is becoming the standard by which claims technology should ultimately be measured.

---

This article provides general information about insurance claims operations and technology. It does not constitute legal, regulatory, or insurance advice.