.avif)
Why the PAS-underwriting boundary matters more in 2026
The relationship between policy administration software and underwriting is the part of the carrier stack most often misconfigured. From the carrier transformations I have advised on, the boundary between PAS and underwriting is where rule violations, audit gaps, and broker complaints actually live.
In 2026, the boundary matters more because regulators now have the tools to examine it. The NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers (December 2023) has been adopted by 25 states and the District of Columbia as of August 31, 2026, and it lists the documentation a Department of Insurance may request about how AI systems are governed and used. The NAIC AI Systems Evaluation Tool, piloted by 12 states in 2026, gives examiners a standard way to ask for it. When an examiner traces an AI-supported decision, the trail runs from submission intake, through the underwriting decision, to the PAS policy of record.
In my 20 years on the carrier side, I have seen three Boards in the last two years buy a workbench and a new PAS in the same RFP cycle, then realize 8 months in that the two systems did not speak the same data model. The integration layer between underwriting decisions and PAS policy-of-record records is where most of the post-deployment defects show up. My take, after sitting in many of those steering committee post-mortems: scope the boundary first, scope the vendors second.
What policy administration software actually does (and does not do)
Policy administration software owns the policy of record. It handles policy issuance, endorsements, renewals, cancellations, billing setup, and the lifecycle events that happen after binding. It maintains the canonical record of what coverage exists, on what terms, for what premium, with what reinsurance treaty allocation. State DOIs examine PAS records against the rate and form filings carriers submit, most often through the NAIC SERFF filing system, to verify form usage and policy compliance.
What PAS does not do well, by design, is dynamic risk evaluation, rule changes outside rate filings, AI scoring, or document automation on inbound submissions. PAS systems are optimized for stability and auditability of the policy of record, which makes them poor fits for the rapidly changing rule logic and AI scoring that underwriting requires in 2026. This is why modern carriers split underwriting decision logic out of PAS and into a dedicated underwriting workbench.
The major U.S. policy administration software vendors (Guidewire PolicyCenter, Duck Creek Policy, Majesco, Insurity) all support this split. Some have underwriting modules of their own, but in the vendor evaluations I have run for mid-tier carriers, the bundled underwriting modules consistently trail standalone workbenches on AI capability and rules-engine flexibility.
Where underwriting depends on PAS and where it should not
Underwriting depends on PAS for
Product definition and rate tables (what coverages exist, what the filed rates are, what states the product is approved in). Policy history and loss data (the back-book the underwriter scores new risks against). Reinsurance treaty allocation rules (how new policies map to treaty layers). Compliance with state DOI rate and form filings. Billing setup feeding into commission and premium accounting.
Underwriting should not depend on PAS for
Eligibility rules (these change too often to live in PAS rate filings). AI risk scoring (PAS systems are not built for ML model versioning, A/B testing, or explainability layers). Document automation for inbound submissions (PAS is record-of-truth, not intake processing). Real-time portfolio exposure dashboards (PAS reporting cycles are typically batch). Per-decision audit trail for AI Bulletin compliance (PAS audit trails capture policy events, not underwriting decisions).
Carriers that pushed all of these into PAS in the 2010s ended up with bloated, expensive, slow-to-change systems that satisfied neither the underwriting workflow nor the policy-of-record audit requirement. The 2026 architecture separates them cleanly.
The PAS-workbench data flow in a modern 2026 stack
In a modern carrier stack, data flows bidirectionally between the underwriting workbench and the policy administration system. Submissions arrive at the workbench from broker portal, email, or PAS-initiated renewal triggers. The workbench applies rules, runs AI scoring, and produces a decision (bind, decline, refer, surcharge) with full per-decision audit trail. Bind decisions flow to PAS for policy issuance against the filed product and rate. PAS issues the policy and feeds back the policy number, billing setup, and reinsurance allocation. Endorsements and renewals trigger re-evaluation through the workbench rules.
The clean version of this looks simple. The production version involves data model alignment, party master synchronization, exposure aggregation, document handoff, renewal trigger timing, cancellation cascade handling, and a dozen other integration concerns. I have rarely seen a carrier underestimate the policy issuance integration and overestimate it at the same time; they are usually different conversations.
