Context Engine: Traversing the Insurance Landscape

[Biraj_Bhushan_C]

In property and casualty insurance, the combined ratio is one of the metrics that matters most. For a carrier writing $10 billion in premium, each point of combined ratio represents $100 million in underwriting result. Three points mark the difference between a year that can be explained to shareholders and one that must be excused. Because the ratio refuses to accommodate narratives; it cares only whether claims and expenses consumed more than premium collected. 

That mathematical discipline once felt manageable because volatility could be bounded and diversified. The past several years have rendered that assumption obsolete. Global insured losses from natural catastrophes reached $137 billion in 2024, with Hurricanes Helene and Milton, severe convective storms and major flooding driving the accumulation.  This marked the fifth consecutive year that insured losses exceeded $100 billion. Following the long-term annual growth trend of 5-7 percent in real terms, projects losses could approach $145 billion in 2025.  

Catastrophe exposure alone does not explain the full picture. Liability severity is being reshaped by social inflation, which the NAIC describes as liability claims costs rising faster than general economic inflation, driven in part by increasing litigation costs and shifting social attitudes about who should absorb risk.  Further, Swiss Re's Social Inflation Index shows US liability claims increased by 57 percent over the past decade, with annual growth averaging 5.4 percent during 2017-2022 compared to 3.7 percent for economic inflation. Increasing inflation also drove liability insurance losses by more than $230 billion over the decade ending in 2024, according to a joint analysis by Triple-I and the Casualty Actuarial Society. 

The litigation environment has intensified correspondingly. In 2024, juries awarded 135 nuclear verdicts (verdicts exceeding $10 million) against corporate defendants, a 52% increase from 2023. The combined value of these verdicts reached $31.3 billion, up 116% year over year. The trend is unmistakable: between 2020 and 2024, the number of nuclear verdicts more than quadrupled, while the median verdict size more than doubled . 

These are structural shifts. They demand operational discipline that can hold when the environment changes faster than teams can update habits. But the issue is not just more risk; it is whether underwriting and claims can scale under pressure. 

Why underwriting discipline breaks down 

Commercial underwriting is evidence-heavy judgment exercised under uncertainty. That’s because risk does not arrive as clean data. It arrives as a submission packet: ACORD forms, schedules, loss runs, broker narratives, engineering reports, emails, contracts, endorsements and supporting attachments. Underwriters translate that packet into appetite fit, pricing position, coverage terms, and a decision record that can withstand later scrutiny  all under time pressure.  Highlighting this, Capgemini's World Property and Casualty Insurance Report found that 41 percent of underwriters' time is consumed by administrative and operational activity. 

Most carriers already run modern data warehouse and lakehouse platforms that provide portfolio-scale governance: territory mappings, class code catalogs, guideline versions, endorsement libraries, catastrophe and accumulation analytics, reinsurance guardrails, pricing indications, and performance benchmarks. These platforms help leaders see the book and steer appetite. Yet the file-level decision still turns on the submission packet, because the most consequential risks live inside documents: the real mix of work, the contractual obligations accepted upstream, and the operational reality of risk transfer downstream. 

Content intelligence in the Hyland Enterprise Context Engine™ is designed for that translation layer. It sits above the data platform, draws on it for canonical reference data and active rule context, and converts the submission packet into structured, sourceable underwriting facts. Every assertion is tied to a specific document, page, and excerpt. The file becomes legible for the underwriter and the evaluation aligns to the same rulebook and reference tables used across the enterprise. 

The five evidence checks behind disciplined GC underwriting 

Strong general contractor (GC) underwriting is a useful lens because the loss drivers repeat and the evidence is scattered. The strongest teams run the same five evidence checks on every file, because those checks force the underwriter to price what is real rather than what is implied. This is where the Context Engine adds leverage by extracting key facts from the packet, reconciling conflicts across documents, and anchoring each conclusion to the clause, page, and excerpt that supports it.    

Now let’s look at how general contractor underwriting offers a clear view into how disciplined insurance decisions are made: 

1. Exposure reality: The first task is to prove the shape of the work, since this is where severity begins. Habitational share, maximum stories, largest project value, subcontracting intensity, and geographic concentration carry more predictive power than any polished narrative.  
   The Context Engine extracts these facts from the project schedule, GC supplemental, and revenue splits, then reconciles them across documents. If the schedule indicates 35 percent multifamily habitational while the ACORD description implies only light commercial, the mismatch becomes a visible underwriting issue with a precise follow-up. Locations are normalized against the carrier's territory mappings from the data platform, so the same submission can be evaluated against coastal concentration limits, venue risk, and catastrophe accumulation. 
   
   
2. Upstream contract obligations: In construction, obligations drive tail exposure as much as operations. Prime contracts and insurance exhibits reveal indemnity scope, defense obligations, additional insured requirements, primary and noncontributory language, and completed operations duration expectations.  
   The Context Engine reads these provisions, ties each to the clause that created it, and maps extracted obligations to the carrier's endorsement library and form constraints held in the data platform. Underwriters who run this check early design programs that fit the real contractual environment. Those who skip it tend to discover the exposure at the worst possible moment, when defense costs on an additional insured dispute turn an ordinary claim into an expensive one. 
   
   
3. Downstream risk transfer execution: General contractors routinely state that they require certificates and endorsements from subcontractors. Claims reveal whether those requirements were executed. The critical distinction is between certificate collection and endorsement proof, especially for completed operations additional insured coverage and waivers.  
   The Context Engine converts subcontract templates, COI tracking exports, and endorsement verification logs into measurable compliance indicators: gaps by subcontractor, weak points by project type. Once risk transfer execution becomes measurable, the lakehouse can treat it as an input to umbrella design and expected loss cost, predicting tail leakage with more fidelity than broad industry labels. 
   
   
4. Loss story: Loss runs are often treated as a list. Underwriting needs drivers.   
   The Context Engine restructures loss runs and large loss narratives into a loss story that separates frequency from severity, identifies open claims carrying uncertainty, and surfaces early litigation signals. For construction, completed operations losses deserve disproportionate attention, since one event can foreshadow years of adverse development. The data platform supplies benchmarking by segment and geography, and internal claims history when it exists. Together these layers let underwriting align pricing and program structure to a grounded view of loss volatility. 
   
   
5. Jobsite control: General contracting is supervision repeated across sites. Evidence lives in safety programs, supervisor ratios, subcontractor prequalification criteria, incident workflows, and loss control reports with remediation proof.  
   The Context Engine extracts control signals and ties them to proof artifacts. The warehouse supplies the carrier's control expectations for the segment, so bind requirements remain consistent across underwriters and MGAs. A GC with documented safety investment, trained supervisors per active site, and a prequalification process that includes financial thresholds earns confidence and flexibility. A GC missing those signals requires tighter terms, higher retention, or narrower authority, and the decision record should say so explicitly. 

When evidence is captured this way, it stops being a one-time review and becomes reusable enterprise memory. Warehouse and lakehouse platforms govern reference data, rule versions, rating structures, and portfolio constraints. The Enterprise Context Engine makes the submission packet legible, maps it to the carrier’s rules, and delivers structured facts with provenance. Underwriting gains a single dossier: economics and constraints at the portfolio level, and evidence, obligations, and execution quality at the file level. 

Decision Intelligence completes the underwriting layer by capturing the bind moment as a structured record. Quote versions, referrals, exceptions, overrides, and subjectivities become traceable events linked to the specific evidence reviewed and the guideline snapshot in force at the time. Renewals start from an accurate prior rationale. Audit questions can be answered without reconstruction. And the data platform can learn from outcomes because the decision and its supporting proof are stored in a form that analytics and models can consume. 

Why claims decisions are harder to defend at scale 

Claims is where the promise of underwriting meets the reality of a loss. A typical liability claim file (policy, endorsements, notice of loss, adjuster notes, medical records, expert reports, litigation filings, recovery documentation) fragments across systems in ways that make coherent evaluation difficult at scale.  

Carriers run strong analytics platforms for reserve monitoring, fraud scoring, and subrogation tracking. But the file itself is where the costly work concentrates. Disputes turn on specific clauses. Reserve volatility turns on what documentation shows about injury and causation. Recoveries depend on whether evidence can be assembled fast enough to change an outcome. 

The Context Engine brings file-level structure to claims by mapping facts to policy language, endorsements, limits, deductibles, and key conditions. This drives faster coverage analysis, stronger fraud or intervention timing, and better audit and regulatory defensibility. 

The Context Engine also supports the evaluations experienced claims teams already perform: coverage basis, liability and causation, damages and reserve drivers, fraud and inconsistency signals, and recovery opportunities. Decision intelligence then logs each coverage determination, reserve change, settlement authority move, special investigations unit (SIU) referral, and denial rationale against the governing policy version, creating a decision trail that auditors and regulators can follow without reconstruction. 

NAIC annual statement data shows US insurers recovered nearly $51.6 billion across major auto lines in 2021, while industry analysts estimate missed subrogation opportunities cost the industry approximately $15 billion annually. Insurance fraud costs the US economy an estimated $308.6 billion annually, with property and casualty fraud contributing roughly $45 to $90 billion, according to Coalition Against Insurance Fraud.  

However, Deloitte predicts AI-driven technologies could reduce fraudulent claims and save between $80 billion and $160 billion by 2032.  Capturing this value requires a file structured enough to act on early, and a decision trail that can withstand the oversight now emerging from the NAIC's model bulletin on insurer use of AI systems and state-level reforms such as Georgia's litigation funding transparency legislation enacted in April 2025. 

The same gap is driving both issues and it is time to bridge it 

Traversing the insurance landscape reveals that underwriting and claims are two manifestations of the same enterprise gap. Insurance is policy logic applied to messy evidence under time pressure. Without a unified evidence layer and a durable decision layer, carriers struggle to scale disciplined judgment across teams, partners, and cycles. 

The structural environment makes this capability urgent. There is a one-in-ten probability that global insured losses could reach $300 billion in 2025, representing what would be the next peak loss year, according to estimates by Swiss Re. The reinsurance market is well positioned to absorb such scenarios. Global traditional reinsurance capital is currently estimated at around $500 billion, per Swiss Re, with alternative capital including approximately $50 billion from the catastrophe bond market contributing additional capacity.  

But capacity alone does not create discipline. In an era where catastrophe losses remain structurally elevated and liability severity is being reshaped by litigation dynamics, the carriers that outperform will be those that can operationalize judgment as a managed enterprise asset: evidenced, explainable, and improvable rather than dependent on individual memory and post-hoc reconstruction. 

The combined ratio will continue its unflinching assessment. The question is whether your organization can explain what produced the number, or only account for it after the fact. 

What do you think? Leave a comment and join the discussion!