Why claims management system selection matters for U.S. P&C carriers in 2026
Selecting a claims management system in 2026 is no longer an IT project with a procurement timeline. It is the single largest infrastructure decision a VP Claims makes in a five-year window, and it shapes combined ratio, NAIC examination readiness, and J.D. Power ranking through the duration of the contract. In my experience working with VPs of Claims at U.S. P&C carriers, the 2024-2025 cohort that selected on demo polish alone is the same cohort writing replacement RFPs in 2026.
When this piece was first published in February 2025, the discussion around CMS selection focused on feature lists. Twelve months later, three macro forces have rewritten what selection actually means. Deloitte’s 2026 Global Insurance Outlook reports premium growth declining through 2026 under heightened competition and cost pressure, with 90 percent of executives recognizing the need to upskill for human-AI collaboration but only about 25 percent having acted. The NAIC AI Model Bulletin has been adopted by 24-plus U.S. jurisdictions as of August 2025, and the AI Systems Evaluation Tool pilot launched in January 2026 across 12 states. CAT volatility from 2024-2025 wildfire and hurricane seasons broke the lifecycle assumptions of carriers that had not built surge capacity into their core systems.
Together these forces mean claims management software selection in 2026 is gated on three things: operating-model fit (not feature checklist), NAIC AIS Program readiness (not optional), and integration architecture for AI-era claims (not a deferred phase 2). The carriers I work with that hit those three correctly compress LAE and protect combined ratio. The carriers that miss them rebuild in 36 months.
This article is the operating buyer’s framework, not a vendor comparison. It assumes you are a VP Claims, Head of Claims, or Chief Claims Officer at a U.S. P&C carrier in active or planned procurement. For the underlying architecture and FNOL revolution context, the complete 2026 guide to AI in insurance claims processing is the longer reference.
What a 2026 claims management system must do (the 7 capability layers)
A 2026 CMS is seven capability layers, not a feature list. In my experience working with VPs of Claims who have run 2024-2025 selections, evaluating by feature checklist produces what looks complete on an analyst-style feature matrix and then breaks at the first integration. Evaluating by capability layer produces infrastructure that compounds across the next decade. The claims management system features below are the ones I require when running a 2026 selection.
Layer 1: Structured intake across channels
Every channel - mobile, web portal, phone via IVR or live agent, broker submission, IoT trigger - must produce one canonical claim record with the same schema. OCR and NLP run inside the intake layer to handle unstructured artifacts. Identity verification, first-pass coverage check, and duplicate-claim detection are intake-layer responsibilities, not adjuster work. A 2026 system that requires adjusters to re-key data across channels is not a 2026 system.
Layer 2: Decision-grade triage and routing
Triage decides where the claim goes and at what authority level - severity, complexity, fraud signal, vendor needs, and policyholder profile, jointly. Routing logic must be editable by claims operations leadership, not IT. Higson, our rules engine, executes triage and routing rules at sub-millisecond latency for the carriers I work with. The capability test for this layer is whether your Head of Claims can change a routing rule on a Tuesday afternoon and see it in production by Wednesday morning without an IT ticket.
Layer 3: Investigation and evidence assembly
Investigation in 2026 is mostly document orchestration: medical reports, police reports, repair estimates, photo evidence, witness statements, expert opinions. The claims management software must assemble the evidence chain in machine-readable form, cross-reference external databases (NICB, ISO ClaimSearch, OFAC), surface gaps to the adjuster as actionable tasks, and flag fraud signals at submission rather than at closure. The Anti-fraud Solution sits inside this layer.
Layer 4: Decisioning and reserves
Coverage adjudication runs in the rules engine against structured intake plus evidence. Reserve recommendation runs as a separate ML model (frequency-severity, GBM, or similar) with mandatory adjuster validation above a documented authority threshold. Litigation and subrogation potential are flagged in this layer, not at closure. The capability test is whether the system can show, for any given claim, why a reserve was set at the amount it was set at - the explainability requirement that NAIC AIS Program puts at the center of compliance.
Layer 5: Settlement, payment, and subrogation
The settlement layer handles the money plus the paper trail. Auto-pay for STP-eligible claims within authority guardrails. Subrogation identification driven by analytics on payment, fault, and recoverable third-party patterns. Audit trail with full NAIC-aligned documentation. In my experience working with Heads of Claims, this is where most carriers leave money on the table - subrogation recovery rates remain well below what underlying loss data could support, because the flagging happens after settlement rather than during investigation.
Layer 6: Vendor orchestration
Repair shops, appraisers, medical case managers, defense counsel, vendor scorecards, SLA tracking, cost reconciliation - all inside the CMS, not in separate tools. Vendor SLA breaches must surface in the real-time dashboard (Layer 7 / control tower), not in a monthly report. The carriers I work with that orchestrate vendors inside the CMS rather than across email and shared drives close claims 25 to 40 percent faster on average.
Layer 7: Audit trail, compliance, and the control tower
Every AI-influenced decision logged with input data, model version, output, and adjuster override (if any). Governance documentation maintained as a living artifact, not a one-time PDF. Real-time dashboard for VP Claims with drill-down from portfolio to single-claim level. This is the layer where NAIC AIS Program governance lives operationally, not as a compliance afterthought. The Operational Data Store is the data foundation that makes the control tower practical without a full data warehouse rebuild.
Build vs Buy vs Hybrid - the decision framework
The build-versus-buy decision for a CMS in 2026 is not a binary. It is a three-way choice between full custom build, off-the-shelf platform, and hybrid - and the right answer depends on portfolio size, product complexity, and the carrier’s tolerance for vendor concentration risk. I’d recommend Heads of Claims run this framework before any vendor demos, because the demos are useless if you have not already chosen the architecture path.
When full custom build makes sense
Full custom builds fit carriers with portfolios above roughly $2 billion in DWP, complex specialty or program business with unique workflows, and internal engineering capability of 30-plus dedicated developers. Total cost of ownership over a 7-year horizon typically lands 20 to 40 percent above off-the-shelf, but the carrier owns the roadmap and the integration surface. Carriers below the scale threshold who build typically run out of internal capacity around year three and end up rebuilding on a platform anyway.
When off-the-shelf platforms make sense
Off-the-shelf claims management software fits carriers between roughly $500 million and $2 billion in DWP, standard P&C lines (personal auto, homeowners, small commercial), and limited internal engineering capacity. Implementation typically runs 9 to 18 months depending on integration scope. The trade-off is roadmap dependency on the vendor and the integration tax for anything outside their pre-built connectors.
When hybrid is the right answer
Hybrid architectures - off-the-shelf core for standard workflows plus custom-built layers for differentiated capabilities (fraud scoring, reserve modeling, specialty product workflows) - fit a majority of U.S. P&C carriers I work with in 2026. The pattern: buy the operational core, build the differentiators, integrate via API. The Decerto Claims AI System sits in the hybrid pattern as the AI and decisioning layer alongside an existing core claims platform.
The total cost of ownership reality
The license cost is typically 15 to 25 percent of seven-year TCO. The remaining 75 to 85 percent is implementation, integration, custom development, training, and operational change management. Carriers that select on license cost alone are looking at the wrong number. The carriers I have helped with selection consistently underestimate the change management line item by 50 to 100 percent.
NAIC AIS Program as selection criterion
NAIC AI Systems (AIS) Program readiness is a non-negotiable selection criterion in 2026, not a phase 2 deliverable. The NAIC Model Bulletin on Use of Artificial Intelligence Systems by Insurers, adopted by NAIC in December 2023, requires insurers to adopt, implement, and maintain a documented AIS Program covering governance, risk management, internal controls, third-party AI management, and consumer outcomes. As of August 2025, 24-plus U.S. jurisdictions had adopted full or substantially similar versions. The AI Systems Evaluation Tool pilot launched in January 2026 across 12 participating states, covering four exhibits: breadth of AI adoption, governance framework, deep dive on high-risk systems, and data source review.