Four integration patterns that work and one that does not
Pattern 1: API-based middleware between workbench and PAS
REST or event-driven API integration through a middleware layer. The workbench publishes underwriting decisions; the PAS subscribes and issues policies. This is the most common pattern in mid-tier carrier deployments and works well with Guidewire PolicyCenter, Duck Creek Policy, and Majesco. Build time: 3-6 months for a single line of business.
Pattern 2: Workbench as PAS underwriting service
The workbench exposes underwriting decisions as a service the PAS calls during policy creation. PAS-initiated, workbench-fulfilled. Works well when PAS is the dominant system of record and the workbench is a focused underwriting service. Build time: 4-7 months.
Pattern 3: Phased data model alignment
Workbench and PAS each maintain their canonical data models, with explicit translation at the integration boundary. Avoids the trap of trying to force one system's data model on the other. Build time: 4-8 months. This is the pattern I most often recommend for carriers with mature PAS configurations.
Pattern 4: Pre-built connector with PAS vendor
Some workbench vendors ship pre-built connectors for Guidewire, Duck Creek, or Majesco. These accelerate the integration but should be validated against your specific PAS version and configuration before scoping. Build time: 2-4 months when the connector matches your environment; longer when it does not.
Pattern that does not work: shared database
Direct database-level integration where the workbench writes to PAS tables (or vice versa) creates fragile, brittle systems that break on PAS upgrades. I have seen this pattern fail in three carrier deployments. Avoid it regardless of how attractive the short-term build time looks.
PAS choice and underwriting workbench compatibility
Most mid-tier U.S. P&C carriers in the $500M-$5B GWP range run one of four PAS choices: Guidewire PolicyCenter, Duck Creek Policy, Majesco, or Insurity. From the integrations I have advised on, all four support modern workbench layering, but with different characteristics.
Guidewire PolicyCenter is the most common PAS at the upper end of mid-tier ($2B-$5B GWP). Its API surface is mature, well-documented, and supports the API-based middleware and PAS-underwriting-service patterns directly. Workbench integration cost is predictable.
Duck Creek Policy has a smaller market share but strong configuration flexibility. Workbench integration through Duck Creek's APIs works well; the gotcha is usually that Duck Creek configurations vary widely between carriers, so a vendor reference customer on Duck Creek may not predict your integration cost.
Majesco is common in mid-tier specialty and personal lines. Its cloud-native architecture makes API integration straightforward; workbench layering on Majesco is typically the fastest of the four.
Older on-premise installations, whether Insurity, Duck Creek releases from the Accenture era, or custom in-house PAS, typically require more integration build. For these, Pattern 3 (phased data model alignment) is the safer approach.
For the detail on integrating workbenches with legacy systems specifically, see our legacy systems integration article. For the full feature checklist when selecting a workbench, see top features to look for. For the comparison between modern workbenches and traditional underwriting tools, see workbench vs traditional tools. For the regulatory side of the same boundary, see our NAIC AI Bulletin compliance playbook.
Frequently asked questions
How does an underwriting workbench integrate with policy administration software?
Modern underwriting workbenches integrate with policy administration software through API-based middleware or pre-built connectors, with bidirectional data flow. Submissions arrive at the workbench from broker portals or PAS-initiated renewal triggers. The workbench applies rules, runs AI scoring, and produces an underwriting decision with audit trail. Bind decisions flow to PAS for policy issuance against the filed product. PAS returns policy number, billing setup, and reinsurance allocation. This split (the workbench owns underwriting decisions and their audit trail; the PAS owns the policy of record) is the 2026 standard architecture for mid-tier P&C carriers.
Can you run a modern underwriting workbench on a legacy policy administration system?
Yes. Modern workbenches are designed to layer on top of legacy PAS systems through API middleware, pre-built connectors, or phased data model alignment. Higson, for example, is API-first and technology-agnostic: it manages underwriting, eligibility, and pricing rules and is not itself a PAS, so it sits alongside the existing policy system instead of replacing it. The integration complexity varies: cloud-native PAS systems like Majesco and modern Guidewire support the fastest integration, while older on-premise PAS may require 6-9 months of integration build. Direct database-level integration should be avoided regardless of PAS choice: it creates fragile systems that break on PAS upgrades.
What data flows between underwriting and policy administration software in 2026?
