Anyone who has worked inside insurance technology knows that systems running the core business rarely get replaced quickly. Across the industry, IBM i (AS400) environments continue to power many mission-critical insurance platforms, supporting policy administration, claims processing, billing, and commission calculations.
What has changed is the complexity of the claims workflow around those systems. A typical auto or property claim can involve dozens of steps including intake, document review, validation, routing, fraud checks, settlement evaluation, and customer communication. In many insurers, a large portion of these steps still require manual intervention.
Industry automation demonstrations show that more than 70 tasks can exist within a single claims workflow. Many of these steps happen before the core system even processes the claim.
Because of this, the core system is rarely the real bottleneck.
The challenge lies in the manual work surrounding it. Adjusters and operations teams still spend significant time reviewing documents, extracting information, validating coverage, and re-entering data into core applications. These steps slow down the claims lifecycle and introduce unnecessary operational friction.
Modern claims transformation is therefore taking a different direction. Instead of replacing stable core platforms, insurers are extending them with intelligence layers that automate document interpretation, surface risk signals, and assist decision-making.
Artificial intelligence plays a key role in this shift, not as a replacement for IBM i systems, but as a way to make the claims workflows around them faster and easier to manage.
Where Challenges Usually Show Up in Insurance Environments?
A claim does not arrive in a clean, ready-to-process format. It usually comes with a mix of forms, emails, notes, estimates, reports, photos, and other supporting documents. Someone has to go through all of that, figure out what matters, and get the right information into the system.
That is where most of the operational effort sits.
1. Too much claim information arrives in documents
A large part of claims work starts outside the actual system.
Teams receive:
- repair estimates
- police reports
- medical bills
- adjuster notes
- claim forms
- email attachments
- photos and scanned files
The problem is simple. The IBM i system can process structured claim data, but it cannot work with a stack of documents until someone reads them first.
So the team has to:
- open each document
- find the important details
- check whether the information is complete
- enter the required values into the claims system
This takes time, especially when one claim includes multiple documents or when the information is kept inside long reports.
2. Claim registration gets delayed at the intake stage
For many teams, the delay begins even before the claim is created. If the submission is incomplete or spread across multiple files, intake staff have to stop and piece it together manually. They may need to check the loss date, confirm the policy number, identify the accident location, or understand what actually happened before the claim can be registered.
For IBM i users, this is a familiar issue. The system is ready to process the claim, but the claim cannot move forward until someone prepares the data in the right format. And this can create a bottleneck at the front of the workflow.
3. Manual entry creates avoidable rework
Once the right details are identified, they still have to be entered correctly. That is where another problem appears.
Policy numbers, claim amounts, treatment codes, dates, and coverage details must match what is already in the system. If something is typed incorrectly or interpreted the wrong way, the claim may fail validation or need correction later.
This means:
- going back into the claim
- checking documents again
- correcting the data
- resubmitting or reprocessing the claim
Most of these are not a system failure. It is the kind of rework that builds up when too much of the process depends on manual preparation.
4. Important claim signals are easy to miss early
Another challenge is that some of the most important signals in a claim are not obvious when someone is just reading documents one by one.
A reviewer may not immediately notice:
- repeated addresses across claims
- unusual claim frequency
- patterns linked to certain providers
- details that suggest third party liability
- indicators that a claim may need deeper review
IBM i systems are very good at storing and processing claim data, but spotting patterns across many claims usually requires more analysis than a person can do during intake or first review. As a result, some claims move too far into the workflow before risks or recovery opportunities are noticed.
5. Experienced staff carry too much of the process in their heads
This is another issue many long-running insurance environments deal with. A lot of claims work depends on experienced people knowing:
- where to look in the documents
- what usually causes validation issues
- which claims need special attention
- how to judge whether something looks incomplete or unusual
That knowledge is valuable, but it also creates dependency.
When experienced adjusters or IBM i users are unavailable, newer team members may take longer to process claims because they are still learning how to interpret documents, navigate exceptions, and prepare data correctly.
Seeing Similar Friction in Your IBM i Claims Workflow?
Seeing Similar Friction in Your IBM i Claims Workflow?
What AI Actually Adds to an IBM i Claims Environment
Here are 8 ways AI can reduce friction, speed up claims, and support better decisions in IBM i environments.
1. FNOL Automation (First Notice of Loss)
The claims lifecycle almost always begins with the First Notice of Loss (FNOL). This is the moment when a policyholder reports an accident, injury, or loss event to the insurer.
In many organizations, this stage still involves manual data entry, phone calls, and email exchanges before a claim is even created in the system. Intake teams often review submissions, gather missing information, and manually register the claim in the claims application. These delays create friction right at the beginning of the claims process.
AI-enabled intake systems change how this step works.
Incoming forms, emails, or uploaded documents can be analyzed automatically to extract key claim attributes such as:
- policy identifiers
- loss dates
- accident locations
- claim descriptions
That information can then populate the claim creation screen in the IBM i claims system automatically.