Audit trail capability - the non-negotiable
Every AI-influenced decision in the CMS must be logged with input data, model version, output, and adjuster override (if any). The audit trail must survive at least the state-mandated retention period (typically 3 to 7 years depending on jurisdiction and claim type). In my experience, carriers that retrofit audit trail capability into their insurance claims management software after deployment spend two to four times what they would have spent including it in selection criteria from day one.
Explainability for adverse consumer outcomes
The Model Bulletin defines adverse consumer outcomes as decisions that impact consumers and violate insurance regulatory standards. Any AI-driven decision in claims that could result in such an outcome (coverage denial, reduced settlement, fraud investigation, delayed payment) must be explainable to the consumer, to the adjuster, and to the state DOI examiner. The claims management software must surface the reasoning, not just the output. Holland & Knight’s regulatory analysis underlines that AIS Programs must incorporate verification and testing methods to identify errors, bias, and unfair discrimination.
Third-party AI management
If your CMS vendor uses third-party AI models (computer vision providers, NLP services, fraud scoring APIs), your AIS Program must extend to those third parties under the Model Bulletin. The selection question is whether the vendor can provide written documentation of their third-party AI governance, model cards, and incident reporting. Vendors that cannot are deferring your compliance risk back to your team.
What the AI Systems Evaluation Tool means for vendor selection
Even insurers in non-pilot states should treat the four AIS Evaluation Tool exhibits as a template. The exhibits will become the de facto examination format across adopted states within 18 to 24 months. The selection question for any CMS vendor is whether their AIS Program documentation maps to the four exhibits today, not in two years.
Integration requirements - what your CMS must connect to
A CMS that does not integrate cleanly is a stranded asset within 36 months. Claims management system integration architecture is one of the top three reasons carriers replace their insurance claims management software after 5 to 7 years, and one of the top three reasons new selections fail at implementation. I worked with a Southeast P&C carrier (around $1.2 billion in DWP, mixed personal auto and homeowners) whose CMS selection in 2022 went well on every dimension except integration. The vendor’s pre-built PAS connector covered 60 percent of fields. The remaining 40 percent took an 18-month custom integration project. By the time integration was done, the carrier’s next NAIC examination cycle was six months away and their AIS Program documentation had not been updated. I’d require integration depth as a selection scoring criterion, not an implementation phase.
PAS integration patterns
The CMS must integrate with policy administration for coverage check, policy details retrieval, endorsement history, and premium continuity into reserves. API-first connectors over batch file transfers, real-time over nightly sync. The integration test during vendor evaluation: ask the vendor to demonstrate a real claim flowing from FNOL through coverage check using your specific PAS, not their generic demo environment.
Billing and payment systems
Payment authorization, reserve releases, settlement disbursements, vendor payments, and recovery receipts all flow through billing integration. The capability test is whether the CMS can issue an STP payment within seconds of decision approval without an overnight batch step. Carriers stuck with overnight batch payment are losing 24 hours of cycle time on every clean claim.
External data sources - NICB, ISO ClaimSearch, OFAC
NICB for fraud cross-reference, ISO ClaimSearch for prior claim history, OFAC for sanctions screening - all real-time API integrations inside the CMS, not adjuster work. The cycle time targets discussed below assume these integrations are running at sub-second latency. Carriers without them lose 30 to 60 minutes per claim on adjuster lookup work that should be automated.
CRM and customer communication
The CMS must integrate with CRM for 360-degree customer view, communication history, and post-claim NPS triggers. Email, SMS, mobile notification, and customer portal updates all from one record. The Self-Service Customer Portal is where this layer touches the policyholder.