Bidirectional data flow covers six categories. From PAS to workbench: product definitions, rate tables, policy history, loss data, party master records, and reinsurance treaty rules. From workbench to PAS: underwriting decisions (bind, decline, refer, surcharge), risk scores, audit trail references, and document attachments. PAS returns: policy number, billing setup, reinsurance allocation, and confirmation of policy issuance. Renewal triggers and endorsement events flow back to the workbench for re-evaluation. Document automation typically lives in the workbench, not PAS, with extracted data feeding both systems.
Does NAIC AI Bulletin compliance require changes to both PAS and underwriting software?
Primarily underwriting software, with PAS playing a supporting role. The NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers, adopted by 25 states and the District of Columbia as of August 31, 2026, expects insurers to maintain a written AI Systems (AIS) Program covering governance, risk management and internal controls, and oversight of third-party AI systems, and it lists the documentation a Department of Insurance may request during an examination. New York applies its own framework, DFS Circular Letter No. 7 (2024), to AI and external data used in underwriting and pricing. Neither document prescribes an audit-trail format, but decision-level records are the practical way to produce what examiners ask for. Underwriting workbenches own the decision logic and the AI scoring, so they own most of the compliance artifacts. PAS needs to provide a policy-level audit trail referencing the underwriting decision, so regulators can trace from policy back to the decision that produced it, but does not typically need to host the model documentation itself.
How long does PAS-workbench integration take for a $1B GWP carrier?
Typical integration timeline for a $1B GWP mid-tier P&C carrier: 3-6 months for API-based middleware integration with Guidewire PolicyCenter or Majesco for a single line of business; 4-8 months for Duck Creek Policy depending on configuration depth; 6-9 months for legacy on-premise PAS using phased data model alignment. Pre-built connectors can compress these by 1-2 months when the connector matches the PAS version and configuration. The bottleneck is rarely the integration technology itself: it is data model alignment, party master synchronization, and exposure aggregation logic.
Should mid-tier carriers buy a single PAS-plus-underwriting bundle or separate systems?
Most mid-tier P&C carriers I work with end up with separate systems because the underwriting modules bundled into PAS suites typically trail standalone workbenches on AI capability, rules-engine flexibility, and explainability. The exception is carriers where the underwriting workflow is very simple (small commercial, narrow personal lines) and the PAS-bundled module is sufficient. For specialty, complex commercial, or carriers planning meaningful AI scoring deployment through 2028, the standalone workbench layered on PAS is the more durable architecture. The trade-off is one more vendor relationship to manage.
Talk to Decerto about Higson
If you are a Chief Underwriting Officer or VP Underwriting at a U.S. P&C carrier in the $500M-$5B GWP range, and your current underwriting workflow runs on a mix of PAS screens, Excel rating sheets, and email-based broker submissions, that is the exact pattern Higson was built to address. Higson layers on top of Guidewire PolicyCenter, Duck Creek Policy, or Majesco without forcing a PAS replacement, and ships a CUO-controlled rules engine that cuts rule deployment from a 4-month IT backlog to 24 hours.
My take, after 20 years on the carrier side: Higson is not the right fit for $5B+ enterprise carriers running multi-region multi-currency books across 30+ jurisdictions. Guidewire is. Higson is built for U.S. P&C carriers in the $500M-$5B GWP range that need rules-engine control and audit-grade documentation without a 24-month PAS replacement.
30-minute tailored demo with a Decerto solution architect. We will map your current underwriting workflow to Higson's rules engine, AI scoring, and audit trail.
Sources
- National Association of Insurance Commissioners. (2023, December 4). Model Bulletin: Use of Artificial Intelligence Systems by Insurers. NAIC.
- National Association of Insurance Commissioners. (2026, August 31). Implementation of NAIC Model Bulletin: Use of Artificial Intelligence Systems by Insurers (state adoption map and reference list). NAIC. Accessed 2026-09-30.
- National Association of Insurance Commissioners. (2026, April 3). Insurance Topics: Artificial Intelligence (AI Systems Evaluation Tool pilot status). NAIC. Accessed 2026-09-30.
- New York State Department of Financial Services. (2024, July 11). Insurance Circular Letter No. 7: Use of Artificial Intelligence Systems and External Consumer Data and Information Sources in Insurance Underwriting and Pricing. NY DFS.
- National Association of Insurance Commissioners. (n.d.). System for Electronic Rates and Forms Filing (SERFF). NAIC. Accessed 2026-09-30.
Subscribe to receive the latest blog posts to your inbox every week.
