Is your insurance core AI-ready?

About a year ago, my team published a whitepaper titled “Insurance Core Platforms in the Age of Agentic AI”. I’ve re-read it this month, and almost nothing in it needs updating. That says something...not about how clever my team was (though they actually are!), but about how persistent the underlying requirements are. 

Here’s the short version. 

AI does not remove the need for a core insurance platform. It raises the bar on what a core has to do. The platforms that succeed in an AI-first operating model will possess eight capabilities. Insurers evaluating their current stack, or assessing a new vendor, should be testing against all eight: 

1. Robust integration and interoperability: API-first, microservices, event-driven architecture, and MCP readiness. If an AI agent can't invoke your core's business functions cleanly, the rest doesn't matter. 

2. Dynamic workflow orchestration: Workflow engines that embed AI-driven decisions at designated points, route around agent failures, support human-in-the-loop gracefully, and produce audit-grade explainability. 

3. A flexible product and pricing engine: Multi-tier product structures, rapid configuration, and the ability to ingest AI-generated product specifications directly into the platform. Hyper-personalisation is impossible on a rigid product model. 

4. Dynamic scalability and performance: Real-time processing, elastic scaling, high concurrency, and cost-aware governance of expensive AI calls. AI agents can multiply transaction volumes. The core has to absorb these volumes. 

5. Stringent security, access control and observability: Fine-grained authorisation with identity propagation across agent chains, comprehensive logging of every AI action, and protection against AI-specific threats. 

6. Unified data management and governance: High-quality, unified data access across modules; systematic data lineage for AI outputs; strict compliance controls including data masking and alignment with privacy regulation. 

7. Multi-tenancy and multi-country support: Reusable agents, shared training data, and unified governance across geographies. This is only possible on a genuinely multi-tenant core. 

8. Easy maintainability and upgradeability: AI is evolving fast. Platforms that age well are those designed to absorb change through regular upgrades, configurations and extension points...not difficult to maintain code forks. 

The truth is that most insurers' self-assessments would reveal that their current cores materially fail at least four of the eight dimensions (and have some shortcomings among all of the others). The follow-up question is whether incremental remediation closes the gap, or whether the honest answer is a comprehensive modernisation programme. That is a board-level conversation, not just an IT conversation. 

At Peak3, we've built our core platform Graphene to support these eight capabilities as the scoring rubric from the outset. Not because we predicted agentic AI, but because a core system that fails any of them ages badly for reasons that have nothing to do with AI. Agentic AI just accentuates the requirements.

-Bill Song, Peak3 Co-Founder and CEO