Vendor and partner ecosystems
API connectivity to repair shop networks, medical case management partners, appraiser networks, and defense counsel platforms. The capability test is whether vendor SLA tracking, scorecards, and cost reconciliation live inside the CMS or in separate spreadsheets.
KPIs a 2026 CMS must deliver
A CMS is not judged on its features. It is judged on the KPIs it moves. The carriers I work with that evaluate vendors on KPI delivery rather than feature checklist consistently select better and implement faster. The KPIs that matter for a 2026 CMS selection are the ones that move combined ratio, NAIC examination outcomes, and J.D. Power scores - not vanity metrics like total claims processed.
Realistic Year-2 mature deployment targets by line of business
STP rate gates - what they actually depend on
The single biggest predictor of STP rate is data completeness at FNOL, not engine capability. A carrier with 40 percent digital FNOL will not hit 50 percent STP on personal auto regardless of how sophisticated the decisioning layer is. The selection question is not “what STP rate can your platform achieve” - it is “given my current intake mix, what STP rate is realistic in 18 months”. Vendors who answer the first question without asking the second are selling.
Reserve adequacy and adverse development
Reserve adequacy gates on model training data - 18 months minimum, ideally 36 months of clean claim outcomes. A new CMS deployment does not deliver target reserve adequacy in year one even if the model architecture is excellent. The selection criterion is whether the vendor’s reserve model can ingest your historical claim data during implementation, not whether their generic model is “industry-leading”.
Customer experience metrics
NPS post-claim, first contact resolution rate, and digital channel adoption rate. J.D. Power’s 2025 U.S. Claims Digital Experience Study confirms that digital experience and proactive communication are now the top drivers of post-claim satisfaction, with mobile app status updates identified as the strongest performer among notification channels. The CMS must deliver these capabilities natively, not as a customer portal add-on.
10-point vendor evaluation scorecard
Vendor demos optimize for what looks good in a 60-minute slot. Selection optimizes for what survives the next decade. In my experience, the claims management system scorecard below is the one that produces the most reliable claims management software evaluation outcomes when running 2026 CMS selections with VPs of Claims. Weights reflect the relative impact on cycle time, NAIC compliance, and seven-year TCO.
The 10-criterion weighted scorecard
How to use the scorecard
Score each vendor 1 to 10 per criterion, multiply by weight, sum to a percentage. Anything below 70 percent is a no-buy regardless of demo polish. Anything above 85 percent is implementation-ready. The 70-to-85 percent zone is where careful POC design matters most.
Red flags during vendor demos
A vendor who cannot demonstrate your specific PAS integration in their demo is deferring integration risk to your team. A vendor whose AIS Program documentation is “in progress” is deferring NAIC compliance risk to your team. A vendor whose rules engine requires IT change tickets for routing rule updates is deferring agility loss to your team. A vendor whose reference customers are not willing to take an unscripted call is hiding implementation history.
What to ask references
Ask the implementation timeline reality, not the published one. Ask the integration scope creep, not the initial scope. Ask the change management cost, not the license cost. Ask whether the reference would buy the system again knowing what they know now. The honest answer to the last question is more predictive than any vendor brochure.
Five common CMS selection mistakes
I have watched roughly five recurring selection mistakes derail CMS procurement over the last decade. Each is preventable with discipline. None of them are vendor problems - they are buying-process problems.
Mistake 1: Buying features instead of operating-model fit
The procurement team builds an RFP from a feature checklist scraped from analyst sites and asks vendors to confirm or deny. Vendors confirm everything. The carrier picks the vendor with the most green check marks. Six months into implementation, the operating model breaks because no feature scoring evaluated how the seven capability layers work together. Mitigation: score on the 7-layer framework, not on feature counts.
Mistake 2: Ignoring NAIC AIS Program selection criteria
The compliance team is brought in three weeks before contract signature to “review the AI piece”. By then the AIS Program gaps are baked into the architecture, and retrofitting governance costs two to four times what it would have cost to require in selection. I’d require AIS Program documentation as a gating criterion at RFP stage, not at contract review.