Claims get registered faster, and fewer follow-up interactions are needed to collect missing details. In several implementations, insurers have reported 40 to 60 percent faster claim creation once automated intake processes were introduced.
2. Intelligent Workflow Orchestration
It is important to understand that claims automation is not driven by a single AI model.
Modern claims systems rely on workflow orchestration that coordinates many different tasks across the claims lifecycle.
The workflow becomes the control layer that decides what happens next in the process. Tasks such as intake validation, document analysis, fraud scoring, and settlement preparation are triggered automatically as the claim progresses.
In many real claims environments, more than 70 individual workflow tasks can exist across the lifecycle of a single claim. Automation platforms coordinate those steps and determine which tasks can run automatically and which require human review.
Straightforward claims move through the workflow quickly. Complex cases pause at specific steps where adjusters need to review information or make decisions.
This structure makes it possible to automate large portions of the process while still keeping human expertise involved where judgment is required.
3. Document Processing
Claims processing revolves around documents. Police reports, medical bills, repair estimates, adjuster notes, invoices, and photos all contain important information.
The challenge is that most of this information arrives in unstructured formats such as PDFs, scanned documents, email attachments, or images.
In many insurance organizations, adjusters still read these documents manually and type the information into the claims system.
Modern document processing systems combine optical character recognition (OCR), natural language processing (NLP), and machine learning models to interpret information from these documents.
Instead of an adjuster reading every document first, the AI system analyzes the file and identifies important fields such as:
- medical codes from treatment documents
- repair line items and costs from body shop estimates
- accident details from police reports
- policy identifiers from claim forms
The extracted data can then be reviewed and posted directly into the same AS/400 claim screens adjusters already use.
The IBM i system still manages the claim record. The difference is that adjusters spend less time typing information and more time reviewing the claim itself.
4. Claims Triage
Another area where AI improves claims operations is claim routing and prioritization.
Many organizations still route claims using basic rules. A claim might be assigned based on claim type, region, or a manually assigned priority tag.
But not all claims require the same level of attention.
Some are straightforward and can be processed quickly. Others involve injuries, liability disputes, or potential fraud indicators.
Machine learning models can analyze incoming claims and classify them based on signals such as:
- policy coverage characteristics
- estimated loss amount
- injury indicators
- claim history patterns
- geographic risk factors
These models can evaluate the information in real time and predict the likely complexity of the claim before it enters the main adjustment workflow.
The claim then receives a complexity score.
Simple claims move through faster queues, while more complicated cases are routed to experienced adjusters.
5. Fraud Detection
Fraud detection is another area where AI supports claims teams.
Fraud rarely appears as a single obvious indicator. More often it emerges through patterns across multiple claims.
- repeated addresses across claims
- unusually frequent claims from the same policyholder
- suspicious provider relationships
These signals may be buried across thousands of records in claims and policy databases.
Machine learning models can analyze historical claims data and identify patterns that suggest suspicious activity. External datasets such as provider registries or vehicle records can also be incorporated into the analysis.
When the system detects a potential anomaly, the claim receives a fraud risk score.
That score is then passed into the claims workflow so investigators can review the case.
AI handles the pattern detection. Investigators still make the final determination.
6. Decision Support for Adjusters
Claims decisions often depend on comparing a current case with similar past claims.
Adjusters frequently need to know:
- how similar claims were settled
- what settlement ranges were typical
- which documents were required in previous cases
AI systems can analyze historical claims data and surface insights that help adjusters evaluate the current claim.
For example, the system might identify that the current claim resembles dozens of past cases with similar characteristics. It can then show typical settlement ranges or highlight factors that influenced previous outcomes.
The adjuster still reviews the case and makes the final decision. The AI simply provides context based on historical data.
For this reason, most experts describe AI in claims processing as a decision support layer, not an automated decision engine.
7. Customer Interaction Automation
Automation is also improving how insurers communicate with policyholders during the claims process.
AI-powered assistants can guide customers through claim submission using web interfaces, mobile apps, or chat interfaces. These systems gather information about the incident, request photos or documentation, and answer common questions during the claim intake process.
Because the information flows directly into the claims workflow, policyholders receive faster acknowledgements and clearer updates about their claim.
This also reduces the volume of routine calls handled by customer service teams.
8. Subrogation Opportunity Detection
Another emerging use case involves identifying opportunities for subrogation and recovery.
Subrogation opportunities appear when a third party may be responsible for the loss. These cases are sometimes missed because adjusters focus primarily on settling the claim.
AI systems can analyze claim narratives, repair documentation, and liability statements to detect patterns that suggest third-party responsibility.
When those patterns appear, the claim can be flagged for subrogation review.
Claims teams then investigate the opportunity without adding additional manual workload to the normal claims process.
How AI Is Typically Applied Around an IBM i Claims Environment
In most cases, AI is not embedded deep inside the AS/400 core. It is applied around the existing system through the integration and automation layers that already support claims intake, document handling, workflow movement, and external communication.
That usually happens through a mix of methods:
- RPA for posting claim values into green screens and handling repetitive screen-based tasks
- APIs for reading and writing claim data in real time where direct integration is available
- ETL or ODI pipelines for moving claims data securely into AI, analytics, or fraud models
- Web services for connecting intake portals, FNOL applications, and customer-facing workflows back to the core claims process
- Message queues for decoupling heavy workloads and moving claim events more reliably across systems
In a nutshell, the IBM i platform continues to handle the core claims logic, while AI works through the layers around it to improve intake, document understanding, routing, fraud review, and decision support.
A Scenario We Recently Came Across
In one claims environment we reviewed, the IBM i system itself was working exactly as expected. Policy validation worked correctly, transactions were processed reliably, and financial records remained fully auditable.
The slowdown was happening before the system ever processed the claim.
When a new auto claim arrived, the intake team had to review the submission manually before creating the claim record. Supporting documents often included photos, repair estimates, police reports, and emails from the policyholder. Someone had to read through these materials, identify the required details, and enter that information into the claims system.
Once the claim entered the system, it still had to be routed to the correct adjuster, checked for coverage, reviewed for potential fraud indicators, and communicated back to the policyholder.
None of these steps were unusual. But when repeated across hundreds of claims every week, the manual effort added significant time to the claims lifecycle.
The IBM i platform was not the problem.
The manual workflow around it was.
What We Changed
Instead of replacing the IBM i claims system, we focused on improving the workflow around it.
The core system still manages policy validation, financial records, and claim transactions. The improvements happened before and around that system.
Here is what changed.
Automated FNOL Intake
- Customers submit claims through a digital form or mobile interface
- AI extracts key details such as policy number, loss date, and location
- The claim record is created automatically in the IBM i system
Document Processing
- OCR and NLP read repair estimates, police reports, and invoices
- Important information is extracted automatically
- Adjusters review extracted data instead of manually entering it
Claim Triage
- AI evaluates incoming claims for complexity and risk
- Simple claims move faster through the system
- Complex cases are routed to experienced adjusters
Fraud Detection
- Machine learning models analyze claim patterns
- Suspicious claims are flagged early for investigation
Decision Support
- AI surfaces insights from historical claims
- Adjusters get context before making settlement decisions
What Happened After the Changes
Once automated intake, document processing, and workflow orchestration were introduced, the impact became visible fairly quickly. Within roughly 120 days, the organization began seeing measurable improvements across the claims workflow.
Some of the most noticeable changes included:
- Intake cycle time improved significantly. Claims that previously took about 2.8 days to register were now entering the system in around 7 hours on average.
- Manual document indexing dropped sharply. With AI extracting information from submitted documents, the workload associated with manual indexing fell by about 80 percent.
- Subrogation opportunities increased. Earlier analysis of claim details helped identify recovery opportunities that were previously missed, resulting in a 2.4× increase in subrogation recovery cases.
- Fraud detection moved earlier in the workflow. Instead of investigating fraud only after a claim progressed through the process, the system introduced proactive fraud scoring during intake.
Explore AS400 Application Modernization Strategies and get a clear view of the main modernization paths, what each one solves, and how teams can evaluate the right approach based on risk, cost, and long-term business goals.
What AI Is Already Changing in Insurance [The Evidence]
- Insurers leading in AI adoption have generated 6.1 times higher total shareholder returns than AI laggards over the past five years. (Source: McKinsey & Company ).
- AI-driven transformations in insurance domains such as sales and distribution have produced 10 to 20 percent improvements in agent success rates and conversion rates.
- Organizations applying AI across underwriting and distribution have reported 10 to 15 percent increases in premium growth.
- Intelligent automation and AI-assisted workflows have helped insurers reduce customer onboarding costs by 20 to 40 percent.
- AI-supported decision systems are improving operational quality, with 3 to 5 percent improvements in claims accuracy.
- Real-world claims transformation is already visible. Aviva deployed more than 80 AI models in its claims operations, reducing liability assessment time for complex cases by 23 days, improving claims routing accuracy by 30 percent, and lowering customer complaints by 65 percent.
Read AI for AS/400 (IBM i): Where It Fits and What to Watch Out For to understand where AI is actually working in IBM i environments, where teams are seeing value first, and what needs careful planning before adoption.
Before We Bid Adieu
But there is also friction in adoption, especially when legacy platforms carry years of business logic, operational dependency, and security considerations. That is why the first step is not to apply AI everywhere. It is to map your challenges clearly, understand where AI can genuinely help, and define how far it should go.
If you want to explore what that could look like in your own environment, talk to our AS/400 experts. We can help you assess your current claims workflow, identify where AI can create measurable impact, and map out a practical modernization path around the systems you already rely on.
And that is where we can help you.
With 15+ years of experience, Nalashaa brings AS/400 modernization expertise and insurance domain knowledge across claims, FNOL, policy, and billing. Our experts help you understand where AI fits, where it can create real value, and how to apply it around your existing environment.
Book a strategy call to assess your automation potential and identify where AI can make an immediate impact.
