The Problem: Product Configuration Has Outgrown the Tools Built for It
Every insurer that has tried to launch a new product knows the pattern. A rating change that should take an afternoon instead takes weeks, because it has to move through a chain of specialists who each hold a piece of the product's logic: rate books, loadings, parent forms, decision tables, exclusions. No single person can see the whole structure, so changes are made cautiously, in isolation, and often without full visibility into what else in the product might be affected.
Product versions drift apart as similar changes are implemented differently by different teams.
None of this is a failure of any one insurer's discipline; it is a structural consequence of managing product logic as static configuration rather than as something that can be queried and modified conversationally. The opportunity is not another layer of tooling, but a different way for product teams to interact with the complexity that already exists.
How the Agentic Product Configurator Works
The Agentic Product Configurator, part of the SAS Insurance Life Cycle Accelerator, is built around a simple premise: product design should read like a conversation with a knowledgeable colleague, not a support ticket. A product owner opens SAS Viya Copilot and asks, in plain language, to see every home insurance product in the portfolio. The system responds with the full inventory product name, ID, storage location, line of business, and version pulled directly from the underlying decisioning environment. A follow up question, asking to see the configuration of a specific product, brings up its complete structure: rate books, loadings, and parent forms.
Figure 1. SAS Viya Copilot retrieves the full home insurance product portfolio and, on request, the detailed configuration of a selected product.
Once a gap or an opportunity for improvement is identified, the same conversational interface handles the change. A request to adjust a single parameter a profit margin moved from 8% to 10%, for instance is applied as a targeted update: only that element changes, every other part of the configuration is left untouched, and the platform automatically creates, names, and persists a new product version as a decision object. Versioning stops being a manual discipline and becomes a system guarantee.
Before a revised or newly assembled product goes live, the agent can generate its own test coverage. One publish command makes the product available for simulation, after which the agent constructs customer profiles varying by age, area, payment method, and household and computes the resulting pure and gross premiums for each, explaining in narrative form why the scenarios differ. This turns scenario testing from a manual, spreadsheet driven exercise into something a product owner can review in minutes.
Figure 2. The agent generates multiple customer test scenarios, computes premiums for each, and explains the drivers behind the differences.
The same approach extends to assembling entirely new combination products. Asked to create a bundle combining a base home product with hail and wildfire extensions, the agent identifies and assembles the components, prices each one, and packages the result as a single bundled decision object ready for publication a single guided exchange a nontechnical product owner can initiate, review, and approve without leaving the conversation.
Figure 3. The agent assembles a multi coverage combo product from a natural language request, pricing each component before packaging it as a bundled decision object.
Business Impact: Why This Matters Now
The case for rethinking product configuration is no longer just intuitive; the evidence for AI assisted configuration, testing, and underwriting workflows is accumulating quickly across the industry.
In its April 2026 analysis of agentic AI in insurance core system modernization, McKinsey found that AI assisted target configuration supporting customization decisions on policy and product platforms through automated generation and validation delivers typical productivity improvements of 15 to 40%, while AI driven testing, reconciliation, and defect cycle compression delivers gains of 15 to 90% (McKinsey & Company, 2026, “Agentic AI to Modernize Insurance Core Technologies,” McKinsey Financial Services Practice). These are precisely the two workflows the Configurator targets: turning configuration change into a validated, generated step, and turning scenario testing into an agent run, explainable process rather than a manual bottleneck.
Consistency, not just speed, is the second lever. BCG's 2026 research on AI first life insurers found that AI assisted workflows improved risk assessment accuracy by 30% (Boston Consulting Group, 2026, “The AI First Life Insurance Company”). A guided configurator that enforces a single, auditable path for every product change addresses the same underlying problem from the design side: fewer manual, ad hoc edits means fewer opportunities for the silent configuration drift that erodes consistency across versions and lines of business.
Speed to market compounds these gains. BCG's parallel analysis of AI first property and casualty insurers found that AI assisted intake and prefill workflows cut time to quote by 30 to 40% (Boston Consulting Group, 2026, “The AI First Property and Casualty Insurer”). The same principle applies further upstream, where the product itself is designed rather than quoted: replacing sequential manual handoffs with an AI guided workflow is what lets a product owner move from portfolio review to a published, tested revision within a single working session instead of a multiweek cycle.
Taken together, this evidence points to the same conclusion from three angles configuration acceleration, testing compression, and underwriting accuracy. The Agentic Product Configurator is designed to capture value on all three fronts at once, rather than optimizing one workflow at the expense of the others.
Governance and Consistency: An Asset, Not a Tradeoff
A natural concern with any AI assisted configuration tool is whether speed comes at the cost of control. For the Agentic Product Configurator, the answer lies in how it is built: every change is targeted, every version is automatically derived and persisted as a traceable decision object, and every AI generated test scenario comes with an explanation of the reasoning behind it. The guided workflow does not remove the underlying rule based, auditable structure of the product configuration it makes that structure easier to interact with and inspect.