Mistake 3: Underweighting integration complexity
The vendor’s pre-built PAS connector “covers most fields”. Most is not all, and the 20 to 40 percent of fields the connector does not cover always live in your hardest workflows. Integration scope creep typically adds 6 to 18 months to implementation timelines. Mitigation: require the vendor to demonstrate your specific PAS integration with your real data during evaluation, not their generic demo.
Mistake 4: Selecting on demo, not on proof-of-concept data
Vendor demos optimize for the 60-minute slot. POCs with your actual claim data optimize for production reality. The vendors that survive POC are the vendors that will survive implementation. The carriers I work with that skip POC because “we have already seen the demo” consistently end up in the 36-month replacement cycle.
Mistake 5: Treating implementation as IT project, not operating change
The IT team owns the implementation. Operations gets trained two weeks before go-live. Adoption stalls because adjusters were not part of design. The CMS technically works, the operating model does not change, and combined ratio does not move. Mitigation: VP Claims owns implementation, IT supports - not the other way around. Change management budget at 100 to 150 percent of license cost is realistic for a CMS replacement.
FAQ - Claims management system selection
What features should a 2026 claims management system have?
A 2026 CMS needs seven capability layers: structured intake across channels, decision-grade triage and routing, investigation and evidence assembly, decisioning and reserves, settlement and subrogation, vendor orchestration, and audit-trail compliance with a real-time control tower. Evaluating by capability layer rather than by feature checklist consistently produces better selection outcomes and shorter implementation timelines.
How do you evaluate a CMS for U.S. P&C carriers?
Use a weighted scorecard across operating-model fit (15%), NAIC AIS Program readiness (12%), PAS integration depth (12%), rules-engine flexibility (10%), decisioning AI architecture (10%), real-time dashboard (8%), vendor orchestration (8%), implementation track record (8%), TCO (10%), and vendor stability (7%). Score 1 to 10 per criterion. Below 70 percent is a no-buy regardless of demo polish.
What is the difference between claims management software and a policy administration system?
A policy administration system (PAS) handles the policy lifecycle - quote, bind, issue, endorse, renew, cancel. Claims management software handles the claim lifecycle - intake, triage, investigation, decisioning, settlement, subrogation. They integrate via API for coverage check, policy details, and premium continuity, but they have separate functional scope and separate claims management system requirements during selection.
How does a claims management system integrate with PAS?
Modern claims management software integrates with PAS via API-first, real-time connectors covering coverage check, policy details retrieval, endorsement history, and premium continuity into reserves. The integration test during vendor evaluation is whether a real claim can flow from FNOL through coverage check using your specific PAS, not the vendor’s generic demo environment. Batch file transfers are a 2026 red flag.
What NAIC requirements apply to AI inside a CMS?
The NAIC AI Model Bulletin (adopted December 2023) requires a documented AI Systems (AIS) Program covering governance, risk management, internal controls, third-party AI management, and adverse consumer outcomes. As of August 2025, 24-plus U.S. jurisdictions had adopted. The AI Systems Evaluation Tool pilot launched January 2026 in 12 states. Audit trail capability and explainability are non-negotiable CMS NAIC compliance requirements.
How long does CMS implementation typically take for P&C carriers?
Off-the-shelf claims management software implementation typically runs 9 to 18 months depending on integration scope, line-of-business complexity, and change management readiness. Custom builds run 18 to 36 months. Hybrid architectures fall in the 12 to 24 month range. Change management work (training, adoption, process redesign) is typically 30 to 50 percent of the calendar time and is the most underestimated component.
What KPIs should a CMS improve for U.S. P&C carriers?
The KPIs that matter for combined ratio are cycle time by line of business, STP rate on clean claims, reserve adequacy (within 10 to 15 percent of final paid), first-touch routing accuracy, reopen rate, subrogation recovery rate, and post-claim NPS. The KPIs that matter for compliance are audit trail completeness, AIS Program documentation currency, and adverse consumer outcome traceability.
Should U.S. P&C carriers build or buy their CMS platform?
Carriers above roughly $2 billion in DWP with complex specialty business and 30-plus engineers can rationally build. Carriers between $500 million and $2 billion in DWP with standard P&C lines should buy. A majority of U.S. P&C carriers in 2026 are best served by hybrid - buy the operational core, build the differentiators, integrate via API. Build versus buy depends on portfolio scale, product complexity, and internal engineering capacity, not on philosophy.
Talk to Decerto Claims AI
Every additional month spent in the wrong CMS is a point of LAE that does not come back, a NAIC examination risk that does not get smaller, and a J.D. Power ranking that drifts the wrong direction. Soft-market premium pressure means expense ratio scrutiny is now monthly, not quarterly. NAIC AIS Program adoption is moving faster than most carriers’ internal compliance work. The window where slow selection is acceptable closes during 2026.
I run a 30-minute CMS selection assessment with VPs of Claims and Heads of Claims at U.S. P&C carriers. It is a working conversation, not a sales call. We score your current platform (or your shortlist) against the 7-capability framework in Section 2, identify the integration and AIS Program gaps that gate realistic outcomes, model what 12 to 18 months of replacement or augmentation would deliver against your specific portfolio, and walk through what hybrid versus full replacement looks like. You leave with a written assessment whether or not we ever do business together.
Not a generic demo. The first call is technical Q&A. Your portfolio data stays on your side under NDA - we work through what is possible against the gaps you already know about. You might wonder whether selecting Decerto means replacing your existing core claims platform. The honest answer is usually no. The Decerto Claims AI System and Higson rules engine sit alongside what you already have for most carriers in the hybrid pattern from Section 3, accelerating intake, triage, decisioning, and audit trail without forcing a core system replacement.
Decerto’s track record outside the U.S. illustrates the architectural pattern. The centralized claims handling deployment at BNP Paribas Cardif (part of BNP Paribas Group; in production with Higson since 2016) consolidated claims handling onto one platform where business teams edit forms and rules without IT involvement. The lesson transfers cleanly to U.S. P&C selection: get the rules editable by claims operations, not IT, and the selection compounds for the next decade.
What is included in the 30-minute assessment:
- Working session with Marcin Nowak and a senior solution architect from the Decerto team
- 7-capability framework scoring of your current CMS or vendor shortlist
- NAIC AIS Program readiness gap analysis against the four AI Systems Evaluation Tool exhibits
- Free access to the Decerto Claims AI sandbox and the 10-point vendor evaluation scorecard from Section 7
We onboard a limited number of new U.S. carriers per quarter. Q3 2026 capacity is filling.
Sources and citations
- Deloitte Insights. 2026 Global Insurance Outlook. December 2025.
- National Association of Insurance Commissioners. Model Bulletin on Use of Artificial Intelligence Systems by Insurers (adopted December 4, 2023).
- National Association of Insurance Commissioners. Implementation of NAIC Model Bulletin: Use of Artificial Intelligence Systems by Insurers (August 2025 state adoption tracker).
- Holland & Knight LLP. The Implications and Scope of the NAIC Model Bulletin on the Use of AI by Insurers. May 2025.
- J.D. Power. 2025 U.S. Claims Digital Experience Study. 2025.
- McKinsey & Company. How data and analytics are redefining excellence in P&C underwriting.
- Quarles & Brady LLP. Nearly Half of States Have Now Adopted NAIC Model Bulletin on Insurers’ Use of AI. March 2025.
- Datos Insights (formerly Aite-Novarica Group). Straight-Through Processing in Underwriting and Claims: 2023 Update.
Subscribe to receive the latest blog posts to your inbox every week.
.png)
.png)
