The Pulse
Peak3 launches Graphene v4 and AI delivery lifecycle for insurers
Peak3 has launched Graphene v4, an AI-native insurance core, alongside an AI-driven delivery lifecycle for building and maintaining insurance systems. The Singapore-based company says the platform lets agents configure products and rules wh
AI.info Team ·
Peak3 is asking insurers to let AI do more than process claims or answer customer questions. Its new Graphene v4 platform gives agents direct, governed access to insurance products, calculations and rules, while a separate AI-driven delivery lifecycle applies agent teams to the work of building, testing and maintaining the underlying systems.
The Singapore-based company announced both products on September 15, 2026. Peak3 says the delivery system is the first global insurance AI delivery lifecycle, while Graphene v4 is designed to let agents operate across life, property and casualty, and health insurance. The launch puts a familiar industry dispute in sharper form: whether insurers should place AI on top of existing systems or redesign the core so agents can act within it.
Peak3 wants agents to change the core, not just use it
Graphene v4 includes the Graphene Agent Platform, which Peak3 describes as model-agnostic software for building, running, observing and governing AI agents. Its marketplace includes agents for medical underwriting, conversational first notice of loss, intelligent document processing and fraud detection in claims.
The company’s larger claim concerns design-time work. Agents can reconfigure insurance products, calculations and rules from unstructured documents and natural-language prompts, using the insurer’s own knowledge and data. Peak3 says work that traditionally takes days can be completed in minutes or hours, although the announcement does not provide independent customer measurements or identify the deployments behind that comparison.
Graphene exposes its capabilities through application programming interfaces, the Model Context Protocol and a command-line interface. Peak3 says access management, attribution and detailed logging are enforced in code, rather than left to instructions given to an AI model. The distinction matters for insurers because an agent that changes a product or claims rule needs an auditable record of what it did and under which permissions.
Graphene v4 targets insurers that cannot replace everything
Peak3 is not positioning Graphene v4 only as a replacement for an existing insurance core. The platform can also run as a standalone middle layer between AI agents and an insurer’s systems of record, a deployment option intended to add agent-ready capabilities without a full core replacement.
The release expands Graphene’s stated coverage across several specialised areas. Those additions include the full group-health lifecycle with provider-network management; commercial insurance with co-insurance and reinsurance managed inside the core; indexed universal life and crypto-denominated policies; and native Takaful support alongside conventional insurance.
Peak3 says it serves more than 50 insurance clients in over 20 countries across Europe, the Middle East, Africa and Asia-Pacific. The company also says more than 1 billion policies are transacted each year on its software, figures supplied by Peak3 rather than independently audited in the announcement.
The delivery system turns Peak3’s own methods into software
The second release applies AI across the software delivery process. Peak3 describes a virtual scrum team made up of digital versions of key delivery roles, including business analyst, architect, engineer, tester and reviewer. Those agents are connected to a curated knowledge base and to AI-driven testing, migration, diagnostics and continuous integration and delivery tools.
Bill Song, CEO of Peak3, said: “We are disrupting ourselves. Peak3 is moving from selling software and related services to providing the production line that builds it — our entire AI-DLC.”
Peak3 says it already builds and maintains Graphene with the delivery system. The company is making the capability available to selected implementation partners that deploy Graphene and plans to offer it to clients over the next six to 12 months. Peak3 also says it is expanding forward-deployed engineering teams that will work with clients on both Peak3 software and their existing technology.
A 50% cost claim carries the announcement
Peak3 says the delivery lifecycle has reduced end-to-end new-feature development and core-system implementation costs by 50% where it has been applied. The company says delivery timelines have fallen alongside those costs and sets an ambition of reaching an 80% reduction within 18 to 36 months.
Those figures are company claims, and the release does not specify a baseline project, customer names, contract values or the number of implementations used to calculate them. The immediate commercial test will be whether insurers and implementation partners can reproduce the result outside Peak3’s own development environment, particularly in projects involving legacy data, regulatory review and multiple systems of record.
Peak3’s proposal gives insurers two connected choices: use agents to run operational work through a governed core, or use agents to alter the products and rules that govern that work. The first group of partners will get access before clients, and the client rollout is expected within six to 12 months. The September 15 announcement leaves the next proof point to those deployments: whether an AI delivery system can cut insurance software costs without weakening the controls that make the systems acceptable to insurers and regulators.