If you have difficulty in submitting comments on draft standards you can use a commenting template and email it to admin.start@bsigroup.com. The commenting template can be found here.
Standards Development
This Technical Report specifies a collection of representative use cases and guidance for exposing insurance contract, policy and related data — modelled according to prEN 18356-1 — to AI systems (e.g. large language model-based assistants and agents) via the Model Context Protocol (MCP) or comparable AI-tool-integration mechanisms.
The scope includes:
- A description of the relationship between the prEN 18356-1 data model/API schemas and MCP tool concepts (tools, input schema, output data, resources).
- A common use case template comprising: business description/context, actors involved (e.g. insurer, intermediary, third-party provider, customer), input data, output data, required/mandatory data fields, applicable business rules, and security/consent-related aspects.
- A selection of concrete, illustrative use cases based on the insurance product lines covered by prEN 18356-1 (e.g. motor, property, liability, personal accident, legal expenses, travel, financial loss, life and pension insurance), including but not limited to:
- Retrieval of customer and contract/policy information (search, listing, filtering by type, status or validity period).
- Retrieval of aggregated/statistical portfolio information.
- Illustrative guidance on further action-oriented use cases (e.g. quotation requests, contract change requests) at a conceptual level.
- Example MCP tool definitions (name, description, input schema) referencing prEN 18356-1 data elements, together with sample request/response data, to illustrate how AI systems can be guided to select and invoke the correct tool.
- Guidance on naming conventions, tool descriptions and disambiguation practices that support reliable tool selection by AI systems.
- Considerations on versioning, backward compatibility and the relationship between semantically stable core data objects (as defined in EN pr18356-1) and more volatile technical implementation details.
Out of scope:
- Definition of a new normative API or data model (this remains the role of EN pr18356-1 and any related CEN/TC 445 deliverables).
- Standardisation of the Model Context Protocol itself, or of any specific AI/LLM technology, vendor product or programming language/framework.
- Detailed security architecture or authentication protocol specifications (reference to existing standards/guidance only).
CEN/TC 445 has developed prEN 18356-1, a data model and API-based interface specification for the exchange of insurance contract data, primarily to support the implementation of the EU Financial Data Access (FIDA) project. prEN 18356-1 defines the semantic structure and API schemas ("FIDA Schema") for retrieving policy and contract information across several insurance product lines (e.g. motor, property, liability, personal accident, legal expenses, travel, financial loss, life and pension insurance).
While prEN 18356-1 focuses on the core use case of contract/policy data access between data holders and data users, market feedback and practical implementation experience indicate a growing need to describe, in a standardised and best-practice manner, how the FIDA Schema can also be used to expose insurance data to AI-based systems and applications, beyond simple contract inquiry.
The rapid adoption of Large Language Models (LLMs) and AI agents in customer service, intermediary support and back-office automation has established the Model Context Protocol (MCP) as a de-facto open standard for connecting AI systems to external data sources and tools. MCP defines how an AI application ("MCP Host"), acting through an MCP Client, can discover and invoke standardised "tools" exposed by an MCP Server, which in turn accesses underlying business APIs and data.
There is currently no harmonised guidance on how insurance-specific data, such as modelled in prEN 18356-1, should be exposed through MCP tools: which use cases are relevant, which naming and description conventions should be used so that AI systems reliably select the correct tool, which input/output data structures apply, and which business rules, consent and security aspects need to be considered. Without such guidance, individual market participants risk developing incompatible, proprietary MCP integrations for insurance data, undermining the interoperability objectives of the Regulation and of prEN 18356-1 itself.
This proposal aims to develop a Technical Report that documents, at a semantic and illustrative technical level, a set of representative insurance use cases (e.g. contract/policy retrieval, policy change processes, claims-related information, customer and consent-related queries) implemented as MCP tools referencing the prEN 18356-1 data model. The deliverable willprovide a common use case template (description, actors, input data, output data, required data fields, business rules, security/consent aspects) together with worked examples, to guide implementers, AI solution providers and insurance undertakings in building consistent, standardbased AI integrations.
The work is intended to remain primarily at semantic level, referencing prEN 18356-1 data structures, while providing a light technical layer (example tool definitions, request/response samples) to make the guidance directly usable, without prescribing a specific technology stack beyond MCP as the illustrative protocol.
Note: in case the WI is based on documents from other organizations than ISO/IEC, please specify it here
This proposal builds on and references prEN 18356-1 and is not based on documents from other organizations than CEN.
Required form fields are indicated by an asterisk (*) character.
You are now following this standard. Weekly digest emails will be sent to update you on the following activities:
You can manage your follow preferences from your Account. Please check your mailbox junk folder if you don't receive the weekly email.
You have successfully unsubscribed from weekly updates for this standard.
Comment by: